They were upfront about the split-dies from the very beginning. This is how it was designed to function and they made no allusions to the latter, I'd hardly call that a scam.
They weren't upfront about how jank the system was for making sure games only run on the vcache CCD. In this video the non-vcache CCD didn't get used at all until Gordon fully loaded cinebench to 32 threads (that's sub-optimum to say the least). Intel did the engineering to implement a hardware scheduler and worked closely with Microsoft for years to get big little working in windows, AMD is just like "mIcRoSoFt GaMe BaR WhItE lISt AnD cOrE pArKiNg!1!!1".
reminds me of hard drives that give you 1TB but than its less than that. There is also advertised ram speeds which don't run out of the box, and you have to setup yourself in the bios.
@@FakeGordonMahUng Outrage Pony™ riders after Ryzen 5800X3D launch: how dare you not give us 3D vcache 5950X3D, AMD!!! Outrage Pony™ riders after Ryzen 7950X3D launch: how dare you give us what we asked for, AMD!!!
Loved Wendel's deep dive but also love this super simple "live demo" of it just working! (got it it's not perfect but it's all of 48 hours old and working pretty darn good for something still so uncommon)
How is it just working? Did I miss the part where he said you can't start/stop threads using the fast CCD while you game on the cache CCD without impacting gaming? He just said you can't do that. Ecores do that.
@@fhpchris like I said it's not perfect, but it shows that the other CCD isn't just "shut off", it prioritizes the CCD with extra cache in games and prioritized the higher frequency CCD with CineBench (which uses all the frequency it can get but doesn't need the extra chache), then when he "ran everything" all the cores were used. Again I admit it's not perfect, we can't one-click-pin tasks to whichever CCD we want but the rough functionality IS there. E and P cores aren't perfect either, I'm pretty sure some comon hypervisors still have issues (or just don't even try) to distinguish between threads and try to treat them all equally. Admitedly these chips would probably have a similar problem so, like I said, not perfect but a really good start for a relatively novel CPU design/chip technology on a new socket/platform that's just a few months old. I guess what I'm saying is that this could have easily been so much worse and in my opinion AMD pulled it off pretty well! In the end I think we the consumers are winning, AMD has some benchmarks they can point to where they "beat" Intel so Intel can't get lazy and price gouge like the Skylake and 14nm+++++ debacle.
@@fhpchris The main misconception that people still have is that 2nd CCD is disabled period and you cannot access it. Like on an original Threadripper with Game Mode where you shut off a CCX/CCD and reboot and Windows literally cannot see it or use it.
@@FakeGordonMahUng The reason that game mode disabled the cores and not parked them is because there is no hardware scheduler or affinity keeping the processes on that CCD. Its very possible as long as the cores are active and not parked or disabled that the threads could be assigned to the wrong CCD. If you let threads share the cache CCD you can hurt your game performance ( very much so with R23 threads!) Ryzen is also limited in memory bandwidth with only one CCD active too. Ecores don't have this problem. Many people expected this to work like a 13900k does. ( ~200 cheaper at microcenter) My 7950x doesn't even post at 6000 mhz without changing every option in my bios ( and its slower than my Skylake X with the same GPU due to C state boost limiting) so you can understand if we don't trust AMD to get it working perfectly.
@@fhpchris Yeah, remember everyone dragging Intel for having to create this Thread Director thing just to make its Hybrid design work when it launched? The main point here is help show people that you are indeed getting 16-cores. This isn't about any design decisions or performance--but there are literally people who do not believe you can access all 16-cores while gaming.
@PCWOrld I like your work, but for many this is NOT the problem. The question many have is this: if I game on my PC, then I run up lets say OBS Studio in the background, for streaming and recording, will that somehow in a major way impact the performance of the game? Because either the scheduler is buggy and put it on the 3D cache part (and use up the 3d cache), or because the mechanic between the CCDs impacts the performance of the game What many people want is the game run on the 3d cache CCD, and every other program put onto the non-3dcache part. This has not been clearly tested. No one has benchmarked this, as far I have seen. Also its not strange at all there is confusion here, because initial reviews were talking about non3d CCD getting parked.
It "spills over" if the load is too much. Not good for gaming / OOBE. You can manually do it yourself with process lasso but most people aren't going to do that. AMD didn't add a scheduler and relies on windows Game Bar... Not only that but many people consider Game Bar bloat and disable it. After some time all these overlays are going to cause issues (Gamebar, NVIDIA Share, MSI Afterburner, Discord, Steam / Epic / Origin / Battlenet).
One example I can show, mincraft dungeons with 12700k and 4090 was giving out 170-300 fps and with 7950x3d, all updated bios and drivers, stock had same results, but when I only set priority to cache CCD 0, not disabling second CCD, just priority, fps went to 650-900fps dinamic. To make it clear, other 8 core still.being used, just way less in games, and more use from cache chip
I think something else is up. The 7950x is faster but even with core disabled it isn’t 2-3.5x faster than the 12700k, it’s at most 15-20% and that’s if the game LOVES cache.
Yeah, I saw Wendell's video and thought oh dang, he beat us already and you can't out nerd Wendell, but we decided there was so much confusion over this, it's worth repeating because once the wrong info train leaves the station, it doesn't come back.
Would love a followup to this with some benchmarks around forcing apps to use the 2nd CCD via something like Process Lasso and measuring the performance difference. Thanks as always for the great content, Gordon.
You know, multi-tasking performance is very difficult to run and replicate and the variances can be all over the place from Tuesday to next month's Monday. I think we could all construct scenarios that highlight weaknesses of each design in the end. My thought is people probably shouldn't over think it. Because yes, the next step is A+B+C on X/X3D/K. I'd say the safest without any numbers right now is X is probably the leader for an all-core performance build where you say, getting it done slightly faster with a cheaper chip is job one. X3D is probably for someone more biased toward gaming performance rather than getting the absolute best all-core performance (and willing to pay a premium for it too over the X). K is probably best for many content creations apps (Adobe and things that hug QuickSync) and games that don't need the cache obviously. All of these parts are great for high-core clock as well as high-core count applications obviously. KS is best for your friend who has something better than you obviously. "Oh, you have a K? I have a KS."
