Intel's Hallock On Arrow Lake Core Ultra 200S Processor Performance Fixes

[@Hothardware]

More details here at HH! hothardware.com/news/exclusive-intel-promises-arrow-lake-fixes

[@Gindi4711]

Yes there is a 5% difference we saw between 285K gaming numbers and official Intel data and it is good for Intel addressing that, but I do not think this is the main issue here. ARL problem is that it does not provide a performabce uplift over 14900K when 9800X3D is already 30% ahead. Not getting slower and using a bit less power is not enough. Maybe you get another 10% with OCing ring bus etc., but that voids my warranty while 9800X3D is still 20% ahead at stock. Regarding SMT: ARL MT performance is heavily carried by Skymont which deliver a lot more performance while not increasing in size relative to P cores. But how did removing SMT from P cores help in that regard because the P cores are still not great in both area and and absolute performance and E cores seem to not have needed that extra die area.

[@thetheoryguy5544]

I think Windows core scheduling is just not handling these cores right.

[@lePurpleDragon]

@@thetheoryguy5544 100% Intel's fault, then.

[@Tdisputations]

They really just need a separate cpu to please gamers. Sounds like they may have something coming in the future with extra L3.

[@Hothardware]

We've been told that the fixes that are coming will be significant, more than just a few percentage points. We can't say more than that, but stay tuned.

[@BatuhanGuvenc-eq2em]

@@Hothardwarethank you for info , if Intel make me dissapointed again , i will not forgive them. Higgins expectetions again

[@ARiverSystem]

This was a really good interview, thanks. First time seeing Robert Hallock, and it was nice to see someone obviously knowledgeable comment on a lot of these things.

[@MarsMan1]

That was a great conversation. Very helpful.

[@Hothardware]

Thanks for watching!

[@dtsdigitalden5023]

It's a little strange seeing Robert speak on Intel's behalf, yet he brings his usual integrity to the game no matter which team he's playing for. He didn't share much, but his demeanour and attitude are reassuring. He's a good front man for the company, and we can only hope those behind the scenes are taking notes too.

[@coffee7180]

hothardware is the perfect title to describe intel cpu´s.

[@coffee7180]

@MasterChief-m3m stop crying about it, it's not your mom we are talking about.

[@coffee7180]

@MasterChief-m3m how am I, a bad person after you insulted my family. I am just telling the truth, and I have Intel stocks, hoping they will do better. you should stop feeling insulted when people make fun of big corporation, who don't give a fuck about you or your family, grow up.

[@perhansen3959]

What a great interview Thanks !.

[@JYMBO]

TechLinked (LTT) brought me here!

[@DJTempa]

Haha just saw that, came straight here to see the whole thing ;)

[@Hothardware]

​@@DJTempa Hit thumbs up and subscribe! :)

[@Hothardware]

Hit thumbs up and subscribe! :)

[@j340_official]

Even if there are performance uplifts the question is why is there such a disparity between Intel’s internal testing and the findings that reviewers have ? Quality assurance and quality control is down at Intel. We saw it with early firmware versions of Maple Ridge, which had significant connectivity problems, lack of hot plugging for various devices and so forth. We saw it with raptor lake. Why did it degrade why didn’t QA/QC catch the bugs in microcode? Why didn’t the silicon design teams catch the susceptibility of the raptor lake core to this vmin shift issue? What scenarios are the internal teams testing and why do the scenarios appear to not reflect actual retail motherboard settings from RoG MSI AsRock Gigabyte? Does Intel even test its CPUs using retail motherboards and their default settings? Intel’s standing is declining it is no longer being seen as a premium desirable brand anymore. I don’t trust Intel. And I have been a fan since 286. Intel you need to regain consumer trust by releasing solid products that work that don’t have these teething issues. I think your decision to change sockets so often is part of the problem. All the bugs you found and fixed with raptor lake and z690/z790, you throw all of that work away to start on LGA1851, only to encounter bugs all over again. Why do it? Why couldn’t arrow lake, which consumes less power than raptor lake, why couldn’t arrow lake run on lga1700, and benefit from 2+ years of experience and debugging ? There’s no way I am spending money on buggy lga1851 to deal with a cpu with performance regressions and no guarantee that lga1851 will support newer Lakes like Nova Lake. Intel, arrow lake is so bad it is making me consider AMD 9800x3d for gaming. I have never built an AMD system but i might have to abandon team blue and go Red for a while until this is sorted.

[@shvrdavid]

The biggest difference I have seen is the memory used in many reviews. DDR5 is considerably slower than using CUDIMM, by a lot. Most of the reviews I have seen use the same speed DDR5, 6000 or so. Intel would have faster CUDIMM to test with, and most stores don't even have part numbers for them yet... Also keep in mind that gaming is less than 5% of the cpu desktop market.... Multicore with CUDIMM on a 285 basically has no competitor, anywhere, even with the present problems it has.... This is no different than when AMD went tile based and has massive issues either.... The AMD 9000 series hasn't been at all it was cracked up to be either..... The Ryzen 9 9950X has an inter-CCD latency of 180 ns which is more than twice as much as the 76 ns of the Ryzen 9 7950X. And people are complaining about less latency in the 285, while completely ignoring the fact AMD has an even bigger issue with multicore......... This is an entirely new platform, and Intel knows there are fixes to be made.... And they are not hiding that fact either..... A Has AMD talked about the 9000 series latency issues yet?

[@thephantomchannel5368]

@@shvrdavid Considering that Intel is pushing their new platform as "production" rather than gaming I would find it interesting if somebody would bench these against the lowest tier Threadripper. While AMD Threadrippers cost more they also provide a huge advantage not only in multi-threaded workloads but also have I/O bandwidth and connectivity that makes both AMD and Intel desktop parts look more like mobile extreme rather than actual desktop platforms. The cost of the new Intel chip is ludicrous for how gimped it is other than a handful of special use cases, which again the Threadripper will outperform. What we need is a return to the HEDT platform to replace desktop at a price around the same as AMD and Intel's highest end desktop chips. Gaming CPUs need to be cheaper and be paired with platforms more along the lines of "console extreme" or "mobile extreme" for small form factor builds. As it is, both Intel and AMD platforms are gimped for having only dual channel memory controllers which makes most motherboards useless for having four DIMMs which can't run at full bandwidth if all are used. Desktop is also gimped for not providing enough PCIe gen5 lanes direct from the CPU, they should have at least 32-48 otherwise there really isn't much of a point in having motherboards as large as they are when most will just use one PCIe x16 slot for a GPU because if both slots are used they default to x8. To your last point about AMD latency issues- the new Epyc CPU's use a new I/O die which will be used in the Ryzen6 as well as the new Threadripper's whenever they come out. The Epyc's also have memory speeds getting close to what is offered on desktop which hasn't happened yet as far as speed parity of workstation/ server parts with desktop. If AMD or Intel can come out with something to compete with the Threadripper but also be decent at gaming for around $500-700 there will be a market for such a thing.

