Nvidia’s Coming for Intel & AMD

[@UFDTech]

► Thanks to Silverstone for sponsoring today’s video! Check out their new IceMyst AIO coolers & IMF70’s here: tinyurl.com/yr8sfay8

[@CautiosulyOptimistic1440]

I read a report that states it's coming out on the z790 and that they will discuss this in computex. Also I hear that they're going to make it flat. Like a large rectangle.

[@ArouzedLamp]

DDR6 with APUs go BRRRRRRRRR

[@ArouzedLamp]

Also did we get an update on AMD's development with Quad data rate memory?

[@RobloxianX]

By 2025, Mini PC's are going to become great gaming machines. Killer fast APU's with killer fast memory. And, Thunderbolt 5 for eGPU connections that wont bottleneck even a 5090! This is a great year for portable desktops.

[@ranjitmandal1612]

🤣

[@CovenantAgentLazarus]

@ArouzedLamp super gay and not funny comment.

[@ArouzedLamp]

@@CovenantAgentLazarus ur mums super gay Lighten up bud

[@WINDOWSXP9013]

Ddr 6?!?!?! I have ddr 3 :(

[@breyefokalagloire5393]

Rolling on DDR5

[@reallycoollondon]

Not joking, I have Ddr.

[@WINDOWSXP9013]

@RAM_845 I had an i5-2400 last year, this year I upgraded to an i7-3770K, and I am not planning to get a new motherboard :)

[@WINDOWSXP9013]

@@reallycoollondon I can relate

[@zbatchDOC]

3 is time for an upgrade. I just made it to 5 literally this week

[@itsdeonlol]

I'm still waiting on DDR7 to come out.

[@Penguingaming69]

Nvidia bouta drop that gddr7 on ya

[@DustyCruz]

Nah fam, we're going TRIPLE. TDR1!

[@ranjitmandal1612]

💀💀💀

[@yahm0n]

Yeah, with the rise in AI and more powerful integrated graphics, 2-3x speed increase for DDR6 is not nearly enough.

[@aguywithaytusername]

​@@DustyCruzthere's already qdr(quadruple data rate) which is used in gpu memory since gddr5

[@00Drizz00]

AMD already has an Arm cpu in the works: Soundwave.

[@R-Tech_Gaming]

Soundwave superior... Intel & nVidia inferior... 😂

[@DementedLessons]

TEAM RED RAHHHH

[@LucidStrike]

Yeah, I'm surprised he didn't hear about that already.

[@JonLaRue]

The RAM Speed comparison was weird it’s like saying you’re faster when you’re running vs Usain Bolt sleeping.

[@DustinShort]

I said this on the last one too, Bette Midler v Ford motor company 1984. They wanted her to sing a song, she said no, that hired an impersonator. She sued and won in a federal appeals courts setting a pretty strong precedent. Open AI asked Scarlett to do the voice and she said no.. They claim they talked to the other voice actor before asking Johansson but I suspect once they asked Scarlett she even the her tweet sunk any defense that it's coincidental which is why they pulled the voice so quick.

[@tabiserebour5912]

They are probably going to add "AI" to the DDR6 name at some point in time. I'm a time traveler.

[@DNM37]

DDR6! That means Dawid does tech stuff not freaking out on single channel ram anymore?

[@NMModder]

Impossible

[@LRK-GT]

Nope, everyone should still be upset with Single Channel RAM implementations/configurations. Even tho DDR6 will clock a *lot* faster, latency will be higher, and the overall bus-width will remain 128-bit ('dual' channel*). edit: I was wrong. The Bus' bit-width is widening. IIRC, due to a more error-tolerant encoding scheme. *DDR6 further develops on DDR5's 'split sub-channel' architecture. Still only 128-bits wide, in total.