@@FakeGordonMahUng Cheers for the reply. I completely agree with your assessment. Everyone is looking for the one size fits all but clearly one needs to look at the workload and get the appropriate chip to handle that. I personally am looking to eek out as many frames as I can on single threaded VR games like DCS and MSFS while still being able to use the entire die for work-relayed heavy multi threaded rendering and batch image processing when I'm not gaming.
Some people have already done this. It actually runs worse since the cache is only on one ccd. Its basically a high strung chip that cant handles all the cores trying to run that fast. Contrary to this video it is a rip off. Yes you get the same productivity with all 16 cores that arent reliant on the V cache. But in gaming it can only efficiently run as half a chip. Most people would be fine with normal productivity and high end gaming with the 7800x3d which conveniently is postponed by a month so they can first rip off the suckers for a while with getting only half a chip while paying for a full one. Cause they know the smart people wait anyway. This is straight up sketchy shit.
I think people need to realize 7800XD will perform similarly to the stacked die here but then fall off a cliff on multi-threaded performance and also will have lower clock boosts for lightly threaded tasks too. We don't know what the delay is, but I think people are building up a narrative that won't match reality because you will still need: new AM5 board, DDR5 RAM.
@@FakeGordonMahUng AMD sais "best gaming cpu ever", but games don't utilize 16 cores. Also nornal gamers don't want to pay extra for 8 more cores, they won't even use.
@@MegaLol2xd I think I need to go look at the marketing claims but you can shave this many ways. The 7950X3D provides V-Cache performance which you can clearly see can outperform X/K/KS in some titles. And when you need to do a 3D render for your channel or do AV1 CPU-encode using all cores, it will outperform CPUs with a lower core count too. However, your north star should be what works for you. And you probably should not try to dictate what someone else's north star is either right?
@@FakeGordonMahUng I might be wrong but it's pretty clear to me that the 7800X3D is meant for people who don't care much about tasks other than gaming and some browsing and media consumption and office work on their PCs. Which is actually the majority. People like me who need both multi core performance and gaming performance are much rarer than that. AMD tries to use that to their advantage for impatient people who don't want to wait for the 7800X3D to be released, giving in and buy the much more expensive part that wouldn't matter much to them. That's not really a scam though. That's just market observation and marketing execution. Nothing evil. Anti consumer? Yes. But not a scam. A scam would be illegal.
I think they are more concerned with “oh what if I want to game and then in the background do video processing or some work load productivity”. So as Gordon as shown, if that productivity is using 32 threads, all cores will be assigned, probably impacting the gaming performance. If productivity in the background is like 8 thread or something, it’ll just share the vcache 8 core with the game so it’ll still impact gaming performance. As I said before, I’ll just wait for 7800x3D. Much cheaper and I’ll be happy with 8 cores
Not true. This is why people destroyed the i9-11900K even losing against a 5900X. Troy Total War saga scales pretty well with cores and threads. i7-12700K is 61% faster for 1% lows (76fps) than a 5800X3D (39fps).
@@slimjimjimslim5923 Yup. That's why I'm a bit confused by all this. You get 16-cores so you can do more things faster of one thing, more things faster of multiple things than lower core count CPUs. If you are primarily a gamer and your experience is 1 game only, not heavy CPU encode or background tasks 7800X3D, 5800X3D, Core i5, 7600/7700X are more than enough. I do wonder what people expect of 7800X3D when capturing host-based encodes at 4K with many layers versus a 7950X though. Which do they think will be faster?
7950x3d would seemed to be a great cpu for streamers, using processlasso and assign obs to the 2nd ccd while keeping the vcache ccd free for gaming. Also not using your gpu nvenc encoder to increase gaming performance
@@pcworld luckily u set it up once on proces lasso and dont think about it ever again. Not like some programs or apps need more compute power than 1 whole ccd
Definetly i don't see the point for the 3D cache Ryzen 9s, Core i9s have WAY SUPERIOR multitasking because of thread director. Definetly minimum effort by AMD for this launch, obviously 7800X3D will outsell the 7900X3D and 7950X3D.
Totally agree! I can't decide now if not having a decent scheduler is better so that you could assign things yourself or if AMD should have released it with a real scheduler LOL.
@@pcworld Learn scripting and embrace everything a computer has to offer. It takes like 2 lines of powershell code to identify a process and set its affinity.
Doesn’t Cinebench runs faster on non stacked CCD with higher clocks?, I like running purely on vCache for games and other software like chrome on the other.
There are two types of persons, Who talk because has something to say, and who says something because has to talk. Thats the problem in internet. Thanks for be in the first type of guys PCWorld.
so people disabling the second CCD in Bios ( the non 3D stacked ) and are getting high fps in games seen some comparisons, is it true and why would it be wrong to disable that second CCD as you said when playing ?