[@RobBCactive]

Then there's bending CPU IHS's.. Everytime I mount a 3rd party cooler I compare relatively simple instructions and mounting with AMD while Intel has often far more complex arrangements and many different options. Again a result of over tinkering for the sake of changing things, without a real cost/benefit analysis to customers. The ARL LGA1851 had plenty of time, ARL was late, MTL was late and the desktop "S" version cancelled because it performed poorly compared to 13th gen, with limited power envelop where it had better efficiency (high end laptop). I can see moving to tiles needed a new socket from the monolithic, but I don't agree that any of it was bold new development not based on years of other projects eg) Ponte Vecchio & MTL, effectively the excuses are saying Intel's not competent at design, they pushed AVX512, then E-cores monolithic, now lots of tiles and they had years of delays to polish this stuff but they delivered a real Richard and so the reviews stank.

[@shvrdavid]

@@thephantomchannel5368 Not sure if you know this, But CUDIMM has already been proven to work with the 285 at crazy fast speeds. Also keep in mind there is another generation of chips being left out of the discussion here. And I notice you didn't compare the Epic to the new Xeons either.... If you want high end, buy high end. Most people don't understand that running 4 memory channels is great, but it isn't always twice as fast as two, its not even close to that.... Even still, Intel has made chips that support 4 channel affordable since Haswell E, 10+ years ago. The twice as fast, is theoretical... Like kids that clean their rooms, it doesn't always work out as planned. My setup I am typing to you on, presently has 3 gpus in it (4 if you count the disabled one in the cpu), and I am well aware of how little difference X4, X8, or X16 makes with limited pcie lanes. The difference is negligible, at best.... I can play cyberpunk or any other newer game at 4k, with 3 gpus on the motherboard running at 4x on each one. It only uses one of them to play the game, and at 4x.... It usually stays at the refresh rate of the monitor, which is 60 fps.... I have no need for more, the monitor wont display it anyway. No one with a 120-144 fps monitor needs 500 fps either, no one..... Most people can not tell the difference between 120 and 500 to play the game on a good monitor either.... When I compile models, program, use the gpus for compute, etc, having 4 true 16x would show an improvement, but only if the rest of the system was overkill in a way that it could feed it that fast.... In that scenario, to show massive speed differences you need one processor per gpu before you see any great improvement... People seem to forget that ram on the system, still hasn't reached 100 gig a second. And every time you divide that band width, it falls drastically faster than you might think due to refresh times that can be in the hundreds of clock cycles between reads... More channels doesn't always help as much as people think it does... Especially when filling gpu memory.... Transferring memory to anything, only occurs after the refresh cycles are done.....

[@thephantomchannel5368]

@@shvrdavid I am not comparing Xeon or Epyc parts to consumer/ prosumer grade parts. I mentioned the Epyc platform to illustrate that AMD has a new I/O die that is proven to be faster than what is currently used on Threadripper and Ryzen. All I am suggesting is that either Intel or AMD need to bring back HEDT to a market that doesn't need to spend more than a thousand dollars for productivity and if it can be decent at gaming then that is a bonus. A scaled down Threadripper like system from either vendor would have a demand if they could price it similar to what AMD and Intel sell their top tier desktop parts for which in my opinion are overpriced for what they offer. As I stated earlier, current AMD and Intel desktop is more like a "mobile extreme" platform rather than a true productivity platform. Sure you can force them to operate with multi-GPU's but at what cost in performance especially going forward when GPU's will actually start to saturate the PCI-e lanes. I agree that systems need a lot more R&D to get the I/O in a place that can actually feed data to other parts of the system with more speed and lower latency. I liked the idea of Intel Optane for what it was able to do compared to the average NVME drive. I was hoping that technology could be carried over to memory where a system could have a separate single or dual channel memory controller which would allow a hybrid non-volatile drive on DIMM so that your OS drive would be carried by the memory controller rather than through the PCI-e sub system. After that I would have liked to see the system memory go for a true quad channel separate from the single/ dual channel controller and have RAM that would be more akin to HBM memory on DIMM which would vastly improve bandwidth and latency of RAM which in turn would feed more data faster to your peripherals. These are of course pie in the sky fantasy scenarios but with proper research could be a possibility to achieve especially if graphene were to ever to be implemented for electrical signaling/ paths. The cost of course would be a lot more than what I am suggesting should be the target for an HEDT system at 500-750 dollars but eventually as technology improves the cost will come down just like how CU-DIMMs will cost a lot for first gen but once adopted on multiple platforms will likely replace U-DIMMs with a price parity about the same once it matures.

[@jonathanjones7751]

That smile when he was talking about V cache. We have rumors about adamantine cache. INTC has unquestioned leadership in L2 cache (hence its productivity performance. If you need an example look at the cache structure of IBM power mainframe chips it’s all L2). Throwing 100MB of realistically L4 cache off the Compute die would be crazy. As a nerd I got to play with Xeon Phi back in the day. It had 16GB of on die memory. Programming headaches aside, effectively using that memory (I was writing in C so pray for me) MASSIVE performance increase.

[@eg9574]

Talking about SMT being a relic from the past while presenting a new product performing like it’s a relic from the past, simultaneously talking about how this helps them keep costs down for their product line while charging premium for all parts and coming a distant second blows my mind… intel needs to wake up.

[@RobBCactive]

Especially as the BETTER example of "waste of space" for the majority of desktop consumers is E-cores. HT was justified based on power efficiency gains, but you don't get to run 2 threads on a P-core often when chips have a crapton of hungry E-cores screaming "pick me!" to the Windows scheduler.

