You Shouldn't Overthink Ryzen 9 7950X3D

Gordon is joined by Wendell from ‪@Level1Techs‬ to talk about AMD's Ryzen 9 7950X3D and figure out why people are getting so confused and angry about it.
How Ryzen 9 7950X3D works: • Ryzen 7000 V-Cache: Ho...
Buy PCWorld merch: crowdmade.com/collections/pcw...
Follow PCWorld for all things PC!
-----------------------------­---
SUBSCRIBE: kzbin.info?sub_c...
TWITCH: / pcworldus
TWITTER: / pcworld
WEBSITE: www.pcworld.com
#amd #ryzen
00:00 - Intro
00:38 - Why So Confused?
02:48 - Core Parking
07:22 - Are We Overthinking?
10:48 - AMD Thread Director?
18:43 - Xbox Game Bar
22:03 - Developers
23:54 - 3D V-Cache On Both?
26:06 - 3 Takeaways

[@SnuSnu91]

Okay, let me get that straight: basically the 7950x3D is the 7800x3D (8 cores using v-cache) just with another 8 cores if you are up to higher demanding tasks where 16 cores are helpful. Let's say you are gaming and you have 100+ tabs open in Chrome, you watch a movie on the third monitor and while queuing for the game you do stuff in photoshop - that will enable the other cores besides the v-cache? so... 7950X3D - if you are a gamer AND use it as a workstation 7950X - if you use it as a workstation (because higher clock speed) 7800XD - if you are a gamer overall the X3D are much more efficient in terms of power usage while at the same time giving a lot of performance. Am I correct or... ? This is all so complicated for me as I have no idea about all this stuff lol. Recently started to show more interest in this kind of stuff since I wanna know what I spend my money on. ( please excuse my english 🙂)

[@RogueWraith909]

@@pcworld I think the problem is a symptom of alot of the reviewers not making that idea this clearly. I'm looking at building my first new PC since the early 2000's (been running laptops last few gens due to cost and simplicity) so getting back into looking at reviews and working out what parts I want and costing everything out before buying anything is more complex now due to how many tech changes have happened that I haven't been keeping up with... it's really interesting though and I'd like to thank people like yourself, Gordon and Wendell and the guys at GN and JtC for helping me figure this mess out again.

[@SnuSnu91]

​@@grimfist79 That was also my question, but if I understand it correctly - yes, the other cores will enable as soon as they are needed

[@Lemonpinesol]

@@pcworld generally if a consumer doesnt understand a product that is the failure of the company not the consumer

[@TheJacklikesvideos]

this all sounds right.

[@IrocZIV]

It does seem like game bar should have a default setting, but also let the user decide what CCD the game uses.

[@kelownatechkid]

@@pcworld That is frustrating. Users need to wade through and manually set masks

[@profosist]

@@pcworld Technically your not forcing it onto the cached die. You are marking it as a game and then it works with the platform to do that default action for a game. The checkbox is really there for games that were not detected as a game. Also why the checkbox is not there for things it knows are a game for sure. What im more interested in is does this or can/could it affect other multi CCD/CCX chips and potentially improve performance there. Say you have a standard 5950X generally it can boost 8 core more then 16, also latency should be better so it doesnt need use infinity fabric. Now why doesnt the gamebar detect that its another type of program and assign it an application moniker and run on the faster cores? Well its the GAME bar... Also if telling it once thing you dont necessarily need to tell it the other. The other action can be assumed. 3D v cache is the minority in chips currently thats why this is an alternate mode of operation. Id agree with Wendell the way were going with different types cores and even Intel is going to chiplets. This will be the default mode of operation in short order. Newer and newer chips have only tried to automate things more and more. This way the 99% has a better experience. For the 0.01% there's always been ways like process affinity to manually handle things.

[@ADB-zf5zr]

​@@pcworld I kept on hearing "Game Bar" in this discussion. Looked it up and remembered that it was one of the first things I nuked in my W10 installation.! Can it actually benefit me in some way as a "PC only" gamer.?

[@IrocZIV]

@@ADB-zf5zr If you use the 7950x3d or 7900x3d Game Bar is what windows is using to dictate which CCD is being parked. I think it also might have benefits when working with Intel's Thread Director to make sure the P cores get assigned properly.

[@mortlet5180]

​@@ADB-zf5zr I have compared the default BIOS and Win10 (all were clean installs from the same image after zeroizing the SSD, with direct die watercooling) performance to an optimized and 'de-bloated' Win 10 setup (also manually disabled CCD1 in the BIOS, as well as all power saving features, SMT and unwanted features such as virtualization) on both a 13900k and a 7950X3d and I still keep getting *much* better 0.1% lows together with a bit smoother subjective experience. YMMV since I never have anything else running while I game (I also disconnect from the internet and disable my antivirus for SP games ) and I have no use for more than 8 cores or SMT either (the difference in performance between 6 and 8 active cores is very small in the games I play. While SMT never increased performance, it didn't greatly reduce it either, except for limiting my maximum overclock on the 13900k).

[@UnlimitedDiarrhea]

I would love it if you could use the gamebar/Ryzen master software to assign all programs to the high frequency core by default, and to run games, and games alone, on the vcache. For now I will be able to do that on the bios + process lasso but I imagine AMD will add this functionality at some point.

[@bes12000]

I use process lasso for it and use CPU sets to push certain games to vcache, but it doesn't always work well, some games get lower FPS doing that.

[@earthtaurus5515]

The other alternative is adding the ability in settings under Display, you can choose which applications are attached to 'high performance' or 'powersaving' on both windows 10 and windows 11. So, the additional setting would be a simple toggle 'Frequency' or '3D V-Cache'. Allowing people more granular control regardless if they have a AMD gpu or Nvidia gpu. With AMD gpus, AMD can add this ability in Andrenalin since only Zen 4 and beyond will have integrated graphics (unless AMD decides to remove it in Zen 5). Then settings and Ryzen master would be the best bet.

[@vabenchu]

How is this working for you? I'm looking into getting the 7930x3d and using BIOS to get everyhting in normal CCD and process lasso to tie my games that prefer vcache folder to vCache CCD

[@Myselfsama]

@@vabenchu Also curious how this is working with manually configuring games in process lasso?

[@bobo-cc1xw]

For a game that could benefit/miss out from this have you tried Galatic Civilizations. A turn based 4x game that uses threads for AI to calculate in between turns which means it can fully load a 64core processor because it will just spin up AI threads. On the forums they recommend reducing multithreading because it can cook you processor.