[@ranjitmandal1612]

😂

[@DNM37]

@@LRK-GT In another sense it provides a lot more flexibility like at first I got one stick of 16g 6200mhz then later down the line got another but that'd be a nightmare on DDR4 and so far going from CL10 ddr3 to CL16 DDR4 and now CL36 DDR5 so far nothing has gone that wrong. So yeah, let's see what happens in the future

[@HanSolo__]

It was funny, but the single channel is like drinking with a straw.

[@danehutchins3845]

Still rockin DDR3. One day, Ill be able to afford an upgrade. I like lying to myself.

[@jackielinde7568]

I don't know what Brett is getting worked up over DDR6 for. I, for one, am looking forward to Dance Dance Revolution 6.

[@dirkjewitt5037]

As soon as DDR6 comes out, APU's will finally matter.

[@Angel7black]

Probably not no

[@LRK-GT]

Only thing that'll 'rocket' APUs would be on-package HBM or X3D (maybe, a dedicated MCD or two for hardware compression/decompression and cache) The latency on each new generation of DDR along with the same-old 128-bit bus limitation means iGPUs don't see huge improvements. IIRC, Intel had (broadwell->most of 12nm++++++++++++++++++) -and, is returning to eDRAM for their CPU's iGPU. Entirely because of the need for a small very low latency cache.

[@dirkjewitt5037]

@@LRK-GT It's a few years away. We'll see.

[@dirkjewitt5037]

@@Angel7black we'll see.

[@Mervinion]

AMD is preparing Strix Halo APUs. They are rumored to be up to 40CUs and 256bit memory bus. Basically quad channel RAM. It will definitely make APUs matter. It's just a shame that this chip is to big, with too many pins to fit on AM5 restricting it to laptops and maybe mini-PCs. One posibility for it to exist in DIY is on TRX platform but with up to 24c and 40CU nobody expects it to be cheap...

[@seannesbit2218]

The AI thermal paste reminds me of when we had to put stickers on printer paper that said "Windows98 Compatible" at CompUSA because people kept asking if paper would work with Windows98!

[@arthurmingo]

My nephew and I love watching this show together. It’s our thing.

[@NobodyImportant-it6ii]

Damn, these videos have been dropping really quickly lately 😂

[@barrym426]

This video reminded me that nvidia NForce chipsets used to be a thing. Also "Soundstorm". Nvidia used to do more things.

[@ZaberfangX]

Nvidia been slowly working on Linux drivers getting better and opening up the driver more, one aspect to keep a eye on that part as well, as we have box86 growing to better support arm with x86 base games.

[@xrafter]

You can't forget box64. It is box86 but for x64.

[@ZaberfangX]

@@xrafter true

[@DustyCruz]

When will we get to TDR1 - Triple Data Rate One?

[@ironicdivinemandatestan4262]

There were experiments with QDR RAM, but cost and complexity meant that it never saw mass adoption.

[@themomaw]

If you want to be really depressed about the progression of GPUs, don't just look at at absolute performance, look at performance PER WATT. You can always throw more silicon at a problem, perf per watt is the REAL "it got better" metric. 1080Ti was a 250 watt part, while the 4090 is a 400 watt part. So you can divide that whole chart accordingly.

[@lanwoah]

I have heard the comparison between scarlett johansson and the openai voice. It doesn't sound like her AT ALL. Its a little similar but don't forget we live with 8 billion on this planet, the chance of someone having a similar voice is very high.

[@tomb3782]

Hey Brett, did you ever think of doing a Retro tech section?

[@Snafu2346]

my first 7800x3d had a internal memory controller fail on me and I had to RMA it, and I was told it was because of running expo at 6400 mt's. So now I dont even run expo and run it at the stock 4800 mts. IF DDR5 is not even stable now should I ever care?

[@LRK-GT]

Yup. Don't much care about DDR5+, when RAM manufacturers are still allowed to advertise the "outside spec, OCing" clockspeeds and latency.

[@RobloxianX]

iGPU's are going to get a CRAZY performance boost from this, if Strix Point Halo is going to be a 4070 with DDR5-8600, it's going to be a 5070 with ANY DDR6 memory module.