[@lePurpleDragon]

They don't need to wake up, they will continue to drown themselves while a couple hundred Incels worldwide buy this like they buy NVIDIA's gimped trash to run shiny shadows and reflections at 720/1080p to satisfy peer pressure and nothing else. Raptor was the greatest PR disaster in their history until Arrow Lake. Arc Alchemist is entirely bested by both NVIDIA AND AMD and Arc's less than 1% market share fell off to "undetectable", so there is no way Battlemage will even put a dent in anything. No one bought this. Look on the sites where it's sold. There's less than a dozen reviews in 3 weeks. Look at mobo reviews. There's like 2-3 dozen reviews for all of them and that might be the same people mobo hopping to find one that works. That's one of the worst paper launches in tech history. The thing shows 80-100ns RAM latency, the worst RAM latency in tech history. It sinks down the list of UserBenchMark, so even Intel's #1 fan can't make it look good. Most of the time when it loses to Intel in benches, it uses as much if not more power. Meaning even if they somehow optimize this mess it will still not be more efficient than AMD and the RAM latency means lows% and input latency will suffer, no matter how many PR shills/bots comment saying it's "normal". Already, input latency on everything RKL and newer is worse than AMD, because no speed runners, the clearest knowers of input latency, use Intel's newer platforms. They're all on Haswell, Skylake, AM4, AM5. Where are "competitive Raptor Lake gamers"? They don't exist. Even if cores/voltage are not the issue for someone, no one has addressed the faulty IMCs on Raptor Lake, and that only gets worse with Arrow Lake. Intel hasn't had a good IMC since Comet Lake. They redesigned the thing so it shoves more bandwidth through for benchmarks, but doesn't actually perform better... Why does Intel need 7200-8000MT/s to be on AMD's level at 6000-6200MT/s? Had that been the other way around, AMD would be chastised to the grave...

[@lePurpleDragon]

​@@RobBCactive E-Cores are the most moronic and nonsensical desktop design of the 21st century. Just fix the regular cores. Intel's solution is to jam voltage and wattage through "big" cores for any detectable load, and impose ridiculous amounts of latency upon the system/user to use atom cores the rest of the time or mix the two (obviously bad idea). And what's the alternative? Turn them off? Now you have no SMT which is always a disadvantage (maybe not on Intel because their implementation sucks, read: DIDN'T WORK on 10nm, so it "only adds heat"). A desktop is not a laptop. Real engineering would make a hybridized core like AMD, that can idle/sip little to no voltage and scale up per load. E-Cores demonstrate Intel's inability to engineer. Intel does not have modern power delivery and proper clock stretching so instead they shuffle back and forth between netbook cores and oversized desktop cores. Talk about a real "BOOMER" design... No SMT = oops, we can't get it to work at all so let's just shelve it and use the big bad cores as an excuse. SMT is synonymous with multi-core; it is a 21st century (for the average user), breakthrough technology no one rational wants to see removed, only improved.

[@SaschaRobitzki]

Please add an additional 8 E-Cores to the 285 and put AVX-512 back.

[@EbonKim]

Do they not do internal testing on their CPU's? Mind boggling that they can be surprised with reviewer results.

[@EbonKim]

On voltages/frequencies, if you're overclocking, temps are important (staying decently below max temp), but also the amount of voltages you're pushing - what the CPU is capable of, Motherboard VRM's, PSU quality. Basically, set a frequency, and set voltages to Auto, set proper LLC, lax RAM freq and timing, test for stability, then lower voltages until you encounter instability. If you're undervolting, then you set the voltage you want, first, lower the frequency, and work your way up until you encounter instability.

[@bandanaocelot45]

I swear Robert has been the face bad cpu launches or poor AMD GPU performance for the last 15 years. I remember when he was forced to put on a brave face the AMD FX CPUs and first generation Ryzen.

[@SaschaRobitzki]

Any idea why Asus pulled the bios updates with the new 112 microcode update for their Z890 motherboards from their support pages?

[@Superior85]

Looking forward to the performance fixes, I wonder how much performance can be clawed back. Hoping things turn out well...

[@Hothardware]

We are hearing the performance gains, after a series of BIOS and Windows Update fixes are in, will be "significant."

[@RobertSmith-lh6hg]

Hallock is the guy you bring in when you need to backpedal into last century

[@NAMOR5000]

Two thumbs up! I absolutely loved the interview with Robert Hallock and the transparency from Intel. It made me smile from ear to ear. Now, I just need to get my hands on the Ultra 9 285 to start my new build. It will be great to have all the non-chip issues resolved so we can experience what Intel’s engineers have been working on. I’m also looking forward to CES 2025, as I have a strong feeling there might be more exciting announcements. I’m glad to hear the chips are performing well, though they are a bit pricey. Hopefully, the prices will come down soon. Thanks again, and have a fantastic weekend!

[@RobBCactive]

You better be patient, people who need to do work now are just buying the 9950x. What planet are you on?? Intel used TSMC 3nm and have their *bleep* handed to them by cheaper 4nm "glued together" CPU range. Intel QA had let through CPU bending sockets, over-volting in apps that memory stall, failed to validate partner boards leading to a wild west of default over-clocking.

[@Hothardware]

Thanks for watching!

[@Valk-Kilmer]

Why the fuck would you buy a 285K when far better options are available from AMD? Utter madness to go Intel. And don't say 'muh productivity' because the 285K will be further obliterated by the 9950X3D.

[@NAMOR5000]

@@Valk-Kilmer Why do you care? Ain't none of your business. I don't owe you explanation, to each his/her own. Live and let live.

[@Valk-Kilmer]

@@NAMOR5000 Because I don't like to see people wasting their money only to regret the purchase they made.

[@alseick]

26:43 first week of December is over, failure after failure

[@Core2]

Absolutely love the interview! We can clearly see that these CPUs have performance left on the table.

[@impuls60]

I seriously DOUBT that. I've seen ALL the oc vids from Derbauer and Skatterbencher. Insanly indept oc vids and theres no fix for this this cpu.

[@Core2]

@@impuls60 Come on, as much respect we have for these great guys you mentioned, we're talking about Intel themselves, the engineers, the people who actually designed the CPU, the people who made the CPU, the people that knows every detail about it and have all the tools one could dream of for debugging, if someone can find missing performance, it's them, at the very least you should give them the benefit of the doubt and not go all in 100% on a opinion based on just some youtubers (with all due respect to these great youtubuers you mentioned of course).

[@Hothardware]

Thanks for joining us!

[@Hothardware]

@@impuls60 and you'd be wrong. That's exactly what Robert was detailing here. There will be BIOS and Windows Update fixes that will fix all this and return gaming performance to expected levels.

[@NAMOR5000]

@@impuls60 You forget that they were working with errors Intel said needing t be fixed. So no wonder they got bad results. Lets see in a few weeks how or what. No one company is perfect, not even AMD, not even Derbauer lol.

[@stevensalter9697]

I still have a i7-10700K, Will look for next cyce product now, Nova Lake. I was hopeful for Arrow Lake. =\

[@TheAkecheta]

I have the same CPU, which has been good to me for over the past 4 years. My plan was to do a build with 265k, but going with 9800x3d, which is my first AMD build after 20 years.

[@ThisBytesForYou]

Great interview guys, I loved it and Robert is awesome as well, gave tons of great information

[@Hothardware]

Thanks for watching!