[@morosis82]

I think Stellaris does also, Wendell I'm pretty sure has used it before for these talks.

[@yol_n]

Would love to see a video that uses temps as well

[@theshazman]

Thank you both for creating this follow-up video!

[@Amfibios]

hardware unboxed showed that the whole process of choosing the right ccd was a little hit and miss. so people not trusting the "auto mode" is understandable

[@potatorigs2155]

New tech comes with adaptation just like e-cores

[@hedroxthelegend]

Went for 7950x3d and 30C 6000mhz DDR5. This is hands down the best CPU I've ever had. Not only the performance is silky smooth, but the frame time is a straight line with zero hitches whatsoever. This is the first time I have 0 hitches in games, especially demanding ones like Cyberpunk. GPU is 4090. What's even more interesting is that the CPU runs 70-80Watts max and utilizes the GPU very well!

[@J-Kimble]

I did ended up buying a 7900x3d. At first I was concerned with the die-to-die switching during gaming, but I'm happy to same it's working fine. I only get stutters if I stay in the game menu for a longer time (minutes or so). In productivity tasks it rips though!

[@mleise8292]

Anecdote from me: I used core parking on an unlikely candidate, the i5-9400F, to put 2 cores mostly to sleep. That way the boost clock goes from 3.9 GHz up to 4.0. All you need is a separate power plan in Windows for which QuickCPU is a great tool. Then you can quickly switch between profiles to maximize either single-core performance or multi-threading. That's why I also think the XBox Game Bar thing wont really be a requirement for long until pro users find other ways to trigger the core parking that more suit their use cases.

[@leonrein7393]

We have Process Lasso!

[@mleise8292]

@@leonrein7393 I have heard of that one, but never checked it out. 👀

[@stephenreaves3205]

This sounds a lot like when a car has a V8 engine (not the js kind) and when you're on cruise control it'll stop using some cylinders to use less gas, but when you need more power it'll use all 8 cylinders again

[@RyTrapp0]

Except this is probably going to be WAAAAY more dependable than those garbage cylinder deactivation systems lol. Maybe Ford found a way to do this with the gen 4 Coyote without creating other engine probelms, but I'm not getting my hopes up. ...oh, right, so CPUs...

[@saricubra2867]

But you lose aceleration and overall snappy feeling of the car because of the constant turning on and off of cylinders. Just have all cylinders always active and improve efficency of the engine. Or go straight to the point and make an electric vehicle (on CPUs, the equivalent would be optics or optical circuits, not electronics, Moore's Law is dead.) I always use high perfomance mode (no core parking too) on my i7-12700K DDR5 system that i use for music production. Because everytime a core parks, it can produce an audio stutter OR the delay between the core turning on and off is just annoying for the stability of perfomance... Big-Little combustion engines for the best of both worlds?

[@bradhaines3142]

​@@saricubra2867 big/little is the exact same thing as core parking but its in clusters like gas engines do. they always use deactivation on V block engines because what they usually do is just shut off 1 side of it. chevy for a long time now has had a habit of making it happen on the same side too often and wearing out half the block while the other side still looks almost new. big/little would be insanely hard to do practically, and that's why they're just doing gas/electric hybrids. electric has a lot of torque which is perfect for lower speeds, and it can keep things rolling while the gas engine spools up when power is needed

[@griffin1366]

@@saricubra2867 This exactly. It's personal preference. Same with the "engine turn off" when idle (ie in traffic etc.). It's unnerving and you just have to trust the manufacturers...

[@saricubra2867]

​@@bradhaines3142 Big-Little is not the same as core parking, just different cylinder sizes.

[@foxfff123123]

I’d like the scheduler to default any gaming loads to the 3DVCache die by default or to be able to default that in the Ryzen master software. The problem right now is that it’s early in the hybrid CPU lifespan and the kinks have not been completely sorted. I see this CPU as a best of both worlds in that it has the functionality of a 16 Core 32 thread processor for slightly less performance in non gaming loads than the X variant, but has the best gaming performance of any processor on the market. It is also more efficient than both the X and all the high end Intel processors on the market right now. It is just in need of better software support and the processor is extremely well designed and thought out, day 1 reviews are not really accounting for future scheduling fixes and the fact that it’s both a gaming and productivity CPU and it does both tasks extremely well, just not better than the closest relative CPU for the latter.

[@gurugamer8632]

For video editing in apps like adobe premiere , after effects and DaVinci Resolve which CPU is the best for performance? I9 13900K , 13900KS or AMD 7950X or AMD 7950X3D?

[@saricubra2867]

​​@@gurugamer8632 i9. The iGPU features make a huge difference, then you pair it with a beast NVIDIA card.

[@gurugamer8632]

@@saricubra2867 i have the rtx 4090 … which igpu and cpu is best?

[@saricubra2867]

@@gurugamer8632For the CPU? none (for 1080p gaming ) for the 4090 (it's way too fast, not even the GTX 1080Ti did a massive jump in perfomance for one generation). iGPU definetly Intel UHD770, productivity, yeah, pick up the i9. Get a very good cooler to tame that beast.

[@hadleys.4869]

@@gurugamer8632 are you gaming only or gaming and productivity? I honestly think the best cpu that makes no compromises in both fields is the 13900k/ks. If you are just gaming I would either get the 13900k or the 7800x3D.

[@1981AdamGs]

This is a useful video. People get so worked up about these things. And to be fair, there are a lot of tech channels that blow this stuff way out of proportion. That doesn't help. I won't name any channels specifically because I don't want to start an argument. But a lot of channels are very alarmist about this stuff.

[@SirCrest]

I play Factorio on 4x speed so 240ups is useful for me :) I guess my area of disappointment is that we're already using game bar to essentially tell where something should run, but instead of telling it where to run it's telling which cores need to go to sleep in the hopes that the Windows scheduler doesn't screw things up. It sounds like it hasn't had the updates to just behave and keep games on CCD0. Why park the cores if the scheduler would just know to not randomly migrate threads from one application. This appears to be an issue already on 7950x where games that don't benefit from 16cores will still be balanced between chiplets and disabling a CCD results in much better 0.1%. Which makes me think: A) the scheduler still isn't where it needs to be and this is AMD's 90% viable solution. And/or B) 7950x could also benefit from this but AMD said multiple times there's no problems with the windows scheduler on dual CCD Ryzens. But as we're seeing in this discussion the problem isn't just V-Cache vs HF it's about migrating games between chiplets and getting a latency penalty. Which sounds like Dual CCD chips from 3000, 5000, and 7000 should have had this already. Both dies having V-Cache being a bad idea makes sense if we're also trying to avoid chip to chip latency. I'm with AMD on that. This just feels like a very inelegant solution. I certainly hope MS improves things natively.