What i'd like to see is the division 2 running on the 7950 and especially the 7900X3D, before i set up my server, i actually tested the 3950X as a gaming CPU, and i was shocked to see The Division 2 using ALL of the cores and threads with the average CPU usage reaching as high as 90%, or a freaking 32 thread processor. If the division 2 only sits on those 8 cores then yes, the scheduler is locking the game to 8 cores, but if it allows that overflow then i'd say it is working as you'd expect a 2 node CPU to work(the 3950X basically acted the same as the dual processor X5690 it was replacing) Ideally, you'd have a scheduler that monitors the cache misses, and how often one thread is dependant on another thread. Say If thread 22 doesnt care about cache and really wants frequecy, as long as it doesnt always talk back and foth between other threads to the point where moving it to the other CCD would be detrimental, then you could move it to the non3D CCD, it gets higherr frequency, and the threads that like cache get more of it to go around. I think this one of the things AMD is going to use Zylinx and AI processing for, maybe it will not come until Zen6, but i expect there will be an AI hardware thread scheduler that will constantly test and prod thread location for best performance
To be fair, Division 2 latches the cores but doesn't make efficient use of all of them. That said, most PC Crash Race Div 2 players are intimately familiar with Process Lasso, as we try and remove contention from EAC (Easy Anti Cheat) from the game for perhaps an ounce more of crash-prevention. So having to now manage the game on the 3D cached CCD would just be an extra rule to enable. IMO
It kinda depends. Wendell feels the Windows scheduler would be smart enough to send it to the non-stacked die, but I get the feeling this mechanism is pretty blunt. Everything looks like it is biased to the stacked die when a game is focused. OBS using NVENC/Radeon Encode is probably light enough to run on the stacked part with most games. If you do host-based encode, however, at 4K and game, it looks like it's enough work that it 'spills' over to the second die.
Determining the optimum scheduling for a task really requires an understanding of how that task is coded ( how many threads are started? What do the threads do? How is data shared between threads? ). Currently there is no mechanism for developers to provide that information to an OS, so for modern asymmetric architectures an OS either observes the task behaviour and makes a best guess, or provides tools to the PC user to do it themselves. Most PC users are not likely to want to micromanage all tasks, so OS like Windows and Linux will have to improve their automatic task management which is sometimes not optimal. Ideally on the 7950X3D, a game should have exclusive access to whichever CCX is more beneficial for it to run on, with everything else running on the second CCX. If a game benefits from more than one CCX, then it should use both, but I haven't seen this behaviour in any 7950X3D videos yet. For most people using AMD CPUs, an SKU with a single CCX may be a more consistent option ( unless you REALLY benefit from the extra cores from a second CCX ).
Ehhh... the problem for many, including me is exactly the last part of the video. If I'm already having to deal with this, I'd like game process to run on 3D V-Cache CCD and rest of the stuff including browser, streaming, downloading and other 435 BS services in the background on the non 3D V-Cache CCD. And the thing is just not doing it in a perfect sense - which is what I'd expect off 700 USD CPU! And what's worse, if hypothetically I was chilling in game menu that barely uses CPU and running off-game stuff second monitor. Would this mean my 3D V-Cache CCD got populated with that off-game stuff and when I decide to do some 20 mins game session while CPU intense work being done, it will be forced to go to CCD2? That's basically what ticks people off including me. I'd wish they cut the crap and released 7990X3D with both CCDs getting equally same dedicated 3D V-Cache modules. Sure it'd mean that workstation tasks would probably run a smidge slower, but I wouldn't have to deal with anomalies like above, whether it's psychologically or actually. Make it cost 900 bucks, I'm game. And yes, I'm with you that these are mostly edge cases, but it is a pinnacle consumer product at top premium price. It should be a "no compromises" product.
Not optimal behavior. AMD needs to copy how Intel thread director works. If the app is not focus all the threads go to the e-cores, while the focus app get the p-cores. It will take a while to amd to fix that without thread director hardware.
Christ, watch some more videos so you can understand how this works. Putting the in-focus app on vcache cores will only result in better performance if that app is a game. If it's not, it'll likely to better on the non-vcache. So it SHOULDN'T be done like Intel does it.
@@Doughillman Not exactly the same. Intel has it way to do it with hardware putting services on the background on the e-Corea but it use it with thread director and windows 11. Amd will only use the other cores if the work load exist in mean while everything will be on the b cache cores. Amd needs to have something like thread director in which knows which cores to use base of the app. Currently they don’t have it. That is why it will take a while. At bare minimum they should have an option which you can indicate which cores to use on their software for the games you are looking. I think the 7950x3D is dead on arrival. The best option will be the 7800x3D on this generation. Also the review mark games with 1080p with níveos 4090
why cb 4 threads is running on the 3d-cache ccd, that makes no sense... its an application it should run on the 2nd ccd, did u test it with the newest chipset driver ?
That was with the driver provided for review. It appears to be a pretty 'course' mechanism that when in game mode, tries to usher all threads onto the stacked die and when not in game mode, does not park, but the driver favors the non-stacked die. But I think a deeper look would need to happen.
The 7950x3d made a lot of sense for me. I wanted the 16 cores for workloads. But I also wanted the gaming performance of the 3d v-cache. This CPU is meant for people like me. Not a scam at all. If you just want to game this CPU makes absolutely no sense over the 7800x3d. People just like to complain about things they don't understand.
Process lasso can solve it, if I want a regular app to run on the non vcache ones when im running a game at the same time just use a CPU set and put the regular app on the non vcache ones.
I would love some testing between the 7950X and the 7950X3D. I think where the confusion has come in and this still needs answers, let’s say for example you game and stream at the same time, and it only loads up one ccd while the other is parked because the scheduler does not see it beneficial. Does that mean it takes a performance hit, as compared to the 7950X which would not exhibit this behaviour. Would it actually make any real world difference??
The "parking" of the non-stack dies are only for a moment. As the game loads, so it picks the right cores. Moments later, the rest of the cores are back to normal. Then other programs, like streaming, can use them.
@@KB-1976 need to look into the scheduler method of how it works. By parking cores when a game starts up it can lock it into the "correct" cores. Other programmes can and do use the other cores. The game is still on the correct ones.