[@Kapono5150]

Intel did this to themselves and they have lost the trust of DIY. I bet they won’t try to blame Motherboard venders again and shame on you for taking 6 months to come clean to your customers. Put one finger in the air for Intel bc they are screwed

[@RobBCactive]

Oh it was more like a year, there's data center people who had ALL their 13th gen replaced with 14th due to issues they had with all their CPUs going bad and the 14th gen failed even faster within months.

[@MathieuPrevot]

At 39:00 still no word about the NPU and the benchmarks which are not taking into account this compute power, or the impact on AI apps/computations. You had an Intel AI PR with you. This is not a well prepared interview.

[@TAH1712]

Looking at CPU pass benchmarks, Nov 9th, I see very encouraging and leading scores all around for Arrow Lake. I just work with 3D CAD and FEM plus making content videos. Whatever hasn't gone to plan, I'm 100% sure Intel will do their best to maximise it's potential. In weeks to a couple of months, they'll be improvements all around. From today's Pass marks, there's nothing honestly stopping the purchase. The cheapest 245KF has a ST score 4790 whilst the 14900KS has a score of 4869 and the 14600K is 4282 - to me, that's a big generational change and all at less power. The Ryzen 9950X ST score is 4734 which is lower than the 245KF at 4790.

[@ssaini5028]

Yeah but nobody cares about benchmarks or synthetic workloads

[@ferdynandkiepski5026]

One thing that should stop your purchase is that 3D CAD and FEM are extremely parallelizable applications. A 9950X wipes the floor with the Intel CPUs in these workloads. Passmark is also completely synthetic. Moreover 3D CAD and FEN can benefit from avx512 which Intel decided not to include in their desktop CPUs since a few generations. The same with making videos. If the program is properly optimized AMD will be faster. Don't base you decision on synthetic benchmarks. Keep in mind that AMD might also get an uplift in performance as zen5 support isn't yet in compilers, so they are basing their cost tables off of zen4, and they have changed as now avx512 isn't double pumped.

[@TAH1712]

@@ssaini5028 Surely not nobody. I can understand that if you have a particular application or two in mind, and it's publicised what is best for that, then any synthetic test is second best - that I get.

[@RobBCactive]

Try real applications, nobody is saying Intel has't produced CPUs that score well in benchmarks.

[@TAH1712]

@@RobBCactive Thanks for your thoughts. I'll have to ask the application support help for their recommendations at the due time. Whilst taking onboard what you've said, that doesn't explain the general derisory remarks about AL - it just suggests AL, applications, OS and Bios aren't yet in sync or there's some conflict going on as this video suggest. OK it's like a yoyo as to who's the current leader - for what I do, I won't notice a few percent off the best as thinking takes most of my time.

[@CarlosDeBernardMrAgapi]

fantastic interview. Thanks Robert and HotHardware guy! i did have the same question as you did. i know was not memory related. the what how and why will know later. the why every single bench was so high and what was the issue FPS was not. those are the Q and A will know later. For what i do those number are what expect to be. We wait... good luck intel i know is not easy. the rest i, as you, what ever is the outcome we wait. for what i see.... this is just the start

[@JoyTrader999]

I purchased Intel Core 9 - 285k with Gigabyte Z890 Aorus Master motherboard and it's an absolute mess Lots of driver issues, BIOS issues After days of tweaking, the only way I can run my new desktop is by disabling my inbuilt Graphics (CPU), disable RAM XMP Profile and run my memory at 3000 MTs instead of 6400 (My MB can go as high as 9600).. And still my mouse shows noticable LAG... Totally unacceptable.... I wouldn't NOT recommend purchasing this Intel new series And guys don't get scammed by these benchmarks.. These all are staged or tweaked in a laboratory like setup Instead go with AMD 9000 series and you will have none of these issues... And pay special attention to Robert: You'll notice that he has to spend extra time to careful craft his verbal script (To ensure he won't accidentally spill the beans)

[@RobBCactive]

Oh dear!! Reviewers were saying hard pass until Intel sort the platform out, the AM5 has efficient options, has matured and has a future with Zen6. But there might be a surprise 16c/32t with V-cache as the 9800x3D posted some surprisingly promising results Laggy desktops has plagued the P&E architectures but doesn't show in benchmarks.

[@Hothardware]

There's a fix incoming from Intel on all these issues, in the next few weeks, early December latest.

[@NAMOR5000]

RESPECT! 😎. Each CPU has / had its issues / challenges and everyone has a choice, which are personal and tailor made to his/her usages. I find it sad and funny the hate that people spew without knowing the entire in and outs just repeating what is already said by some. Being judgmental, a sheep, no disrespect intended, is all ok, lets agree to disagree and still love each other as PCDIY people. It is sad IMHO that some people do not see and hear what this stream was all about, and the direct ownership and transparency was offered and taken almost instantly. One forgets that AMD had similar issues and those also got fixed. I love both AMD and INTEL and future to come CPU companies. It is all about pushing and overcoming challenges for consumers. The non consumers are light years ahead of these 'Toys' lol. I am looking forward to the next few weeks once the fixes are out to see for myself how or what. PEACE 🤛

[@iceresilient]

"Must be a bad motherboard, because I'm running an Intel 265K with the MPG Z890 EDGE TI WIFI and 6400 DDR5. It is very fast with zero lag. "Try clearing the CMOS. It might help with some of the issues you are having."

[@RobBCactive]

@@NAMOR5000 nope it's not hate it's based on what people saw and observed. There were reasons Meteor Lake desktop was cancelled and Core Ultra also followed Sapphire Rapids, Intel had PLENTY of time to sort things out. Years. Their past platform featured QA problems like bending CPU heat spreaders, the wild west of partner motherboard settings being out of spec by default to inflate benchmark scores, the overvolt on code that was light on the CPU but stalled on memory. Intel have a single shot platform requiring expensives motherboards, their CPUs are competitive in productivity but catastrophic in gaming. That is NOT hate that is recorded observed facts. Everyone should wait and see if Intel do fix the product, but how they handled 13/14th gen does NOT inspire confidence. Intel remember lost a lot of performance due to poor security in their chips, features which they sold and chips they sold with performance features. Eerily similar to the infinite power settings and turning off of chip management features, to juice benchmarks on motherboards.

[@yellowjoe.2000]

Is Arrow lake core ultra 200h 3nm TSMC and is it big improvement battery life over Meteor lake H?

[@JudahRichardson1]

Honestly anything that isn't CPUs failing is a dub rn. Fix it in post and get the next launch right.

[@LeeSeanSullivan]

Robert comes across as a smart and genuine guy, hope they get it all worked out.

[@lePurpleDragon]

Literally, he's a marketer, which means a professional BSer. So F for judgement for you.