[@JoeWayne84]

Yes the problem is they are using chiplet instead of monolithic because they want to be cheap. We will most likely always have a latency issue with infinity fabric /ccd interconnect . I don’t understand why AMD didn’t just make a single 16 core ccd for there flagship.. Oh that’s right because it’s AMD and they like being the inferior product.. if they were a cheaper option to Intel I would think a lot higher of them but at this point they are the premium priced option and they are a experimental at best chiplet design that has been finally refined enough to compete after 5 generations of having issues. Basically the 8core amd part is the premium option for enthusiast. The 16 core parts have been to pad the numbers on artificial benchmarks for bragging rights to increase Sales. But at least all the CPUs nowdays are amazing for gaming even if they could be better

[@SirCrest]

@@JoeWayne84 Is this meant as a reply to me?

[@JoeWayne84]

@@SirCrest no my bad I was commenting on the video mainly

[@profosist]

GameBar is not telling it to run anywhere. It merely marking the game as a specific type of task. Windows scheduler understands for that config of the cpu it should keep that task in a certain place if possible. And yes I wouldnt be surprised if this applies to other dual CCD chips or really anything with a asymmetrical or chiplet approach. Since we switch to UEFI its only been better for the software to be better aware and more correctly controlling the hardware. This is just second gen. Last time we didnt even get proper turbo and only had one CCD. Now we have full turbo just no manual tuning and another CCD. 3rd gen maybe some more of the negatives can be worked out. Think about it we never even got the 5600X3D that was nearly released. This time around 2 of those make up a 7900X3D so wouldnt be surprised if in another 6 months we see that chip. The biggest issue with both dies being v-cache is that youd lose out on a lot of top end performance. These cores dont boost to where their X counterparts can which still matters a lot. Youd be paying more for a cpu that would perform worse pretty significantly in things that cant utilize the v-cache.

[@cheesyriceo4]

@@JoeWayne84intel is coming out with chiplets too.

[@ADB-zf5zr]

I kept on hearing "Game Bar" in this discussion, Looked it up and remembered that it was one of the first things I nuked in my W10 installation.! Can it actually benefit me in some way as a "PC only" using gamer.?

[@victorbart]

I choose the 7950X over the X3D for my video editing workstation and gaming system. 100 euro cheaper and fast. I don't think X3D will make my system faster. I rather wait to upgrade my AM5 in the future to see what they do with X3D

[@BaneSIlvermoon]

The whole ongoing dialogue around this is weird to me. For me personally, as a gamer, the whole separate CCD was the selling point of the processor for me. So when there IS something else running on my system that's taking up significant CPU time, there's somewhere for it to go that doesn't interfere with the game

[@jiazhoushen5447]

Just something in case people don't know: cross-CCD data transfer is not fast. It is extremely SLOW. It adds another penalty of ~60ns, which is more or less the same amount of latency for assessing data from a DRAM. So DRAM latency 60ns + L3 latency is the time for a core to get the data ready (ignoring L1 and L2 they are so blazingly fast). Which is the same amount compared to the cross CCD core to core latency (about 80ns, roughly 2xL3 latency + cross CCD penalty). Zen 3/4 does not fetch any data from the other CCD's L3. Whenever a cache miss happens that CCD requests data directly from the DRAM. L3 is only shared within a single CCD. This has been tested by many people and confirmed by AMD. Parking cores isn't different from setting core scheduling priority, which windows has done already for decades to schedule dual CCD Ryzen and later Intel 12th/13th gen CPUs properly. Parking cores just allows you to set this priority manually when you believe Windows does the wrong thing (apparently it can't always be right because this is no way for a OS to know if a running app is cache sensitive or frequency sensitive in real time). And people have been using 3rd party tools to do so for 12/13th gen Intel for quite a long time because windows is dumb as always.

[@jiazhoushen5447]

Some people lacking knowledge of the architecture has delusions of these CPUs. Just like the recent PCMag review saying 7950X3D has significantly better igpu performance compared to non-3d version, which is completely wrong. Apparently they have fixed the review. The iGPU is in the IOD die and can't read data reversely from L3 on a CCD, this is not how Zen sereis' IOD works. The iGPUs for 3D and non-3D version have identical performance in reality.

[@saricubra2867]

@@jiazhoushen5447 I haven't watched anyone doing DDR5 vs DDR4 for the Intel iGPU on Alder Lake/Raptor Lake

[@jiazhoushen5447]

@@saricubra2867 Raptor Lake is not very memory hungary as Zen 4 as hardware unboxed showed. 13900K and 7700X perform almost the same at 6000C30 but if you use 4800 DDR5 7700X suffers a much larger performance hit compared to 13900K perhaps thanks to Raptor Lake's large L2 cache. Also 6000C30 on 13900K performs similarly to 3600C16. Only in some newer bandwidth hungry titles like spiderman DDR5 performs much better compared to DDR4. I remember 13900K with DDR4 even beat Zen 4 with DDR5. In my opinion Zen 4 is so bandwidth hungry that it should pair with 3D v-cache by default. Zen 4 is an architecture designed for server market and designed to have 3D v cache from the very beginning because their crappy IOD bottlenecked the whole bandwidth/throughput. Coming back to 13900K, you can probably gain another 10% overall in games going from 6000C30 to 7200C32, you will need a pair of hynix A-die RAM. From 7200 to 8000 it starts to vary depending on how good the motherboard is and how good the quality of the IMC of the CPU is. In order to know how much you gain from 7200 to 8000 you can check frame chaser's video, not too much. I'm not sure if it is worth it though because for people without any memory knowledge XMP works petty for 7200. Beyond that XMP is no longer reliable you will have to tweek the voltage by yourself. If you want to make a decision: choose the cheeper RAM sticks depending on your market price. If 3600C16 is cheaper than 6000C30/32, the DDR4, otherwise DDR5. 6800C32/34 hynix a die currently is the same price as Samsung B die for people rocking at 4000C14 DDR4 but B die is only 2x8 while D5 a die is 2*16 so in the fast RAM regime there is no reason to pay for a pricy DDR5. In a word, choose the cheaper one.