Its beena while, but if I remember correctly, most if not all game engines cannot use all 32 cores, i dont think they could fully use 16 cores at once. the threads will hop around from core to core, so might as well have them use the vcache cores. thats why usually 8 cores and higher clock speeds do well in gaming and why people are waiting on the 7800x
How will this Amd 7950x3d cpu work as hypervisor like Xcp-ng. I expect that I will not be able to take advantage of v cache. But will all 16 cores and all 32 threads be available in the hypervisor. Do you have any recommended amd cpu to use in a hypervisor server no gaming
Do Intel P and E-cores behave the same way? Like for example if I were gaming and streaming at the same time, will the streaming try to put the load on the P-cores instead of the E-cores?
I wonder if AMD could use Ryzen Master software to instruct their CPUs to direct executables to certain core complex, parking cores to maximize PBO on the remaining ones etc, just like they optimize their GPU drivers.
No, that's up to the CPU scheduler in Windows. On Linux often Intel and AMD modify the scheduler. But on Windows - the source isn't available so they have to collaborate.
@@_nom_ I never had to worry about that many cores or heterogeneous multi core processors but I watched recently a video about this subject at Level1Tech and Wendell talked about core parking, the benefits of leaving energy plan in balanced, how you could micromanage core affinity using an utility called process lasso or SetThreadAffinityMask function, so I don't see any reason for AMD to develop a similar tool (process lasso) or game developers to use that function if they test their game will run smoothly bypassing Windows Scheduler until it get better at handling these newer kind of CPUs.
Thanks ..this enlightened me ..the question that users should ask is :why there is only 32gb ram in the laptops market for a processor that can address more than tera byte of ram?
Anyone getting mad about core parking clearly doesn't ever use a laptop for anything. I have a i7 12800H for work, so that's 6P cores and 8E cores, and at any given moment doing normal office things, most of those cores are parked. That's just how they work.
@@FakeGordonMahUng use the "the best gaming cpu" marketing really didn't help since I feel many are like I don't need that extra die if I'm just gaming.
What inel is doing is considerably more complex and can also cause more issues since the cores they have are actually different where as here it's just the cache and speed but otherwise no difference.
i'll be happy if/when in the future the behavior is improved to isolate Processes better. that's what the Chips need to be doing. so that People that like, idk, run Visual Studio, run Databases, run Graphics Software, or whatever - but also might launch a game at the same time, can get about the best peformance the CPU can offer without doing a lot of hand work to force it that way. i'd be surprised if they didn't get to that soon, but until they do, multi-tasking WILL be a little rocky.
Exactly. I run Visual Studio, VMware and some other stuff in the background and game at the same time regularly. Their intention with 7950x3d was combining gaming and workstation/productivity together which is great in my case, but at the current state it looks like a lose-lose situation when it comes to multi-tasking. If the regular ccd is parked, I won't have the performance that this cpu offers for my non-gaming tasks. And if both ccds are active, both the game and non-gaming threads may be scheduled on the wrong ccd which once again results in less than desirable performance. So it only kind of achieves what it's supposed to do, yet.
@@Life-cb5dz This is why I bought the 7950x. I need to run a bunch of VMs but I also like games at 4k. I made a decision to get 7950x because I figured the my GPU will be the bottleneck anyway. From some benchmarks, it looks like I was wrong. Anyway, I have enough issues with running VMs on this thing. I don’t want to add split die features.
@@smokeyninja9920 it’s like first gen intel PE design back in alder lake era. Intel and Microsoft do optimized the thread director in raptor lake gen so that’s why intel showed off last year exactly how while gaming and productivity, P used gaming and E used productivity.
@@smokeyninja9920 P and E core scheduling is also done by hardware. No driver patches can fix the 7950X3D's multitasking for a no compromise gaming experience
There's really no user intervention required--and in fact--some users want more intervention. However, is it as elegant a design as Intel's chips with Thread Director? Arguably no.
How about virtualization? VirtualBox, VMware Player and HyperVisor thing. I'm using VMWare player and I would like to know is it good to go with 7950x3d and updating to 192GB memory later but starting with cheap 32GB for install phase. Personally I like the idea of being my personal Thread Director to some extent i.e run Mama and like 5 virtual machines. This processor could take me to next level with those because with my 2600K I can run no more that one VM (Win11 for emails). I got carried away and tried 2 vms and then got knocked out very quickly. That Z77 motherboard is going bad and I'm on market for update. I would like the idea of working on web server code on a Win vm and run Linux web server as vm all the times just to be able to move it as readily compatible to some hosting provider's machine.
@@Savitarax Cyberpunk 2077, Battlefield V (i think), Battlefield 2042, Portal 2 (i get way higher fps there than on CSGO and i have an i7-12700K), PS3 emulator, Troy a Total War Saga, the list goes on and on... Yes, that's why the i9-11900K is considered as a garbage flagship and as a high end gaming chip.
Thanks so much for this prompt response Gordon!! I've been struggling trying to explain to people that this is wrong assumption as well. If I could count how many times I've been told last few days that 7950x3d is a scam just wait for 7800x3d, and I knew this was intended way the processor is actually running.
Such scheduling is not much different from the Intel Alder/Raptor Lake big small cores. This is not a scam. I personally don't like either, and would go for a 7800X3D for gaming, or 7950X for general everything (its a great gaming CPU anyway - just not the best).
What you are showing here, proves exactly why this is a scam. When IN GAME the Operating System pushed all new processes to that CCD which eliminates the benefits of having a High Core CPU. The fact that Cinebench forcibly used all 16 threads is not a real world scenario. When you opened new processes, these were forced into the CCD with cache because a game was already running. This does not allow the system to spread the different loads to the other CCD cores so any CPU Load Balancing advantage from having a lot of cores is lost !!!!!
People don't understand it = It's a scam I'm not saying it's a scam, I'm saying when people are ignorant they are quick to judge and frequently make wrong conclusions.
I have literally had people saying "it's a scam" and that they are not getting 16-cores. Look, I don't want to call anyone stupid because I myself am stupid as all humans are in every area they don't understand. But it is not a scam or fraud. Is it disappointing to some? Too expensive for some? That's up to their view of it, but letting the accusations of "scam" spread without trying to rein them in would be wrong..