[@thephantomchannel5368]

Both Intel and AMD need to lower the prices for their desktop platforms for how gimped they are as they are no more than what should be considered "mobile extreme" or "console extreme". What we need is for HEDT to replace desktop and offer quad channel memory controllers and at least 32-48 PCIe gen5 lanes direct from the CPU and cost no more than $500-750 which is what the current top end desktop chips cost. As it is there is already beginning to be a disparity between gaming and production and both Intel and AMD desktop are woefully inadequate when compared to AMD Threadripper for production workloads. If either Intel or AMD could come out with a "Threadripper lite" HEDT that could do both gaming and production I could see there being a market for that kind of a platform. Gaming only chips should be relegated to small form factor builds or medium sized desktops at the most and be priced a lot lower than what they are now.

[@SaschaRobitzki]

Totally agree. The higher-end Z890 and X870E motherboards don't make any sense. Why spend so much money when you can just plug in these limited gaming CPUs?

[@AustnTok]

December 18th and still no 0x114 microcode on anything except the Taichi OCF. Neither my z890 hero or z890 unify-x have received 0x114 and I'm getting frustrated. Still can't even play steam games that have easy anti cheat.

[@Bass.Player]

What I don't understand is how ALL OF THESE reviewers found what Intel could not find.../????

[@knofi7052]

I had an instable 13. gen. cpu from Intel and their support didn't help at all. Intel, never again!

[@jjlw2378]

Is this true? I had a perfectly stable 13900k since Nov 2022, and Intel still gave me a full $599 refund after only a couple of emails back and forth.

[@knofi7052]

@@jjlw2378 It's true! Maybe I live in a different country or had just bad luck with support...

[@julienlauzon1106]

amd is a mess too dude.

[@GEN9100]

i still think the lack of smt is a shot in the foot.

[@twistedandy]

Arrow Lake S looks pretty confusing due to a much higher memory latency and performance degradation in some games. Things get really bad considering that there will be no new SKUs available in a year or even two. I do not understand why Intel decided to reuse the SoC tile and the layout from Meteor Lake, knowing that it shows pretty bad performance. I've been using Intel CPUs since 1998 and now considering switching to AMD, because it's simply faster.

[@TitanPlakInside]

I saw the die shot of both MTL's and ARL's SoC tile and their layouts look similar.

[@ivankintober7561]

@@TitanPlakInside its lunar lake soc...since it use same core layout...every 2 p cores ...there 4 ecores ....they duplicated with that tile to get through every 2 pcores ...i believe ecores are never the problem...they prove doing lot more than p cores doing....OC more and gain more fps ...its pcores slowly down ecore cluster...cause ecores function as a cluster while pcores don't function in any cluster...its possible they can phase out the pcores sooner the better since ecores are becoming much more mature doing bigger work loads now than previous architectures ....royal core architecture is all ecores only ....24cores and 24T....all function as a group

[@teej_youtube]

Great interview

[@mikebruzzone9570]

The "FTC says tech road maps cannot be disclosed?" Well, nothing in Docket 9341 about that, or that I am aware from the SEC. AMD and Intel, others provide road maps all the time that for decades a compliment collaborative prerequisite. In fact a best practice. Mr. Hallock said he did not know and then "fair disclosure" right a material disclosure. Mr. Hallock knows Intel is not ready to respond that is the answer. Mr. Alcorn, Intel side fixes? Well, what about the software compliment optimized fixes? Applications rewriting, recompiling and Intel provides tools for application developers holding their hands out, even extorting Intel to be paid to do optimization on Arrow nascent seats. Media application labs knew and said noting FK THEM. Mr. Hallock why did you not address that fact as old as every new architecture specific OS, RTOS, RTOS think dGPU subsystems and the actual applications. It is very clear all game devs did not recompile for Arrow waiting seats. By summer 2025 there will be plenty of seats. Arrow is what OEMs ordered for enterprise IT; power optimized, 1 HT system security, warranty assurance. This fact of cooperative system design does not happen in a vacuum. Oh, "not the right set of software firmware", you mean with the application and performance guides Intel provided to application software developers did not include in that kit the new Arrow application optimized compiler tools? Yes, firmware needs to be whole entering the hardware application re-compile to optimize applications. Yes, the raced product launch never let Intel PR send OEM specified enterprise part to game gangs before the game applications have been fully optimized. You always start with the application targets and then expand. HH channel bros as of week ending 10.28 Arrow supplied 285K = 20%, 265K = 67.6%, 245K = 12.4%. Supply my specialty all the time and my Arrow full run estimate by quarter is posted at my SA comment spot I will make it easy PLUS a LOT more data than this; q3 2024 = 4,739,866 q4 = 14,325,127 now 19,065,013 the same as AMD R9K second half 2024* q1 = 40,588,460 q2 = 20,927,749 Total = 80,581,202 units Definitely visit my Seeking Alpha comment spot for more data even Mr. Hallock and a lot of employees at Intel will learn something that is why my regular viewers exceed HH subscribers, it's only the whole industry, so as I follow all your news follow mine and then we will all keep up. My bet is P-core branch prediction on targets and like Raptors game devs and commercial application developers not utilizing SIMD using the e-cores as a parallel array but Mr. Hallock would know more than me on those questions. HH its none optimized applications waiting for seats and I heard no address but that's it. Soft on application developers extorting for payola to optimize. I can imagine from game devs especially "we optimized for Skylake +++ and then recompiled for Alder Raptor P/E and now we have to rewrite and / or recompile for Arrow, pay us. But that is not Intel's responsibility in the whole platform hardware software collaborative compliment that is not always collaborative on who takes responsibility for what. You can see from all the benchmarks Intel took responsibility for the canned application and performance synthetics but that not all applications. Right OEMs thinking about commercial sales in q4 so Mr, Hallock never seed gamers when the application target is enterprise until enterprise is established and game gang application optimizations are complete. Arrow is not the first Intel disaggregate chip Mr. Alcorn by a long shot. Recall Haswell, Broadwell, Kaby G MCM. Now Sapphire, Meteor, Emerald what did I forget in SIPS. In the reviewer community, on my framework I would have tens and in their following hundreds of real users with no ax to grind sending platform, AIB, application and performance data into the internal working group months before the launch to insure when Intel Inside paid media said r product was less that it was I COULD SIC the ENTIRE WORLD no, but a substantial constituent base of savvy beta validators onto ZD PC WEEK PC MAG, IDG PCW and others CNET came late basically informing them to get their bund off our back or FK OFF. Mr. Hallock media gang land can be regulated which takes bottom-up market based user preparation because ultimately the audience of gang land calls the shots. mb

[@RobBCactive]

Yep, ARL was developing on MTL .. Intel been doing Ponte Vecchio for years, are we to believe nobody talks to each other at Intel? Then there's all the QA issues, IHGs being bent, the wild west on default settings being out of Intel specs, over-volting in anticipation of memory stalls clearing and getting load, the Arc debacle. This launch is part of a trend, NOT an isolated misfortune caused by over-ambition on a project with little time and investment behind it.