[@aggies11]

I think a lot of confusion stems from the fact that people assume the windows scheduler is doing a lot more then it is. They imagine it's working much like how a human would logically assign cores/ccx on a asymetric vcache design, which we can do very straightforward, while in reality the scheduler is much more "dumb". Don't get me wrong, a process scheduler is a non trivial piece of engineering but it's easy to naively assume that it does more then it can. With that in mind, you can see that AMDs current strategy is one that is designed to work with this "dumb" scheduler in mind, and work around/within those limitations. A "thread director" situation would be more preferable (from a purely idealogical, not practical standpoint), but even intel's thread director and microsoft's concessions to the scheduler, still don't lead to a 100% ideal handling of Pcore/Ecore systems. So from a "what would work for the majority of cases" in a set it and forget it sense, this is a pretty solid first attempt that accomplishes the majority of what most systems would need.

[@Misterz3r0]

Sure, but that wasn't a problem until you have heterogeneous chips with cores that perform differently under different loads. The computer IS supposed to do these things automatically. That is the whole purpose of the scheduler. AMD was responsible for working closely and collaborating with the kernel developers of Microsoft and Linux to get the chip to work seamlessly. Instead, they offloaded it to consumers to figure out. That is a significant problem. A "first attempt" should have stayed in-house until they got it running, not sold as a finished product to be beta-tested.

[@CalebBriles]

Wendell and Gordon is one of my favorite combos, this video is a gem for a compsci person like me. I really hope we see more collabs with these two.

[@Dr._Nicholi_Rasmuson]

I've been a fan of Wendell since he was just a pair of eyes peering over monitors. He improves everything he's in.

[@cromefire_]

I mean ideally you have something like thread director, but instead of heaving to park cores, which let's be real is a solution to work around Microsoft's seemingly bad scheduler, the scheduler just asks thread director whether the process (game) may be moved to the E-Cores and thread director then could just discourage that request. That way you could have for example OBS running on the second CCD without having to dead that the game will jump over to the second CCD. Really interested to see what Linux will do with thread director, if it finally arrives, with it's much more modular (and probably superior) scheduling.

[@giorx5]

There are some people that always need to find something to complain, so don't mind of those. This CPU is MEGA both in performance, thermals and power efficiency. So, price is high as expected for a halo CPU. On the CPU scheduler side of things, I own Dirt Rally and Dirt 4 and both induce stuttering on my Ryzen 5600 CPU that is vanished once I turn XBox gamebar on. So, the windows cpu scheduler is at fault there imho.

[@paulschaaf8880]

I think we really just need all apps to have an option to prefer Vcache or non vcache and have the OS abide by that preference. Maybe that option already exists and just isn't being used consistently, no idea. I will probably wait for the 7800X3D, but I might still decide to get the 7950X3D despite the chance that I might sometimes get some funky performance if an app uses both CCDs sometimes. I imagine that this problem will be less of an issue over time as either apps get updated to specify a preference for 1 or the other or the OS gets updated to make better choices about it when it's not specified.

[@yol_n]

If you only game the 7800X is the best, the 7950X3D is for people that game and do productivity

[@MTDucas]

what i want to see form you guys is to run a game on the 3D chip and at the same time run in the background discord, youtube, recording software, and other apps people normally use....then well see how good the 7950x3d and 7900x3d are for the money

[@miyagiryota9238]

Great discussion with Gordon and Wendell!

[@slimjimjimslim5923]

Awesome the two experts I trust the most are meeting. 😊😊😊 other people are always biased but these two always give balanced review

[@billybigbollox5470]

What a load of crap brother,bald guy is super amd fan and we all know it

[@mysterium364]

I wonder if this core parking will work within a virtual machine. What if I create two gaming VM's each with 4 3dc cores and 2 normal cores. Will the parking still work?

[@profosist]

unlikely until the host is setup to deal with it. From my understanding at this point avoid this imbalanced CPU's for server stuffs or at best you can manually assign certain thing to certain cores. If anything real servers just keep moving to more and more cores rather than different types, unless you get really specific.

[@anthonyrizzo9043]

I would just get 78003d if your mostly gaming, if your doing gaming and heavy workloads I would just get a 13900k and call it a day, if it's not fast enough, Intel overclocks. As far as these guys not understanding people's complaints, the complaint is people thought they were getting 16 cores with vcache and that's not really what your getting, it's advertised that way and that's what your paying for, but evidently and can't make that work correctly. It's a valid complaint to be honest.

[@Lemonpinesol]

Yeah the amount of mental gymnastics everyone is going through for a pretty straight forward complaint is pretty amazing.

[@anthonyrizzo9043]

@@Lemonpinesol yea it's like they want to defend it, and PC world says games don't use more than 8 cores and I get that, but it doesn't mean they won't for the entire life of your owning of this high end cpu

[@joeykeilholz925]

I'd just go with amd either way and have far less heat and power usage with similar to better performance

[@anthonyrizzo9043]

@@joeykeilholz925 and there's nothing wrong with that, they're all good cpus at the end of the day.

[@DaveLanki]

If Amd had never solved their low frequency boosting speeds I think there would be less of an issue, but people feel like they're losing out by games choosing to run on the slower ccd with vcache. I guess it's just a mindset of worrying about leaving performance on the table. I wonder how many of them end up gpu bound a lot of the time anyway

[@profosist]

I think AMD is working on is as this time we get proper turbo and a second ccd unlike the 5800X3D. I imagine the 3rd version will be even better. Not easy developing completely new tech. Its similar in a way to HBM but this makes a far bigger difference. lots of people dont understand how to balance a system and likely run into bottlenecks without knowing it

[@georgejones5019]

People buying this for gaming are just blowing money. This is a workload card. I'll await the 7800X3D.

[@martinopala467]

Money is just paper. Im still satisfied with the performance i gained over a 11900k

[@saricubra2867]

@@martinopala467 Basically anything from the mid to high end right now is better than an i9-11900K, even the 2021 i7-12700 is way better. i9-11900K is just one of the worst Intel chips ever since the Pentium D or the 9700K (the only desktop i7 in history without hyperthreading), 14nm hurted it so badly as a flagship product.

[@bodasactra]

AMD does not have a version of Ryzen Master for Windows 11.This CPU needs user, Gamebar, Windows, and game dev support to function as intended. Sorry guys, sounds like a true shit show one should skip.

[@celdur4635]

All the people freaking out won't even buy the CPU LOL. Its just internet being internet.

[@profosist]

hey I got one coming! Always better to have hands on whenever possible. You also have a better grasp once you use it yourself for your own workloads for a while.

[@neutechevo]

Hello PCWorld. Sleeping Cores or not, Can we see Wendell on the Full Nerd Live? It will sure be nice!