@@FakeGordonMahUng Stupid was a strong word, and I shouldn't have used it. I should have said, when people are ignorant they are quick to judge and frequently make wrong conclusions. In fact I'll change that...
Yeah. But don't tell that to a certain group of people, they just live in their own little reality bubble. It's also why AMD releases the 7800X3D much later than the CPUs which are basically not really meant for gaming only.
SOME gamers. There's no way that AMD is going to let the 7800 outperform the 7950, so it's going to clock lower. It'll almost certainly be the better cost vs frames value, but it's not going to be the outright frames leader. Some people are willing to spend the extra money to have that small performance lead.
So what if i assign 30 threads to a threading optimized game that allows threadcount set in the game settings, how will this affect the game compared to 16 or running the 13900k/ks?
Well, we'd need to see this game and there might indeed be conditions that favor Intel's design over AMD's. But I think you need to realize that very few games can use all the resources of a 16-core or 24-core CPU. It is exceptionally rare in fact for the bulk of today's gaming experience.
Because it is easier to determine which task would benefit from the P-Cores and which aren't. Also Intel put in actual effort in R&D and worked with Microsoft years ahead to make it work properly, stuff which AMD doesn't really care about. They're just rapid prototyping CPUs and selling them for higher and higher margins. Unfortunately there will be no competition from Intel in the coming years. So AMD will raise margins even further and become more and more anti consumer.
@@LawrenceTimme I also get that feeling that AMD's products are all very rushed. But hey, they have success with that and there are many many people who are happy with their products. I had nothing but issues with the TRX40 Threadripper platform and I wanted an upgrade to the snappiness of my system and some multicore speed bump. So I decided to get a 13900KS which is factory overclocked. I undervolted it and I am very happy. No weird USB issues, no weird freezes, no performance degradation with every Windows Update… Yes the X3D parts of AMD are stronger in some games, but does it matter much to me? No, since my monitor is 144 Hz and I don't play competitive FPS at all. Who cares if I got 10 FPS more in some games, double in some outliers? For my NAS and my HTPC I am using Ryzen 5000 and I am happy with it on Linux.
@@kyoudaiken "lso Intel put in actual effort in R&D and worked with Microsoft years ahead to make it work properly" "When not in focus push to ecores" There is literally 1 line of code from Microsoft and this results in render tasks leaving P cores when you open Explorer or Chrome, lol.
Hypothetical question: if a game manages to max out the utilization of CCD1 (the 8 cores with 3D Vcache), will the other parked cores from CCD2 wake up and help out? Or is the game strictly limited to CCD1? I don’t know of any game that can max out 8 cores, but maybe in the not so distant future it’ll happen.
@@MasterKoala777 Though, the other application (say streaming) will try to get its fair share from the 3D Vcache as well, so you will get an performance hit, even if the other non 3D Vcache 8 cores come into play. You want the 3D Vcache cores to not share its cores with other applications.
The people who say the 7950X3D and 7900X3D are scams are 1)have no clue what they are talking about or 2)they're clickbaiting. The way these CPUs are designed is genius as 8 cores is plenty for games especially since not all games engines are not fully optimised to run across multiple cores so there is no way games will be using 16 cores any time soon. Plus games benefit more from cache too, so I hope we do get the ability to set which die applications run on as I'd be setting every single game to run on the stacked die.
Gordon, you should put a game load on that demands all the cores and demonstrate that all the cores will be utilized if there is a need. If a game like that even exists. Alternatively you could define something which demands all those threads as a game which will force the primary core selection and run that load. Saying the CPU is a fraud is like saying the RAM manufacturers are committing fraud because you have 64s Gig of RAM and your game never uses more than 32 Gigs.
"As expected, Twitter comments or Twitter is full of hollow heads anyway." It would be smarter, if people paid attention to the catastrophically high gaming CPU/system consumption on Intel i9 processors vs Ryzen 9 7950X3D.
Multiple core cpus have to be tested in virtualization capabilities also. I can't find a good benchmark for the intel P+E cores solution, same i don't see a benchmark on v-cache cores on amd cpus :( ...
From what it looks like, it's a pretty course or blunt instrument. Gaming on and in foreground? Bias everything toward the stacked die. Running a render in the background? All cores in use plus gaming in the foreground and going to the stacked die.
The focused game will looks to always be pushed onto the stacked die. If you are doing something in the background where the second die is hot, it looks game will run on the stacked die, then the rest will 'overflow' it and onto the second die. Will the game run slower? Yes, but you are doing more things at the same time too.
Could have lower performance though. The benefits of this configuration is that the chip can use its CCDs depending on workload. V-Cache CCD have lower clock, so if the task not cache sensitive it actually a disadvantage. This way the CPU can use normal CCD for higher clock task.
@@pcworld I agree with what you just wrote I just don't agree with the video and how it talks about half the threads not being used is the reason in terms of being utilized, again love ya especially when I see the collabs
@@FakeGordonMahUng I think people are upset because they don't think they are getting an 8 core gaming + 8 core MT when both are used at the same time.
@@FakeGordonMahUng I think the arguments right now are in such a spectrum that nobody is happy for a multitude of reasons, I think we can all agree the 5800X3d being released now would have vaporized this current dissent cycle.
@@marktackman2886 I'm very confused by all this because it looks like it's a pretty simple choice between them: Best gaming value, non-content creation focused, high-core count focused CPU goes to the 7800X3D. If you do indeed need a higher core count while valuing the goodness of the V-Cache do the 7950X3D. And yes, Intel is still very much in the conversation here because 8-core 7800X3D vs. its equivalent priced Intel chip isn't a walk in the park either in some workloads and games. Why the 7800X3D is delayed by a month I don't know. We'll ask though.
you're only getting 8 cores that have the V cache not the 16. But games don't use more than 8 cores anyway either way i'ts kinda misleading but still works fine.