[@lycanthoss]

that's not how it works

[@Angel7black]

I can promise you Strix Point Halo will not be anywhere near a 4070 or even a 3060, especially when theyre having the performance issues theyre having with RDNA3. Youre talking about an advancement of tech that would probably take 15 years atleast to catch up to with APUs, and notice im sayinG ATLEAST. Current iGPU’s still havent caught up to the 1060/580

[@LRK-GT]

@@lycanthoss Not to mention, every new DDR RAM generation, latencies increase. Looking at how old HBM GPUs perform (extremely low memory transaction latency), DDR6 will not particularly improve things. (At least, not without dedicated cache and compression/decompression hardware; as is in Navi31's MCDs)

[@ranjitmandal1612]

🤌

[@ILoveTinfoilHats]

It's not. That's also not how GPUs work.

[@Draal_1]

Water companies, get on this. Name: AI Water Slogan: Become a better you. Description: Vapor distilled water, now with nanobots

[@seanc6754]

Ddr 6 will help APUs that are in handhelds like the steam deck and Ally

[@MaheerKibria]

AMD has been making ARM-based server chips since 2016. They can make ARM-based processors for the mainstream. They just haven't had a need to do so. The Steamdeck and z1 extreme handhelds show that. X86 can be low-powered and performant.

[@LRK-GT]

What is that 'transition' SFX?! It seems familiar. Maybe, shock/electrical noise from an old video game?

[@julmdamaslefttoe3559]

My ryzen 2600 failed XMP due to bad memory controller. seen someone have similar with a 7950x3d

[@jimbodee4043]

That silver stone air with the clip on modular fans looks great but are they noisy low CFM fans?

[@sw1ft3]

It's Friday and after midday but my brettfast was incredible thank you. Crispy bacon, Caramelised onions with scrambled cheese eggs! As a side note can you buy Rhys a light so we can actually see his face?

[@_lonath_]

Eating breakfast? I've had dinner and am trying to sleep

[@wizencoyote78]

Brettfast

[@aykyi2668]

GPU switch makes only sense if: 2x performance at 1/2 watt or 4x performance at same watt. This rule will stay forever. CPU is same at performance/watt switch time. Also a switch could make sense if ram bandwith/capacity increase 4x.

[@jburton7978]

As someone who used to work at Intel, AI was a big thing pushed in the server space. It started with KNL and that didn't really go anywhere and was killed quickly and now Intel GPU's that aren't good enough for training but are good enough for inferencing. A lot of customers don't like working with Nvidia and were willing to get ARC GPU's to move away. Servers are very much x86 reliant at the moment with infrastructure software built for it but with CPU's getting hotter cooling is becoming an issue, whilst I was there we were experimenting with immersion cooling as a solution. If a powerful ARM server CPU can come out with a reasonable TDP I could see it doing very well if the software is there to support it, maybe mix architecture environments at first. If x86 doesn't make a radical change and get power down, I could see an ARM solution taking over like with PowerPC falling out of favour for the same power consumption reasons

[@quarksarranged]

But what is the latency of DDR6 going to be?

[@xrafter]

Should be the same or a little bit higher than ddr5. Cas would definitely go higher, but the first word should be the same.

[@Penfolduk001]

The AI Pin always seemed pie in the sky for the current state of the tech. Perhaps in 10-20 years time a similar product will be viable.

[@NolanWayne-h9n]

Me too. I run a htpc with i3 2100 and 4gb ddr3, and i game on a i5 2400 gtx 960 combo with 16gb ddr3

[@Kmcornell23]

I think AI would be awesome in video games when it comes to NPC's. Like imagine playing something like GTA and you could walk up to an NPC on the street and yell at them through your microphone and they'd actually respond based on what you said to them.