[@larrypaul2462]

"Parked" is a term only old timers remember as we used to have to "park" our hard drives read/write heads (especially when the drive or machine was moved) back in the day. Needless to say, it didn't disable the drive.. For lack of a better term when talking about CPU cores being "parked" Parked fits, even if many don't understand it don't mean disabled. For those that may be unaware. Parking of read/write heads on spinning disc became a thing of the past long before spinning disc began being phased out, they just parked those heads automatically rather than the user needing to park those read/write heads via the "park" command..

[@thenerdysk8er]

People get so worked up on stuff like this because they don’t look at pc parts/tech as tools and expect it to conform to their wishfull thinking of how things should be.

[@profosist]

I always say gotta review something for what it is not what you want it to be.

[@joeykeilholz925]

I mean.... Even cache on X3D should really be coming eventually I hope

[@joeykeilholz925]

I mean... For a very long time, the CPU would just do what it does and you wouldn't have to worry about uneven cache sizes and all this crap. Hopefully they get that fixed next time

[@thenerdysk8er]

@@joeykeilholz925 there was far less complexity involved yes and also less media outlets to cover the topics. But i think that in general none of us really know if the teething issues right now won’t pay off in the future, it’s the engineers and developers that know. And don’t get me started on the industry mambo with exclusivity hardware deals that leaves performance on the table. AMDs latest iGPU is a great example of this. Time will tell and if this cpu doesn’t do what you need it to do there are and will be others.

[@mattgreenfield8038]

Another dud from team red.

[@ole7736]

It would be great if you could easily reserve the V cache die for just the game and let everything else including the OS run on the regular die.

[@bartios]

Maybe that would be great, maybe the overhead of system calls from the game needing to jump across dies to reach the OS/kernel would be greater than whatever you gain.

[@_nom_]

@@bartios Unlikely, system calls aren't typically called the number of times to create a bottleneck. The winsock stack and desktop environment could happily run on the 2nd. Unfortunately the cpu scheduler in windows is too simple

[@Freefallendless]

You can make it do that. Set preference to frequency I'm bios, then use process lasso to set games to cache.

[@DJaquithFL]

I think you guys are really forgetting the audience here for this CPU, they are not the average user that you keep on throwing them into that box . People spending this kind of money absolutely are nailing their CPU to the wall. You're kidding yourself if you don't think they set Windows to "Performance" and they maximize the BIOS! Competitive FPS Streamers are running north of 200+ FPS and can't stand stutter, and they're absolutely are pushing their CPU more than your insinuating. When I'm streaming I am literally nailing my i9-13900K to the wall. CPU-Z JJU8GT Bottom line run some real world benchmarks on AAA games, streaming to multiple sources, multiple monitors .. In other words, like a real professional with set up there PC looking at a $700-800 CPU.

[@big_bird8597]

yea i think homie thinks im going to be using this for excell spreadsheets at my accounting job

[@SirMo]

7950x3d give you the best of both worlds if you do Gaming and Productivity. One CCX is optimized for games and cache heavy workloads, while the other CCX is optimized for frequency. It also keeps the costs down. Because 2 v-cache dies would cost more. It's a win win, which ever way you look at it. If all you care about is gaming, then 7800x3d is your CPU however.

[@danchan5620]

It is actually 13900K that gives you the best of both. Even 13700K is better performing per dollars. Intel hardware thread director solves all such extra layers and latency with gamemode core parking. This is the problem content creators are lazy to test and try to theory themselves out with videos such as this

[@SirMo]

@@danchan5620 7950x3d is faster in gaming, easier to use in productivity, AVX-512 support on all the cores, SMT on all the cores . Twice as power efficient. I fail to see how Intel comes even close in competing.

[@saricubra2867]

@@SirMo You are considering 100% multithreading workload and power efficency there, not multitasking and other stuff. By the way, Intel iddles at way lower wattages. My i7-12700K for example iddles at 5 watts (also blows the 13900K's perfomance per watt out of the water too). Power efficency is kilowatt hours, not watts. Also Intel has an IPC lead vs AMD, it doesn't hurt saving more watts with a 300MHz reduction in clockspeed and still matching AMD's singlethread results. The thing is, both Intel and AMD are running way outside the efficency window with absurd clockspeeds.

[@danchan5620]

@@SirMo More efficient yes. Other than that, 13900K is the better cpu. Amd lost this round. I hope content creators do a video on tuned cpu and realworld multi-tasks workload to really represent what Intel can do

[@SirMo]

@@danchan5620 Nice troll. Flight Simulator runs 30% faster on 7950x3d. 30%!!! Intel is back in the rear view mirror.

[@teddyoflove5897]

If you don't like this processor wait for 7800x3d, it will be very close in perfomance. But if you want best perfomance and you are building high end system that already costs over 3-4k then just pay this extra 200 which is nothing for you and get that extra higher clocks and in the future hopefully 2nd ccd can be used for other processes to give full 1st CCD for games. Anyway if you are hard capped on money for your build should not even be looking at this CPU, if not then pay these 200extra and get slight edge over 7800x3d.

[@gurugamer8632]

When is zen 5 due?

[@MasterKoala777]

This explanatory video will do a lot of good. Plus seeing Gordon and Wendell in a discussion is always a treat. I did see Wendell’s video at Level1techs where he said some cores were “disabled”, and I cringed because I knew some people would misunderstand. It’s like talking in a classroom to students, but in this case there are thousands, and some words need to be chosen very carefully (or explained further). Remember, not everyone can keep up with you guys. But this is in hindsight where everything is clearer. The takeaway for me is that the next time a non-symmetric (a.k.a. heterogeneous) CPU is launched, tech journalists and explainers should explain more slowly and carefully 😉. Cheers!

[@jazzochannel]

yomoma is always a treat in hindsight where everything is clearer. I did see Wendell’s video at Level1techs where he said she was "disabled" and some words need to be chosen very carefully. Remember, not everyone can keep up with you guys. The takeaway for me is that the next time yomoma is released, tech journalists and explainers should ride her more slowly and carefully 😜. Cheers!

[@profosist]

but people want info NAO and zoom fast. Hard to know what will and wont be picked at. I for once think its a bit odd because much of this would apply to Intel chips and even other multi-CCD AMD chips. I think thats why this caught the tech press off guard.

[@TheJacklikesvideos]

omg i see so many comments equating parking with disability.

[@dunastrig1889]

Outstanding, I always love Gordon/Wendell talks. 👍

[@WereCatStudio]

I think the issue is that people expect the separate dies to behave as if they had 2 different PCs. For example the often argument with people is that they are expecting to restrain the game to the 3DVCache CCD and OBS to the non-3DVCache CCD which sounds possible in theory until you realize that OBS and the game have to interact with each other so you would have to have cross-communication between CCDs which would make the game perform less. You could in theory assign the 2nd CCD to VM and run a capture card to do this with VM but I am not entirely sure about how this would behave as well.