6:43 While I never said a "scam" (maybe I did and maybe it is) but I did say Streamers prefer one (1) PC that can both run the **GAME AND STREAMING** apps like OBS and all the other open apps that they require. It would be a very interesting comparison between Intel and AMD. There are a lot of people that look at these high-end CPUs, GPUs and high performance systems for their streaming. It's a big industry.
.. My point is if both the game and OBS are all stuck on eight of the 16 cores. In other words, half of the CPU then that is a absolutely a major problem!!!
@@DJaquithFL There is a comment by PC World here saying it did stick to the first CCD and only spilled over after the load became too much. Ideally that isn't what you want so AMD need to fix that. Ofcourse you can do it yourself manually but most people aren't going to do that.
@@griffin1366 .. This extremely popular Gaming + Streaming scenario needs to be tested ASAP. I beg the differ, I think a lot of the people who are buying these high-end CPUs are indeed Sreamers. In the past a lot of people were purchasing two separate PCs, one to handle the streaming and one handling the gaming because the CPUs simply were struggling too much to do it all-in-one. So there needs to be not just a proof of concept but also benchmarking. At this point I wouldn't trust AMD to perform this task.
@@griffin1366 .. Hopefully he'll run a test the same way streamers play on their PC, the game plus all the capture and chat run enough of one PC. Then compare it to Intel.
funniest detail here is that the game is running better on the 8 cores selected for the task than using all 16. those people just don't understand their own $700 CPU makes me wonder if they're also blowing money on getting the fastest RAM even though that has much less influence on performance compared to the non-X3D AMD CPUs
In the end of the video you have explained what this CPU can’t do and at the same time saying it’s not a scam. That’s literally the reason why everybody are saying it’s a scam. People who are buying 7950X3D want their 3D CCD to run games and games only, while their other CCD is handling background processes, encoding or anything else. Yes, you have paid for a 16 core CPU and you get a 16 core CPU. The problem is that it behaves like 8 core CPU.
Except you are literally getting a 16-core CPU. Believe me, an 8-core Ryzen 7000 CPU will not come even near the performance a 16-core Ryzen 7000 CPU. What you allude to is that in gaming it does behave closer to an 8-core CPU by trying to keep everything on one die--however--if you had the RGB unicorn game that indeed needed more than 8-core / 16-threads, the second die would go to work. Ideal? No, but it is a design decision that AMD took given the cards it had and the state of PC games where the overwhelming amount of games do not need more than 16 threads. And remember: If you bought single-ccd 7800X3D and you are indeed playing a game or running a task that needs 16+ threads, it's out of gas.
RIP Gordon :/
You should make this a series and call it "Gordon vs Internet".
That's a bottomless pit you could fall into...
They were upfront about the split-dies from the very beginning. This is how it was designed to function and they made no allusions to the latter, I'd hardly call that a scam.
Me too. But I dunno where it's coming from, but people keep saying it's a scam.
They weren't upfront about how jank the system was for making sure games only run on the vcache CCD. In this video the non-vcache CCD didn't get used at all until Gordon fully loaded cinebench to 32 threads (that's sub-optimum to say the least). Intel did the engineering to implement a hardware scheduler and worked closely with Microsoft for years to get big little working in windows, AMD is just like "mIcRoSoFt GaMe BaR WhItE lISt AnD cOrE pArKiNg!1!!1".
@@BrianCroweAcolyte 🤣 And Big Little had launch issues as well. Coping or forgetful?
reminds me of hard drives that give you 1TB but than its less than that. There is also advertised ram speeds which don't run out of the box, and you have to setup yourself in the bios.
@@FakeGordonMahUng Outrage Pony™ riders after Ryzen 5800X3D launch: how dare you not give us 3D vcache 5950X3D, AMD!!!
Outrage Pony™ riders after Ryzen 7950X3D launch: how dare you give us what we asked for, AMD!!!
Cinebench is a new first-person shooter or a racing game ? Does it support controller ?
Yeah framechasers has been a big person spreading the misinformation. He’s got like 33% dislike ratio on that video he said it doesn’t work.
He did it for views and engagement
Dude is a joke anyway
Loved Wendel's deep dive but also love this super simple "live demo" of it just working! (got it it's not perfect but it's all of 48 hours old and working pretty darn good for something still so uncommon)
How is it just working? Did I miss the part where he said you can't start/stop threads using the fast CCD while you game on the cache CCD without impacting gaming? He just said you can't do that. Ecores do that.
@@fhpchris like I said it's not perfect, but it shows that the other CCD isn't just "shut off", it prioritizes the CCD with extra cache in games and prioritized the higher frequency CCD with CineBench (which uses all the frequency it can get but doesn't need the extra chache), then when he "ran everything" all the cores were used. Again I admit it's not perfect, we can't one-click-pin tasks to whichever CCD we want but the rough functionality IS there. E and P cores aren't perfect either, I'm pretty sure some comon hypervisors still have issues (or just don't even try) to distinguish between threads and try to treat them all equally. Admitedly these chips would probably have a similar problem so, like I said, not perfect but a really good start for a relatively novel CPU design/chip technology on a new socket/platform that's just a few months old. I guess what I'm saying is that this could have easily been so much worse and in my opinion AMD pulled it off pretty well! In the end I think we the consumers are winning, AMD has some benchmarks they can point to where they "beat" Intel so Intel can't get lazy and price gouge like the Skylake and 14nm+++++ debacle.
@@fhpchris The main misconception that people still have is that 2nd CCD is disabled period and you cannot access it. Like on an original Threadripper with Game Mode where you shut off a CCX/CCD and reboot and Windows literally cannot see it or use it.