[@BigEightiesNewWave]

from Level 1 Techs on why Arrow lake is not as good as it should/could be..."Intel needs user land hooks in OS schedulers in Windows, it is already a thing in Linux. APO is a FIRST STEP, but SHOULD HAVE BEEN THERE three generations ago! 'It's not just a scheduler problem though, it's different geometry in the system" Linux is our canary in the coal mine, to see how much Windows can improve Arrow Lake performance. Intel clearly has the memory controller competetive advantage. If next-gen P and E cores are even better, Intel will have a product to be reckoned with.

[@squatch545]

So Marco Chiappetta doesn't care how much power his desktop uses? I guess he doesn't have to pay for electricity like the rest of us plebes.

[@ETophales]

Most people don't care. Or don't care much at least.

[@squatch545]

@@ETophales Speak for yourself.

[@ETophales]

@@squatch545 The point was that it's most people. If people cared about power efficiency, they wouldn't have bought Intel CPUs. The fact that they have bought them means that they don't care. High end GPUs such as the 4090 are also quite successful. People are simply willing to pay the extra cost involved in this. You might be too poor to afford the electricity, in which case you likely buy low end hardware anyway, but that doesn't mean that you should apply that to others.

[@squatch545]

@@ETophales Nonsense. The new Core S series is more power efficient than the previous two generations. But before that Intel CPUs were also fairly power efficient. The fact that anyone buys he new Core S series means that power efficiency is a factor in purchasing. Especially when you consider that Intel idles at a lower wattage than AMD. I live in a Canadian province with the highest electricity rates in the country. And I buy high end hardware and look for power efficiency when I am choosing. A computer spends most of its time in an idle or near idle state, unless someone ONLY uses their computer for gaming or video editing/ 3D work.

[@ETophales]

@@squatch545 "The fact that anyone buys he new Core S series means that power efficiency is a factor in purchasing." How? If that was a factor, everyone would have bought AMD, which in many cases both performs better absolutely and is also more power efficient. Sure, it might be a minor factor, but you can't argue that people who bought 13th or 14th gen really considered that as even a minor factor. It's also a fact that over time TDPs went up for both GPUs and CPUs, which is a strong indication that people prefer higher performance even if it uses more power.

[@haisan]

Intel strongly recommend you to buy for gaming the new one Amd Ryzen 7 9800x3d !

[@RobBCactive]

Robert's argument about HT wasting space applies far more strongly to the problematic E-cores on CPU like i/U5, they're solving Intel die area and yield problems not solving customers' problems, so Intel can score better in productivity reviews. Independent tests have shown AMD SMT is alive and well delivering significant performance gains, even in the MT productivity apps on the 9800x3D. You get context free thread switching, parallel work and fewer significant security problems than Intel's implementations. Yes useful work uses more power, but still less than ARL and often SMT delivers large performance benefits in things like decompression, where Core Ultra has taken another short term decision like E-cores and marketing efficiency when the ARL P-cores don't hit the previous frequencies. The Core Ultra S launch followed the MTL desktop cancellation, there have been prior tile releases like Sapphire Rapids and MTL, motherboard vendors were showing lots of ARL boards in Taiwan event when probably initially developed for the cancelled MTL desktop of 2023. So Robert's excuses about a new platform left out how delayed MTL & ARL desktop are, and that Intel's QA has yet another fail, compounding CPU bending sockets, the juicing by default on K-series with no clear answers from Intel for months, then the buggy microcode overvolt that follows years of seriously performance damaging mitigations on CPUs, at times requiring disabling of segementation features like HT people paid extra for. Overall there was a lot that could have been pushed back on, and we know from leaks the large cache Adamantine project was cancelled. Intel CPU server tiles have their own memory controllers, so unlike MilanX, GenoaX and TurinX helping defray costs large caches for gaming chips are seen as too expensive with Intel's Forveros large base tile.

[@impuls60]

Amen brother. They made so many bad decisions to make more profit instead of making a good product that people wanted to buy. Give us 12p cores with IMC for 9000 Mhz ram on a gaming cpu, and you can make a productivity cpu for the rest of the market. As a fellow engineer and enthusiast I saw the problems with poor performance coming from a mile away. If you turn off HT performance drops by 40% in the big online popular FPS games. Making up for that is a BIG ask. Putting IMC on different tile while a huge cache was a certain fail. What were they thinking over at Intel? Spinning a narrative that SMT is from a bygone era is just HILARIOUS when it kicks as for AMD. Intel PR bullshit is in full swing here people.

[@itstheweirdguy]

I don't know how it takes such a long livestream just to say that their new product is slower than their old product. And say a bunch of BS to sidestep the fact.

[@RobBCactive]

@@impuls60 They could not give a load of P-cores because of the bloated area, poor scaling to lower frequency and power levels, while also being expensive to produce as large monolithic dies. That's exactly why the i/U5 have cut down on P-cores and give the traditional gaming chip E-cores instead of cache. Then there's the sympathy, when we know ARL developed on MTL's cancelled desktop platform, the AIBs had a host of mobos for ARL on display this summer in Taiwan when Lunar Lake was introduced to the world. There's also Sapphire Rapids, which followed Ponte Vecchio, it's almost like nobody in Intel talks to each other.

[@RobBCactive]

@@itstheweirdguy well his job is to put the best face on the situation, so the talk is of a "reset". The real issue is the state of Intel's QA because even fairly succesful products have been dogged with problems like bending the CPU heat spreaders, that over-volting problem and the wild west of power settings by partners with NO validation, all incentivised to juice their boards for reviews with likely golden sample chips available to any reviewer who hit instability with their single sample CPU.

[@lePurpleDragon]

Intel cancelling the large-cache competitor was really the dumbest thing they've done. They seem to have some aversion to using cache correctly, as if they can't figure it out or don't want to win the gaming segment. It's been obvious since the 5800X3D that Intel needed to put all hands on deck. It's been obvious since they cache-starved Rocket Lake, which just should've been Tiger Lake had they not fetishized laptop production so much, that cache is the final thing holding them back. And they completely blew it. It's as if they are compromised internally and are doing nothing to stop themselves from self-destructing.