[@franklinlora7380]

Thank you for the explanation! Very clear and easy to understand!

[@soapa4279]

The people who are confused are the mass amount of new PC gamers who has zero clue on how actual PC hardware works. Some of them don’t even know the difference between cores and threads, so how you expect these folks to understand 3D V cache 😂

[@big_bird8597]

yup people like me are used to plug and play cpu and now with this one theres are these small print disclaimers AMD is slightly being coy about

[@soapa4279]

@@big_bird8597 Yeah no lie. PC gaming has been a lot easier to get into than it ever has, but it can still be complex to understand

[@Lemonpinesol]

good thing youre here to enlighten the masses with your gigabrain

[@soapa4279]

@@Lemonpinesol Come on, you can be more sarcastic than that. Try harder, F for effort.

[@Lemonpinesol]

@@soapa4279 how egotistical

[@doculab3d]

This conversation furthers the community at large. Thank you.

[@joshua_1_9__]

I just don’t trust it, yet. Not saying it isn’t great and all that, but just need to see some more real world results before I commit to it.

[@1234Jaytay]

This is the same thing that has always happened with AM4. Except now, one ccd(8core chiplet) has all the v catch. Therefore, the system prioritizes that ccd until it needs the secondary ccd(8 core chiplet). All-around champ. Gaming and productivity alike.

[@TheJacklikesvideos]

i'm all on board with this processor, but he doesn't sound like much of a factorio player. the whole point is to build a massive factory and the cpu bottleneck is the deciding factor in how you build late game.

[@raynathanielcaparas8064]

once gamebar detects a new game for the first time, will it not be confuse anymore the second time?

[@sacamentobob]

The other thing is that the 7950x3d is not a pure gaming part. If you're after a pure gaming cpu, wait for the 7800x3d. Then there'll be no such confusion. AMD clearly said that the 7950x3d is a best-of-both-worlds type cpu, gaming + work use, and they also explained the cost-benefits of not implementing V-cache on both dies.

[@BlueRice]

people who plays counter strike will understand hardware matters. i tried it on many hardware and a lot of it doesnt perform as good. i believe its due to render latency, monitor input latency then last would be ping. i test 7950x3D and its one of the cpu that works well with counter strike. i came from other cpu which is over 4.2ghz and it perform bad with counter strike. but the weird thing is, counter strike runs on non vcache ccd default and it perform really good. this is what im puzzle. when i play other games like dota that FPS is not important it runs on vcache ccd. i dont want to mess with lasso since on default, the game runs well for me.

[@sblantipodi]

what an awesome couple and what an awesome content.

[@thumbwarriordx]

lol I wonder if the X3D's V-cache would let me finally resume my X3: Terran Conflict playthrough. Used to be single core clockspeed was the only thing that brought it back. The game is single threaded and I have built a factory and logistics empire that has ground the simulation to a 15fps halt on my previous THREE computers.

[@IronCondorSimulations]

It is rare to see CPU Tech Journalism this good. Every channel with the same crap repeated over and over again. Finally, some context.

[@onizendor8839]

stopped the VoD at 4:27 to write this ... how to explain a Ryzen CPU with 2 CCD's , and especially the new 7950x3d / 7900x3d , for Dummies ( car analogy ) : imagine 2 parallel streets with a river in between. in the olden days, you used a ferry to cross that river. AMD uses those fancy stuff called infinity fabric as bridges, that takes a lot less time ( still takes time though ) . and what about 3d-Vcache? well, the street on one side is wider, so even trucks ( more cargo = more instructions / code ) would fit, instead of a car. ok, but why does AMD make me use only the wide street? well, it's not really AMD alone that does that, but let's stick with that for now... and to answer that question you need to understand that about every game and many programs that can use ( utilize ) many threads / cores, have one main thread for ease of explanation, that one main thread makes sure, everything is executed in order. so... if car #3 changes to the other street, it will slow down everything, because a) it takes time to cross those "bridges" and b) car #1 ( main thread ) will still make sure that the instructions / code on car #3 will be executed after #2 and before #4 ... honking across a river to signal "instructions ready to execute" also takes some time ;-)

[@TravisHershberger]

All this makes me glad I got used to core scheduling eccentricities with SGI servers back in the mid 90s!

[@tIhIngan]

People think more cores is more better for everything. It isn't. Some things benefit from more cores, yes. Some things benefit from more cache, yes. Some things benefit from higher clocks, yes. Define what YOUR needs are, and go get the part that is optimised for YOUR needs.

[@johnpaulbacon8320]

Thanks for this informative video.

[@AshySamurai]

It is a great video and it was really educational for me but when you are saying "people don't trust to run it automatically" - I'm as a hosting administrator (so I had a lot of windows not only as a average user) do not trust windows and MS in general. They screw up on every step they can.

[@joeykeilholz925]

Yep. If you value stability whatsoever, disable windows update and only update 12 months after the fact if you do. Running a 2021 windows 10 build and it's rock solid.

[@vsuhden]

A closer collaboration between hardware and software development is approaching

[@Cheeseypoofs85]

that "coming to america" reference was on point. lmao

[@KeinNiemand]

I don't want 16 cores I want an 8 core with even more cache maybe 2 layers of extra 3d vchache stacked on top of each other no idea how that could be coolded tough, or even a second chiplet with only cache maybe as some kind of L4 cache since cache on another chiplet would have higher latency still probably less latency then going to ram, I want so see how much faster cache limited games like factorio would run if you could have like a few GB of cache to fit the entire game into cache. Also das why do both intel and AMD rely on core parking to make sure games run on the right cores, why don't they just use core affinity.

[@__--JY-Moe--__]

whats so confusing? 1. jump in 2. start.. 3. stomp it & take off! 4. the end!

[@JuanCAraujoS]

Red shirt + landing party = You're dead! (Is better to have Main Engineering as a background) Maybe we're just suffering from banchmark saturation that results in this idiotic obsession with FPS and scores. The 7950X 3D looks amazing for someone who streams while gaming: The V-cache CCD handles the game, the other one handles the stream. At least that's the way I see it .

[@xRaptorScreamx]

It will be could when there wont be a need for Parking, and have 1 ccd for games, and stick them there. And the other ccd could be used for background or streaming, at the same time

[@SuspiciousPixels]

A few questions if Gordons and Wendell's community can clarify. With the 7950X3D is a complete Windows Re-install needed after installing the CPU? Does the 7950X3D benefit more from the scheduler in Windows 11 than 10?