[@Dunkdamonk]

That's why gta6 is taking so long. They are training a neural network for the npc ai. If you know how in depth the programming for rdr2 was as far as npc daily lives and interactions or world building... gta6 going to be revolutionary. They have patents on this and it's exciting

[@corymatthewmouser1543]

Even if they just used someone that sounded like Scarlett Johanson, I feel that they're intentionally misleading people into thinking it is her. Sure, people can often sound pretty similar, but there's no way they weren't planning int making people believe it was actually Scarlett

[@miguelcossyleon7284]

Once you see it, you can't unsee it....Brett Hanks 😮

[@jipillow1]

Nvidia should absolutely enter cpu market - so many people in the comments hate competition

[@awaiswasi354]

and ruin the prices for cpu market?

[@Penfolduk001]

They kinda tried by trying to buy ARM. But the competition authorities around the World wouldn't let them. Besides, an Nvidia CPU would probably make Intel look power efficient... 😂

[@trollmastermike52845]

​@Penfolduk001 I just care about performance. god gave us ac, get your broke butt to work, and afford the electric bill 😂😂

[@alexis1156]

​​@@awaiswasi354This is not how things work. Nvidia would have to compete with AMD and Intel, and unlike the GPU market Nvidia in this case wouldn't have nearly the same market share as they do with gpus. They can't just jack up prices and expect it to work when AMD and Intel are there.

[@antiimperialista5764]

Wtf because we need their predatory practices in that segment too smh

[@deanhantzis9630]

Great brant. The question was marketing foolishness.

[@funkijote]

10:20 Perhaps referring to the fact that Mediatek + Nvidia are releasing a Snapdragon X competitor next year. Mediatek are the biggest ARM soc manufacturer by volume and already power millions of laptops (Chromebooks), and in addition to dominating budget mobile segment, are eating up mid and flagship-tier marketshare outside of US & EU. Entry-level Windows 11 2-in-1s and tablets on the Kompanio 14XX (upscaled/overclocked Dimensity 9400, a mobile chip surpassing Snapdragon 8 Gen 3 in benchmarks, with Nvidia GPU), could be quite compelling if the price/performance ratio is right. Chromebooks similarly become more attractive in this reality with powerful, cheap, power efficient, MediaTek+Nvidia chips.

[@ARTUROOVEM]

So now my new computer setup will be on hold for longer, have to wait for new ryzen cpus, new nvidia gpus and new ddr6 hope to plan all together

[@cybernit3]

So the RTX 4090 has approx 836 TOPS but what would the RTX 4070 and AMD 7800XT have? I think the 7800xt would have half of what the RTX 4070.... it should be more than 40 TOPS right?

[@leophoenix1452]

Nvidia is going to absolutely demolish the competrition in terms of performance on arm... Remember tegra 4 vs the competition? Tegra X1 vs its compettition back then?... They were running circles arround the competition...

[@WingCoe]

Is the connection on the new DDR5 already saturated? Because it would be great if it could just be a drop in upgrade for exisiting DDR5 sockets. Just wishful thinking, LOL

[@LRK-GT]

Very much wishful thinking. Every DDR gen requires new slots and memory controller. Reason being, the 'generational improvement' typ. comes (largely) in the form of new signaling methodology. (necessitating a new trace and pin layout)

[@terrycook2733]

Heat and dye sizes... gotta have smaller and smaller dye's to sell... quality of such products and stability.... its not that they cant give us the performance they are just working on efficiency to keep the cpu's cooler and higher clocks with less watts. EPYC is the frontier so far. But Zen6 is going to be the next alteration for future tech that makes you regret getting anything before.

[@445588997]

Still on DDR3 for my home server and DDR4 for my "gaming" rig

[@kgonepostl]

"We're with a billion Dollars" The market: "Nope......."

[@teetea7734]

Nividia ARM PCs, starting price, your first born, both kidneys, an arm, and a leg.

[@ToxicGamer86454]

Nvidia is crushing AMD atm. Nvidia stock has doubled in the last 6 months while AMD is up 30%.