[@teej_youtube]

50:50 exactly what I have been preaching to all the AMD shills who would buy a 9800X3D and then still game at 4k lmao? This just shows people are just dumb and will believe what ever person is their favorite "beloved tech youtuber" with out ever testing the actual tech they are ragging on for themselves.

[@SonicBlueTyphoon]

Many of us play actually CPU demanding titles. I play Factorio and build megabases. It hits your CPU so hard I was at 30fps with the 7800x3D. We need better CPU performance generation on generation.

[@nossy232323]

So buy an old cheap CPU then, not these expensive new turds from Intel.

[@jjlw2378]

​@SonicBlueTyphoon Pretty sure Intel 13th/14th gen dominates Factorio with end game bases. 7800x3d is only good when game fits in the cache. Once it goes out to memory, X3D dies.

[@nossy232323]

@@jjlw2378 That's funny as the 7800X3D run Factorio way faster than 13th/14th gen Intel CPU's. It's literally almost twice as fast in that game.

[@jjlw2378]

@nossy232323 That's true in early game but late game it's the opposite.

[@jonaswox]

he wants to really stress that they are internalizing the mistakes and wants to fix the issues. So brave coming from a company , who has consistently been pointing fingers at everyone else while they were the problem the entire time for the past many years ........ It does not come off genuine whatsoever. It comes off as "oh now we probably cant tell that lie anymore so well try something else - what about this one; somehow our QC team did not realise that the chips was not working as intended whatsoever, so there is some discrepancy with reviewers". Its just plain stupid, so nobody at intel figured to try out a consumer motherboard on a retail version of windows , or what are you saying exactly mr marketing guy? Until every executive is fired from that company, im not looking at intel as a serious company anymore. Its just pure brand and profit margins. They are the type example of everything wrong with publicly traded multi billion companies. Its so funny how few years ago they be like, "yeeea we just gonna throw money at the problem until we come out on top GPU and benefits from pairing with intel CPU will pwn" , a couple of bad quarters and here we are - instantly the money looses trust in its own plan :D Like , Pats grand plan - a few bad quarters and its all thrown out the window............... and the copers start reasoning WHY it is smart - nobody said that back when the grand plan was presented lol. Its so obviously a shitty argument to cover for a bad situation caused by incompetence. And all people with a normal head knew from the start that the grand plan was just a tad optimistic. Now it is genius to trash an entire node, because the new node is so much better - nobody mentions WHY it is so much better, which is because 20A was so horribly bad that its not practical to mass produce. Its not a choice of genius, its a choice by force of ones own failure.

[@Loudlevin]

Intel has truly lost all credibility and thats coming from an intel fan since the 386 days. I doubt there will be any meaningful improvements going forward with arrow lake.

[@DreadyBearBoi]

We already know there is a ton of performance on the table just fixing the CPU. It’s never gonna be shockingly fast but it’s way worse than it should be.

[@AshtonCoolman]

BIOS and Windows 11 fixes will get it performing better. Intel confirmed this yesterday.

[@godnamedtay]

14900k/ks is the only way at this point. I do have a 14700k in my other PC that works fine but it’s i9 or bust if u want to compete for years to come. Arrow lake = trash

[@lePurpleDragon]

I switched to Intel 12 years ago and now I'm back. I left NVIDIA too, earlier this year, and am shocked how little I miss them.

[@godnamedtay]

@ terrible idea on both ends, truly stupid move. Unless u only use ur PC for gaming, then I hope u purchased a 7700x or 7800x3d, maybe even 7950x3d. Otherwise it was a terrible decision. Regarding the move from NVIDIA…I have no words. Ur just an unknowing and/or ignorant clown with that one.

[@TomTschritter]

With all due respect, it's clearly not at the BIOS or at the OS level; what happened here is incompetence. The problem isn't with the engineers, these problems result from antics in the boardroom and with the executives. So many arrogant and egotistical decisions made by those in power who now refuse to accept responsibility for their unethical decisions, for instance, this video is full of excuses. Spend capital on capital improvements, not more benefits for the CEO and friends. Why should the customers and the staff pay when it's clearly Intel's senior management that screwed up? People deserve much more since we've all paid so much already.

[@nevzataydin1]

they have to(!) release it before M4. they are pushing engineers, while it's clearly not the way as we seeing it at 14th gen

[@NZRanger]

I own an AMD 5700X and it does me well, but I was looking at Core Ultra as a replacement... I actually use and LOVE my Intel ARC A750 and may now wait and see what happens with Core Ultra and instead invest in a Battlemage first. Options are always on the table :)

[@godnamedtay]

A750 is like the worst modern gpu on the market. A770 or nothing. 8GB is garbage in 2024. wtf are we even talking about here?

[@Met1900]

With such honest workers intel gona get on top soon. This guy gives me hope next gen will be awsome.

[@florin604]

Honest 😂

[@thetheoryguy5544]

All new architectures have growing pains.

[@nourddinebakour2163]

Intel is the best technology company for future CPUs, but society now only looks at FPS, this perspective is wrong

[@thetheoryguy5544]

Because society has grown dumb!

[@PainIsntExist]

When I read the latest news about Intel like this, I get the impression that some great unholy conspiracy has been plotted against poor Intel - oxidation, degradation, and now this ..... Or maybe I just got too far ahead with my imagination and this is a video for April 1st. As you listen to it, I catch my head, Intel continues to make idiots of people. What kind of fixes? Or this system of theirs? Does this company test CPUs on their "own" Windows hybrids? For what it's worth, previous tests of older Core have already been done on this "civilian" system.... Going further, we have a corpo that turns billions of dollars - completely (even minimally) did not embrace optimization in time? Seriously? Tests show not some minor performance drops. There is talk of a chasm.... So what, they couldn't even get the performance of such a 285K at least on par with the previous CPU (14900)? People! What cheeky bastards are sitting there!

[@jjlw2378]

I've owned 6700k, 7700k, 8700k, 9900k, 10850k, and 13900k. I've always loved Intel OC, and Intel CPUs have always been superior. However, ArrowLake is none of those things. ArrowLake might have more OC dials, but none of them actually improve performance over 13th/14th gen. Luckily, I picked up a X670E-F/7800x3d bundle for cheap and have a 9800x3d coming in a few days. I hated 7800x3d, but perhaps OC on the 9800x3d will provide me the tweaking/performance increase Im looking for.

[@vorkzen4767]

What you think about you 9800x3D?

[@jjlw2378]

@vorkzen4767 It actually just got delivered about 45 minutes ago. I have quite a bit going on with work/family, so I probably won't get everything set up until after Thanksgiving. From what I've seen from others, it has potential. I'm excited to try out mem OC and all core OC.