[@TheJacklikesvideos]

idk about needed, but if you're upgrading CPU or mobo, a fresh OS install is a good idea in general.

[@macd7743]

Wendell's endorsement is enough for me.

[@LoupFrileux]

At the end of the day, other than vanity, you pay premium for a 16 cores to be able to use 16 cores. If you're a gamer (which most enthusiasts probably are) and cant have a VM in the background, or any other heavy workload, on the other cores - you might as well wait for 7800 X3D. Which everyone is guessing was only delayed so the 7950 X3D could be sold at premium price to unaware customers. Not even mentionning 7900 X3D aberration. The pricing, the delaying, the slight obscuring to unsuspecting customers that the better CPU is yet to come in a few weeks for cheaper. I had a good image of AMD. But there are too many recent signs of tricking customers, with wrong delaying/pricing (X vs non-X, 7950 X3D vs 7800 X3D). These processors are not even the game changers their marketing teams wants you to think, considering you only gain a few FPS compared to a way cheaper 5800 X3D build.

[@mikebruzzone9570]

Right Wendell at 27:10 u said wait for 5800X3d and in my related comments one down look for 5900X3D refresh data that I said much earlier in the broadcart and even gave some news that of course PCW and Level Tech will cover? mb

[@Inb4ify]

It's honestly been pretty frustrating seeing just how much people will run with a mistaken notion. What is going on here isn't even that complicated

[@andrewsearle5845]

Hi guys,great info.Are you saying though that less cores is better,or is that dependant on what the computer is doing?

[@cjroush8563]

Wendell we love you!!!!!

[@vasudevmenon2496]

Last time when i used to unpark core was on win 7 to keep the core from going to sleep on light threaded games

[@kirkanos3968]

Great video thanks for all the info

[@earthtaurus5515]

What's there to be confused about... the cores aren't turned off when gaming on the 7950X3D lol 😶. @10:00 exactly! people who are merely web browsing and leave a browser open whilst a music streaming app is running when they go play their favourite games do not need a $700 dollar CPU lol... 🤦🏽‍♂

[@keyput415]

I watched a 30 minute video on this topic and still don't understand.

[@patric.rogers]

So I bought this you for future essentials, and well I would rather have 16 cores for gaming and extra needs while gaming, streaming and all one system yk? But honestly was it’s. Good idea? I have 7950x3d and 3070ti ftw3 ultra I do like the cpu but unsure if it’s the best of cpus and all?

[@GooberBrainTrollingCorp]

My take, is that this is a decoy product to get enthusiasts to buy now, and buy later on the AM5 platform. They won't make the 7800X3D perform close or worse than it's predecessor. There's performance sitting behind the curtain.

[@DaPoets]

Always happy to see you guys!

[@DaPoets]

@@pcworld I'm actually looking for sponsorships now for it 👍

[@Multimeter1]

AMD I P I P -FrameChasers channel.

[@TrueThanny]

Have to disagree that parking is better than affinity. It just operates on the assumption that the user is _either_ playing a game _or_ doing some kind of production work. That's not how reality works. Somebody with a 16-core processor who wants good gaming performance is likely doing some kind of production work and gaming to pass the time. You don't want half the CPU parked, and you definitely don't want that to be the mechanism by which threads are scheduled. You want your game to run on the v-cache cores only, and everything else to run on the normal cores only. Core parking is the least-effort solution. Automatic affinity masks is the proper solution, which would take more effort to implement correctly. It's already been demonstrated that the least-effort solution doesn't work for some games. Ultimately, anyone buying the 7950X3D or 7900X3D should get Process Lasso and learn how to use it. You'll get better results, especially when multitasking - which is something you should be doing if you're buying these parts.

[@hadoukenocx4746]

The Problems here we have With AMD Infinity Fabric and the 3DV are if you have BIG High Res Loading Texture 4K or 8K, and you play Modern games your 0.1% and 5% lows are way down compare to the 13900KS + 7600Mhz 8000Mhz with XMP on. That means if you are an AAA 4K modern gamer or 8K I recommend 13900KS + 7600Mhz DDR 5 with XPM or 80000Mhz XPM with low CL. many TEC tubers don't get that but the loading texture on the AMD Ryzen CPUs aren't good that's an Architect fail, and I have both Systems here. Believe me or not, or Test it in Modern AAA Games your self and you will see it.

[@mikes567]

o shit gmu and the man behind the monitors

[@TheMessiah1337]

So a weird thing i found out today was when i was playing sons of the forest i checked what my performance would be if i only used the 3d v cache cores and suprisingly enough i had lower performance then just using all cores available, it was about a 20-30 fps difference.

[@balls2thewall724]

if game bar knows what processes are video games in order to park the frequency cores when those processes are focused, then why can't it utilize an affinity mask to make those processes only run on the vcache ccd? the problem with parking the frequency cores is that i'd like them to still be usable for non-game and system applications while im playing a game.

[@AsthmaQueen]

my hope is that gamebar alone with cache preference, will do its best with background applications to prevent spill over, even if it means gamebar is effectively masking the game process temporarily or without user manually having to do it. The background applications are the thing that can spill over.

[@gurugamer8632]

For video editing in apps like adobe premiere , after effects and DaVinci Resolve which CPU is the best for performance? I9 13900K , 13900KS or AMD 7950X or AMD 7950X3D?

[@TheKazragore]

Depends on whether or not you also need a space heater. If you need your room well done, 13900k. if you'd prefer something more manageable, 7950x. If you're actually sensitive to temperature or are easily discomforted, 7950X3D because it's the most energy efficient of all of them. All these CPUs are very high performance.

[@gurugamer8632]

@@TheKazragore 7950x3d good for video editing?

[@ChiquitaSpeaks]

Want to build a rig now for when games finally get more based in utilizing physics so my decision comes down to what about physics and games benefits more on the chip with the cache or the better frequency hopefully get that answer now so can make the right decision for the future… Anybody know?!

[@ashleyjaytanna1953]

Those parked cores will result in the vcache to do some fancy acrobatics

[@kelownatechkid]

This is the dream team right here!

[@lubossoltes321]

So basically developer supplied profiles for numactl are the future ? What a shocker ...