[@ScottyG-wood]

I’m going to wait until the next socket for DDR6. DDR5 was just too much of a hassle for the higher speeds that didn’t really translate too way into real world performance. Benchmarks and iGPUs showed up but not everything else IMO.

[@LRK-GT]

You're *probably* going to have to wait, regardless. I've heard no rumor of a DDR6 IMC on Ryzen 9k. NtM, the new signaling and sub-channel layout may necessitate a new socket (or at least a modification; ala AM3-AM3+)

[@cracklingice]

Nvidia does not need a CPU - as you have already stated, they make the built in NPUs look pathetic. They just need CUDA to be certified in Windows for copilot.

[@TuxLinuxOfficial]

DDR3 on top

[@AtulC-c6d]

question about AI pc is in lines with have a CPU+GPU+NPU+XPU or any other combination which doesn't suck lot of power and not a discrete GPU in the PC/Laptop.

[@deanlonagan1475]

..Ive seen vids on DDR5 Bit Error Rates at Gigahertz frequencies...Id like to see an eye diagram of current DDR6...21kMHz?...how are they making that stable?...

[@prague5419]

DDR-6 21333, I thought it was a type-O. I about blew my drink out my nose.

[@raawesome3851]

I'm surprised you didn't cover the Linus story about how he was cleared about the allegations.

[@ForzaE2]

I wonder what will be sweet spot for DDR6? 17000MHz ?

[@Coconutgl]

Keep in mind Xess scale factor is higher than DLSS/FSR in same quality preset, for example Xess balanced sf=2x, while DLSS/FSR balanced sf=1.7x. Therefore Arc graphic performance boosting is some kind of cheating. Also even the fps of Arc graphic maybe higher in some cases, there are much more stuttering cause the gaming experience really bad.

[@CountryMouseLaser]

You sound like Al Pacino when you said "Tom Hanks".....you got him down...

[@adamjames0710]

Sometimes I think my pc needs an upgrade… but if I imagine I’m a time traveller and only play 10 year old games then I’ve actually got the worlds fastest pc 😎

[@qwertyuiop.]

Microsoft is making Windows remember everything. This comes after making it harder to setup Windows with a local account. That's yet another reason to disable telemetry and not login to a MS account.

[@NicolasChapadosGirard]

Can you see a futur where you just dock your phone and boom PC? Maybe dock has a GPU on it and then you fit a gaming PC.

[@abdulabrahams1322]

im still rockin ddr3

[@Kmcornell23]

I don't think that ARM is going to take off in Windows. The reason Nvidia doesn't have x86 CPUs is because they can't have them. Windows is coded to run x86. I order to get ARM chips to work with Windows you either need some kind of translation layer (which hurts performance) or Windows would need to natively support ARM but that would require a LOT more work on Microsoft's end. I have to assume that Microsoft won't want to remake their entire OS which means that ARM is never going to be able to keep up with Intel and AMD. So if they can't sell ARM based computers at a substantially lower price, people just aren't going to buy them. Why buy a slower product for the same money? Not to mention that other programs aren't coded to work with ARM at this point so they'd all have to recode all of their stuff as well. It's just way too much work for not enough benefit. I think they just need to stick to devices like phones because I don't see this going anywhere on the consumer side of things.

[@TheSundayParty1]

There was actually a lawsuit a few years ago in which I believe Ford or another major brand wanted Bette Midler‘s voice Bette Midler refused so they hired an impersonator of Bette Midler to do a radio commercial instead. She sued and won the lawsuit that it was still not honoring her wishes to not appear in the commercial. So there is a legal framework for Scarlett Johansson to sue open AI if they hired an impersonator of Scarlett’s voice.

[@breyefokalagloire5393]

I'm addicted to your videos

[@ranjitmandal1612]

Have to wait and see what nvidia come out with next near

[@nickyt.creates6637]

I’m only waiting for RDR3 at this point 😅

[@Kifey8183]

18:05 I think they showed their hand when they withdrew the voice model. We will see where this goes I guess.