[@vorkzen4767]

@@jjlw2378 can't wait, take you Time.

[@SaschaRobitzki]

Only 153 likes?

[@marktackman2886]

To see Hallock go from AMD to Intel combined with the trajectory of each company is funny. Talk about the grass being greener on the other side :P

[@edge8616]

Thanks for the Interview! I am curious how things develop regarding future products for Intel. Let's hope they trade blows with AMD, so that we - the customers - have a real competition at the market. What would be interesting for gamers -> a KS for each segment U9 down to U5 with utilizing 3D V Cache to enhance the gaming performance up to 15% while reducing the power draw at the same time. i7-5775C did a great job back in the days with its L4 Cache. After AMD brought 5800X3D, it would have been clever, if Intel introduced a counterpart at least in 2023 for Raptor Lake SKUs.

[@leeloodog]

Intel if you read this, lower power is preferred on my side. I've personal experience playing with my pc. It's just like with Raptor lake actually. I turn down the power might lose 10% performance and i turn it into something that's very efficient. Not everyone is like me; the way you explained that really resonated though. My biggest issue is the 285 is more what i need (for productivity) and the price is too high. The reason i bought Raptor lake over Ryzen was the price and flexibility with Ram. Platform upgradability is another thing that is VERY important to me having come from AMD. I like to see something more like am4. I know you sell quite a lot to businesses, but thsi really cuts down on waste because you can keep using a board. By the way boards have gotten insanely expensive. Raptor lake is a great product in many ways. It actually runs cooler than RYZEN in a lot of workloads especially at moderate activity. Ram is easier, and the IGPU is good. The bios thing was a disaster. It imapcted me. I had to rewrite my board firmware to fix the efi partition and i spent more time agonizing over bios settings than i personally like. I like the idea leaving some on the table a comfortable buffer so when mistakes happen they don't matter because you have a buffer. Another big learning for you. I work at a huge tech firm too. Your reviewers shouldn't be telling you this. The fact that you were surprised you should be internalize. It happened a bit to amd too to an extent but you need good quality control. IT will save you a lot of money in the long run having a bad launch can damage the brand long term. What I'm really trying to say: I think raptor lake has both good and bad lessons for you. Then please consider changing your approach to sockets. Much of what you said at least aligns with this. It's getting very competitive though you guys are going to really have to tighten this stuff up. I do actually worry about Intel a bit.

[@TAH1712]

if not the 285, then the 265 looks almost as good and much cheaper. Maybe I'll go for that.

[@godnamedtay]

No one, especially anyone from Intel, is reading this elongated nonsense, foh.

[@elizabethiradukunda1950]

It's befuddling that a company the size of intel can still make such basic rookie mistakes after 14th gen i thought they learned their lesson this is what happens when you take an engineering company and give it to MBA's who care about nothing but share price i downgraded from i9 14900 k to i7 12700 because intel refused to admit they had messed up with the i9 14000 and refused to take responsibility i'll wait for a few years and then go for chinese Loongson - never again will i spend a dollar on any intel processor

[@vorkzen4767]

What about AMD?

[@대구막창]

루나레이크 짱짱맨

[@teej_youtube]

Thats because people are whiney little cry babies who will shill to their youtube god reviewers nuts all day long without even getting the cpu and reviewing themselves first hand

[@JoyTrader999]

One more thing These new Intel CPUs are partially manufactured by TSMC (Shows Intel's Desperation and Incompetence) So why not purchase a CPU which is 100% manufactured by TSMC (AMD) ?????

[@thetheoryguy5544]

Cost you dummy.

[@Morne_Smith]

So how does AMD make SMT work at LOWER power usage and greater performance? There is something wrong in intel architecture...it might be a good idea to just go back to the design board and come up with a new architecture.

[@Bass.Player]

I understand that the socket will be around for three generations? 2026? I just moved to this platform and if it is a one time socket I moving to AMD which i did once before when Intel Socket life was short. This equipment is way too expensive now to limit the life of a socket, it's wasteful and just plain arrogant. However Intel has been know lately to just PHUCK things up!

[@bluearcherx]

the guys sitting here saying "i don't care how much power my desktop uses" are part of the problem. Meanwhile Apple is blowing everyone's doors off with 1/3 the power, including encroaching on the high end GPUs.

[@Brol1y]

Yea Intel didnt test their processors in games and they are surprised by low performance xd They are liars or they dont know what they actually do. Both are extremally bad. BTW performance in games is worse than 14 generation. Power consumption its Lower than 14 gen but its still worse than AMD had to offer Release processors with Edram + 8 Performance cores and everything will be fine in games

[@godnamedtay]

So what really what went happened, Robert?

[@Erictraiven]

Is this an hardware fix or software fix?

[@SaschaRobitzki]

Software.

[@tibbydudeza]

I don't care about power efficiency for my desktop - laptop/mobile obviously should be geared towards this.

[@nevzataydin1]

some of us care. high wattage means more heat, and more heat leads to expensive coolers and increased electricity bills. + more noise and possible throttling.

[@tibbydudeza]

@@nevzataydin1 I have a solar system so during the day most of electricity comes from the sun and at night I tap my batteries. But I don't think PC's compared to things like a pool pump , water boiler, oven, A/C is that of an expense.

[@stephanhart9941]

Shoulda stayed at AMD! Homie has aged 10 yrs in 2 yrs.

[@julienlauzon1106]

i build myself a 12900k system is going well but few weird details, i think its more windows11 fault, but i'm really looking foward to sell my sytem and build myself a 11700k system when rtx 5090 gonna launch, something is wrong with big core little core, the only thing i'm ''scared'' to returned too is pcie gen 4, i admit that gen 5 download speed is superior but for all the little weirdness i have i wish to return to 11 gen lol its not normal but i really think i will doo.

[@BravoSixGoingDark]

Why do you need to downgrade? Go with the 7800X3D.

[@jordan3802]

Arrow Lake is still far less efficient than 7800 and 9800x3d.

[@tomazgrad8919]

bla bla bla............bla

[@FougaFrancois]

Robert's claims about hyperthreading are inaccurate. There are no better return on investment than SMT when it is done right. Recently, intel implementation of SMT has not been tuned properly, causing him to believe what he is staying.

[@BOKLUK_CHOVEK]

To late,every one looking to buy a 9800X3D 😁 Next time ,intel give the people what they want : fats gaming CPU

[@greatwall9741]

Robert, you were in a good place, what happened? Your happy face ain't saving the blue team.

[@FougaFrancois]

Wow, Pretty sad/pathetic state of the union in chipzilla.

[@mraei]

At least on AMD's side SMT has been beneficial w o much influence on power and with little addition to the area. The only downside was security.