@@FakeGordonMahUng The reason that game mode disabled the cores and not parked them is because there is no hardware scheduler or affinity keeping the processes on that CCD. Its very possible as long as the cores are active and not parked or disabled that the threads could be assigned to the wrong CCD. If you let threads share the cache CCD you can hurt your game performance ( very much so with R23 threads!) Ryzen is also limited in memory bandwidth with only one CCD active too. Ecores don't have this problem. Many people expected this to work like a 13900k does. ( ~200 cheaper at microcenter) My 7950x doesn't even post at 6000 mhz without changing every option in my bios ( and its slower than my Skylake X with the same GPU due to C state boost limiting) so you can understand if we don't trust AMD to get it working perfectly.
@@fhpchris Yeah, remember everyone dragging Intel for having to create this Thread Director thing just to make its Hybrid design work when it launched?
The main point here is help show people that you are indeed getting 16-cores. This isn't about any design decisions or performance--but there are literally people who do not believe you can access all 16-cores while gaming.
@PCWOrld I like your work, but for many this is NOT the problem. The question many have is this: if I game on my PC, then I run up lets say OBS Studio in the background, for streaming and recording, will that somehow in a major way impact the performance of the game? Because either the scheduler is buggy and put it on the 3D cache part (and use up the 3d cache), or because the mechanic between the CCDs impacts the performance of the game
What many people want is the game run on the 3d cache CCD, and every other program put onto the non-3dcache part. This has not been clearly tested. No one has benchmarked this, as far I have seen.
Also its not strange at all there is confusion here, because initial reviews were talking about non3d CCD getting parked.
It "spills over" if the load is too much. Not good for gaming / OOBE. You can manually do it yourself with process lasso but most people aren't going to do that. AMD didn't add a scheduler and relies on windows Game Bar... Not only that but many people consider Game Bar bloat and disable it. After some time all these overlays are going to cause issues (Gamebar, NVIDIA Share, MSI Afterburner, Discord, Steam / Epic / Origin / Battlenet).
One example I can show, mincraft dungeons with 12700k and 4090 was giving out 170-300 fps and with 7950x3d, all updated bios and drivers, stock had same results, but when I only set priority to cache CCD 0, not disabling second CCD, just priority, fps went to 650-900fps dinamic. To make it clear, other 8 core still.being used, just way less in games, and more use from cache chip
Other example is modern warfare 2, where 12700k in 2k was pushing 250ish fps , and 7950x3d did 375fps.
@@Bostonski you shouldn’t have seen that big of a jump on fps between those two cpus.
@@Bostonski 1% lows are what matters not maximum frames when looking at an ocean.
I think something else is up. The 7950x is faster but even with core disabled it isn’t 2-3.5x faster than the 12700k, it’s at most 15-20% and that’s if the game LOVES cache.
@@jolness1 Maybe he is on 720p 😆
I feel like you read my exact comment in the Level1Techs video on core parking and made a video on it. Thanks Gordon!
Yeah, I saw Wendell's video and thought oh dang, he beat us already and you can't out nerd Wendell, but we decided there was so much confusion over this, it's worth repeating because once the wrong info train leaves the station, it doesn't come back.
Would love a followup to this with some benchmarks around forcing apps to use the 2nd CCD via something like Process Lasso and measuring the performance difference. Thanks as always for the great content, Gordon.
You know, multi-tasking performance is very difficult to run and replicate and the variances can be all over the place from Tuesday to next month's Monday. I think we could all construct scenarios that highlight weaknesses of each design in the end. My thought is people probably shouldn't over think it. Because yes, the next step is A+B+C on X/X3D/K.
I'd say the safest without any numbers right now is X is probably the leader for an all-core performance build where you say, getting it done slightly faster with a cheaper chip is job one. X3D is probably for someone more biased toward gaming performance rather than getting the absolute best all-core performance (and willing to pay a premium for it too over the X). K is probably best for many content creations apps (Adobe and things that hug QuickSync) and games that don't need the cache obviously. All of these parts are great for high-core clock as well as high-core count applications obviously. KS is best for your friend who has something better than you obviously. "Oh, you have a K? I have a KS."
@@FakeGordonMahUng Cheers for the reply. I completely agree with your assessment. Everyone is looking for the one size fits all but clearly one needs to look at the workload and get the appropriate chip to handle that. I personally am looking to eek out as many frames as I can on single threaded VR games like DCS and MSFS while still being able to use the entire die for work-relayed heavy multi threaded rendering and batch image processing when I'm not gaming.
So... like using a 7700X?
Some people have already done this. It actually runs worse since the cache is only on one ccd. Its basically a high strung chip that cant handles all the cores trying to run that fast. Contrary to this video it is a rip off. Yes you get the same productivity with all 16 cores that arent reliant on the V cache. But in gaming it can only efficiently run as half a chip. Most people would be fine with normal productivity and high end gaming with the 7800x3d which conveniently is postponed by a month so they can first rip off the suckers for a while with getting only half a chip while paying for a full one. Cause they know the smart people wait anyway. This is straight up sketchy shit.
the real scam is delayed 7800x3d release.
I think people need to realize 7800XD will perform similarly to the stacked die here but then fall off a cliff on multi-threaded performance and also will have lower clock boosts for lightly threaded tasks too. We don't know what the delay is, but I think people are building up a narrative that won't match reality because you will still need: new AM5 board, DDR5 RAM.
@@FakeGordonMahUng AMD sais "best gaming cpu ever", but games don't utilize 16 cores. Also nornal gamers don't want to pay extra for 8 more cores, they won't even use.
Being that it already sold out I'd have to imagine there just weren't enough die's.
@@MegaLol2xd I think I need to go look at the marketing claims but you can shave this many ways. The 7950X3D provides V-Cache performance which you can clearly see can outperform X/K/KS in some titles. And when you need to do a 3D render for your channel or do AV1 CPU-encode using all cores, it will outperform CPUs with a lower core count too. However, your north star should be what works for you. And you probably should not try to dictate what someone else's north star is either right?