[@inkredebilchina9699]

so what's to stop nVidia to make small form factor AIO PCs? I mean CPU is there, GPU is there, RAM as well... like really small, really powerful and very VERY expensive PCs.

[@baltazarcortez7328]

But there are people in the comment section vehemently on the side of AI. You are against Chinese knockoffs and yet turn around and defend AI. There is nothing innovative or intelligent about AI. It’s just an intelligent and new way of copying somebody else’s work and claiming it as your own. It’s intelligent in that it does enough to obfuscate that the goods that it claims as its own were actually stolen.

[@BodySculptTV]

DDR6 is ridiculous, same as PCIe 6. DDR6 will be probably gear 3 or 4 with nasty timmings just to say wow your memory is going so fast when it really isn't. They are having issues with thermal throttling of NVMEs on PCI5, but they are looking at a PCI6 standard. It's all marketing hype, they don't even have a PCI5 video card yet. Now you can see why Nvidia was trying to buy ARM, but the EU stopped the merger. They did it anyways....

[@ZeuSVI]

Why not compare DDR6 to the current generation(DDR5)? 10x increase? Well if we compare it to DDR it has an increase of 200x, so what? Doesn't make it any better or worse if we compare it to old stuff. This is like comparing NVidias 4000 series RTX cards against 2000 series instead of the 3000 series, makes no sense.

[@KevinKnightCanDance]

old McDonald had a data farm A.I. A.I. Oh with a 1001 here, and a 1001 there. old McDonald had a data farm!

[@LucidStrike]

Asking Nvidia to save y'all from Nvidia is wild.

[@aljonzc]

Hey... I'm still on DDR3 and 1070ti

[@douglasknapp4059]

Tom Hanks! Now I can't unsee it.

[@zbatchDOC]

So wait- can I use my GPU for “copilot” things? Or is Microsoft limiting that to JUST cpu stuff? That seems pretty daft if that’s the case.

[@StefanEtienneTheVerrgeRep]

I've been ready for Dance Dance Revolution for years. But I'm ecstatic to announce my new A.I. tweezers and A.I. Static bracelets.

[@thekangaroo1880]

I just got ddr5 sticks... Still waiting on a motherboard and cpu to install it into. Edit: i just realised that ddr5 is already like 4 years old.

[@Wolvey10]

Please take EVERY opportunity to rant about even the most basic of computer programs being called AI. Even as a tech enthusiast, I’m beyond tired of hearing AI marketing.

[@Icefury]

Given Nvidia's long history of abhorrent anti-consumer behaviors, while I'm excited to have a 3rd competitor in the PC CPU space I remain guarded. We will also need to see how well their motherboard chipsets perform, driver support, and security / bios support are. Breaking into established / mature markets as a new contender is hard, just ask Intel with ARC graphics.

[@DenOfTimbsllc]

I bet Apple is going to try to get their hands DDR6 RAM as soon as it comes out.

[@DenOfTimbsllc]

#BrettlookslikeTomHanks

[@ensabrenoir5669]

........woah.......people actually don't have hats for their pet rock? What kind of sick world is this.

[@lostskull7467]

10:56 I got a 4050 laptop... How do I get started? 😂

[@ashig_1]

Bro i just built a ddr5 7800x3d and a 7800xt you cant do this to me!!! Like i just finished the build this monday smh

[@wakzyon]

Old Mc Donald had a farm, AI e I O, and on that farm he had some ASIC power by Nvidia AI e I O

[@EIDJgaming]

I feel bad for MSI, trying to package an Intel APU has got to be hard. And it's not the GPU that's the problem. Intel mobile CPUs are notoriously bad at power consumption. They just don't function great at low wattage. Unfortunately I imagine the biggest setback with those APUs is getting the CPU to a stable clock at low power so it can maintain battery life. Honestly I hope they find a better means of power allocation so they can get the most out of those GPUs. The arc cards are actually pretty impressive, and from what I've seen their lower power variants are also quite good. Hell I run an A380 discreet card at home with no additional power other than what comes off the motherboard and it pushes 1440p pretty okay. It's not amazing or anything, but it reliably does push 1440p.