[@denvera1g1]

15:56 Imagine if developers could flag speciffic threads for cache VS frequency. Imagine an AI hardware scheduler that does all this testing live on your system, figuring out that, hey this thing keeps having to hit RAM try moving it to 3D cache. Hey this thread isnt using much cache throw it on the faster cores Hey this one thread on the faster cores are doing an aweful lot of talking to threads on the 3D cores, move them both back and forth to find which cores are best for this pair of threads, or if having them split between CCDs is the best solution. Edit, this scheduler may work with different CPU designs, from Intel P/E, or ARM big/med/small

[@sojirou]

Great video explanation. When techies need tech support, Wendell comes to the rescue.

[@RobertJohanssonRBImGuy]

Plug and play pretty much. No confusion.

[@chrisj2024]

Hardware has to come FIRST, then software can follow! Amazing how pcworld/ Linus tech tips, gamers nexus, jayztwocents, all failed to understand this when talking about PCI Express 5.0 and future 6.0 You NEED a massive base of HARDWARE first for the software guys to figure out a unique ways to utilize this hardware. Look at the PS5 game developers using the audio hardware chip to process polygons. HARDWARE FIRST ! Then software will find a use for it!

[@HitmanSC2]

Love the 7xxxx3d talk between you two

[@GamingOnArc]

It's quite simple Mr Gordon. If you buy a 7950x3d or 7900x3D for gaming, 1 whole CCX package will park and the CPU will go from 16 / 12 cores to 8. Meaning buyers overpaid for a 7800x3d. Transversely, if you buy a 7950x3D or 7900x3D for production work; none of the apps will benefit for 3D V-cache and you should have just gotten a 7950x or 7900x instead. In a nutshell, 7950x3D and 7900x3D are complete wastes of a products and it is pointless to pay for the x3D feature these have. 7950x3D and 7900x3d are just money grabs and proving this is that AMD released them 1st and withheld release of the 7800x3d till they made enough back on the R&D for the 7950x3D and 7900x3d. They knew people wouldn't buy either if the 7800x3D was released at the same time. Also at 7:23 when you address "People overthinking", i would interject that this is like the iPhone vs Android where iPhone you use them as they are and Android pushed the ability to customize in their marketing as well as word of mouth by users. This applies as much to AMD who has since the launch of Ryzen, touted overclocking ability across all of their products and now, people expect it. Problem is, when it that when products don't meet expectations; the go into torch and pitchfork mode on social media. So i personally blame AMD for creating this culture "excessive expectations" and "Excessive Competition" within the PC enthusiasm space. The amount of entitlement and football team mentality with this space is pure toxicity.

[@POVwithRC]

Everyone survived Intel's plunge into P/E. They'll survive this too. Apoplexy has become far too contagious 🎠

[@saricubra2867]

Intel has Thread Director, is not the same thing.

[@POVwithRC]

​@@saricubra2867 Didn't say it was.

[@slimjimjimslim5923]

@@saricubra2867 agreed but I guess AMD didn’t have enough time nor resource to polish it. And also if AMD worked harder to convince people of how their dual CCD works, people might not be so confused now. But that isn’t an engineering issue, that is a marketing issue.😊

[@jonathansaindon788]

@@slimjimjimslim5923 Yes they could market it better. But at the same time… I feel a lot of people are complaining about very technical things that they do not fully understand and that won’t affect them anyway…

[@RyTrapp0]

...yea, if only AMD could make their CPUs as 'polished' as Intel's 13th gen on Windows 10...

[@reijhinru1474]

I really dont know what to do now… should i keep my 7950X3D or send it back and get a 7800X3D. All im doing is playing Games, having some browser tabs open and discord, teamspeak, other programms like steam and so on.

[@super2k128]

For games only, return it and get the 1ccd x3d

[@reijhinru1474]

@@pcworld thanks ill send my cpu back and ill get a 7800x3d now.

[@TechRIP]

@@reijhinru1474 Before you send it back, remember that if you get the 7800X 3D, you'll only see performance gains in games that want cache. You'll lose performance in games that want higher clocks. The 7950X 3D can do both from switching CCD's. But, if you don't care about the best gaming performance, then all you really need is the 7700X. It's only a little slower than the 13900K but almost half the price. But, any CPU other than the 7950X 3D will have worse performance in either high clocks or cache dependent games.

[@Pleplerhep]

Use my pc as is LOOOOL, it would be a slideshow.

[@JonathanMaple]

I'm glad yall are having this conversation its really interesting to see people that have never had to know how the kernel or internals of the cpu work getting a splash of cold water. This has been fascinating release. I instantly saw how these parts should work because I worked on getting Affinity Scheduling working on Bulldozer for Crays Blue Waters before the Linux Kernel fully integrated AMDs threading issues. We had to provide this for the scientists needed to make sure that the shared Floating Point Unit could only get scheduled by one of the threads and keep all INT based operations on the other thread. This is that all over again, the thing I don't like about the Parking and balanced is if I want to contain all my streaming apps on the core away from the game so they don't context switch. On AMD's post I was wishing that there was an interface to hint or recreate process lasso in Ryzen Master. However this requires a lot of bench-marking and knowing the behaviors of each application but that would be outside the realms of what most are willing to do.

[@mobeenranachannel]

I dont get the fact, people are ok with Intel p core and e core but not with amd v-cache core and non v-cache core

[@slimjimjimslim5923]

It’s because intel did a lot of work to demo and show people it’s ok to rely on thread director. AMD said 3 lines about how the scheduling work and now we have to rely on reviewer to test and understand how it works. Just poor planning and marketing on AMD’s part

[@mobeenranachannel]

​​@@slimjimjimslim5923 agreed. Also i think it may be bad publicity from the other camp because i dont think normal people open the task manager to see which core is running or not.

[@slimjimjimslim5923]

@@mobeenranachannel also people gave a lot of shit to intel for having e cores which they called fake cores lol. So now that AMD made an uneven CCD, they feel their trust has been betrayed. But people need to understand one way or another, we are all heading to heterogeneous cpu with P and E cores or vcache and no cache CCD. It’s the way to optimize for multi thread and save power

[@saricubra2867]

@@slimjimjimslim5923 Intel's Big-Little alone is not for power efficency, it's for perfomance per mm2 of die space for more cores on Ring Bus. IPC and process node do play a huge role on perfomance per watt, also clockspeeed. Apple M1 lineup has the best perfomance per watt on the market because they have WAY higher IPC than Intel and AMD, made on TSMC 5nm and of course ARM needs less transistors than x86 for a job (but more memory intensive). Many of their chips clock at 3.2GHz which is a sweetspot for sillicon, that is also a common speed for laptops.

[@slimjimjimslim5923]

@@saricubra2867 cool, learning something everyday. 🤔.