[@fateunleashed9680]

I don't they most likely got a massive bag of cash to use Intel to get them into the handheld market. Whatever happens good or bad will be the consequences of their own actions. Meteor lake is only more efficient than Raptor lake under 15w, and performs worse than Raptor lake over 45w. Almost all the handheld consoles are in the range meteor lake is worse than raptor lake at. The steam deck and Ally may have similar performance, but will have much longer battery life due to the much more efficient SoC. Same way we don't see Nvidia CPU's in handhelds. Outside of trashtendo.

[@EIDJgaming]

@@fateunleashed9680 Look I'm not saying they weren't incentivized, and I certainly won't go to bat for MSI normally, I personally have had terrible luck with them. All that being said, an intel apu for a handheld is a hard thing to accomplish and they're the only ones trying. I just feel bad because the gpu tech is actually solid and coming along quickly, they just get handicapped by the face that AMD mobile cpus are massively more efficient, especially at lower wattage. It's just sucks to see people trying new and different things and immediately getting pushed out of the market. We need different, else we end up seeing stagnation.

[@markolazarevic6944]

Me and my homies waiting for DDR 10

[@saberknight2474]

I thought Zen 4 was already having stability issues on DDR5 with transfer speeds above about 6000mt/s. I wonder if AMD’s development for new memory controllers for Zen 5 and beyond will be able to keep pace.

[@jemborg]

If you up the voltage to 1.4V 1.45V you'll get better results. But I feel for you man. Running a CPU with RAM on EXPO should not disqualify you AT ALL. I thought my 7700 had died on me maybe and they just handed me a refund. (So I bought a 7800X3D.) But, it was the *motherboard* at fault and it was blaming the CPU. Point being there were no questions asked. _And I've learnt enough to say no to any anyway._ Still, there are no warnings about warranty with EXPO. It's BS. I've tweeked my rig a fair bit and as far as the 7800X3D goes you won't lose running the RAM at 6000 MTS and tightening the timings. Yes, 8400MTs RAM is faster... barely. I kid you not. Like 1%. In practice you won't notice. You see it tested on youtube. It's not considered worth the cost. And it makes even _less_ difference on the 7800X3D... because of the VCache. It's bloody brilliant. I run my "terrible" SAMSUNG at EXPO 6000MTs CL36-36-36-96 RAM at... 30-34-34-76 with VDD = 1.4V. The rest I undervolt. CPU = PBO and - 30mV curve VDDIO = 1.3V VDDQ = 1.3V VSOC = 1.225V VDDP = 1.1V tREFI = 38607 The last one is a _big_ deal to many. It can get pushed to 65535. Don't change the FCLK it's not worth it. I'm actually a bit conservative here BION. Not wanting to risk stability. I run a 4090 too BTW. Take it easy brother. 🇦🇺

[@DaRealEaze]

It was just a matter of time before nvidia tried

[@geekswithfeet9137]

Is Time spy is still even relevant as a benchmark?

[@mcdazz2011]

I can definitely see that Tom Hanks reference. Brett, do you know who your father *really* is?

[@OneKyox]

Intel wll stay with x86, the internal message for the teams is that “ARM is far away from having the same support we have”

[@Blackfatrat]

Intel has been somewhat preparing for X86s for a while now which is basically x86 but sheds some really old stuff, emulating that part instead, and keeping the normal x86 64 bit. I think that's a likely path forward.

[@guitaristkuro8898]

I wonder if that Fortnite increase just came from the UE5.4 CPU update.

[@rwait1287]

GPUs Only 2.8X faster in 8 year Isn't it because of SLI?

[@ZSchrink]

When does 'DDR AI' hit the shelves?

[@kai990]

what you are describing about open ai there is just the modus operandi of every profit driven venture.