@@FakeGordonMahUng I might be wrong but it's pretty clear to me that the 7800X3D is meant for people who don't care much about tasks other than gaming and some browsing and media consumption and office work on their PCs. Which is actually the majority. People like me who need both multi core performance and gaming performance are much rarer than that. AMD tries to use that to their advantage for impatient people who don't want to wait for the 7800X3D to be released, giving in and buy the much more expensive part that wouldn't matter much to them. That's not really a scam though. That's just market observation and marketing execution. Nothing evil. Anti consumer? Yes. But not a scam. A scam would be illegal.
Man people are in a rough spot once they understand that virtually all games only use 8 cores or less, majority doesn't use more than 4-6 cores.
that 7800x3d launch is gonna be a monster chip for gaming only.
I think they are more concerned with “oh what if I want to game and then in the background do video processing or some work load productivity”. So as Gordon as shown, if that productivity is using 32 threads, all cores will be assigned, probably impacting the gaming performance. If productivity in the background is like 8 thread or something, it’ll just share the vcache 8 core with the game so it’ll still impact gaming performance. As I said before, I’ll just wait for 7800x3D. Much cheaper and I’ll be happy with 8 cores
most people know shiet, just know how to throw money without understanding basics. :D
Not true. This is why people destroyed the i9-11900K even losing against a 5900X.
Troy Total War saga scales pretty well with cores and threads. i7-12700K is 61% faster for 1% lows (76fps) than a 5800X3D (39fps).
@@slimjimjimslim5923 Yup. That's why I'm a bit confused by all this. You get 16-cores so you can do more things faster of one thing, more things faster of multiple things than lower core count CPUs. If you are primarily a gamer and your experience is 1 game only, not heavy CPU encode or background tasks 7800X3D, 5800X3D, Core i5, 7600/7700X are more than enough.
I do wonder what people expect of 7800X3D when capturing host-based encodes at 4K with many layers versus a 7950X though. Which do they think will be faster?
7950x3d would seemed to be a great cpu for streamers, using processlasso and assign obs to the 2nd ccd while keeping the vcache ccd free for gaming. Also not using your gpu nvenc encoder to increase gaming performance
@@pcworld luckily u set it up once on proces lasso and dont think about it ever again. Not like some programs or apps need more compute power than 1 whole ccd
Definetly i don't see the point for the 3D cache Ryzen 9s, Core i9s have WAY SUPERIOR multitasking because of thread director.
Definetly minimum effort by AMD for this launch, obviously 7800X3D will outsell the 7900X3D and 7950X3D.
@@saricubra2867 well it is kicking intels ass at way lower power usage and heat so hey are doing something right.
Totally agree! I can't decide now if not having a decent scheduler is better so that you could assign things yourself or if AMD should have released it with a real scheduler LOL.
@@pcworld Learn scripting and embrace everything a computer has to offer. It takes like 2 lines of powershell code to identify a process and set its affinity.
you could use software like process lasso to tell processor which program game exactly to use which ccd on.
oh yes if you have to tweak several settings, it is a sca you dont need to do such thing on intels lineup
Doesn’t Cinebench runs faster on non stacked CCD with higher clocks?, I like running purely on vCache for games and other software like chrome on the other.
There are two types of persons, Who talk because has something to say, and who says something because has to talk. Thats the problem in internet. Thanks for be in the first type of guys PCWorld.
so people disabling the second CCD in Bios ( the non 3D stacked ) and are getting high fps in games seen some comparisons, is it true and why would it be wrong to disable that second CCD as you said when playing ?
What i'd like to see is the division 2 running on the 7950 and especially the 7900X3D, before i set up my server, i actually tested the 3950X as a gaming CPU, and i was shocked to see The Division 2 using ALL of the cores and threads with the average CPU usage reaching as high as 90%, or a freaking 32 thread processor.
If the division 2 only sits on those 8 cores then yes, the scheduler is locking the game to 8 cores, but if it allows that overflow then i'd say it is working as you'd expect a 2 node CPU to work(the 3950X basically acted the same as the dual processor X5690 it was replacing)
Ideally, you'd have a scheduler that monitors the cache misses, and how often one thread is dependant on another thread.
Say If thread 22 doesnt care about cache and really wants frequecy, as long as it doesnt always talk back and foth between other threads to the point where moving it to the other CCD would be detrimental, then you could move it to the non3D CCD, it gets higherr frequency, and the threads that like cache get more of it to go around.
I think this one of the things AMD is going to use Zylinx and AI processing for, maybe it will not come until Zen6, but i expect there will be an AI hardware thread scheduler that will constantly test and prod thread location for best performance
To be fair, Division 2 latches the cores but doesn't make efficient use of all of them. That said, most PC Crash Race Div 2 players are intimately familiar with Process Lasso, as we try and remove contention from EAC (Easy Anti Cheat) from the game for perhaps an ounce more of crash-prevention. So having to now manage the game on the 3D cached CCD would just be an extra rule to enable. IMO
Frame Chasers claiming it's a scam as he doesn't understand or he does that for clickbaits...
Preach, Gordon!!!
Thanks so much for the clarification, you are great!
So, if you are gaming and use OBS at the same time, OBS will run on the 3D V Cached chiplet too? I would thing that impacts gaming performance.
It kinda depends. Wendell feels the Windows scheduler would be smart enough to send it to the non-stacked die, but I get the feeling this mechanism is pretty blunt. Everything looks like it is biased to the stacked die when a game is focused. OBS using NVENC/Radeon Encode is probably light enough to run on the stacked part with most games. If you do host-based encode, however, at 4K and game, it looks like it's enough work that it 'spills' over to the second die.
Determining the optimum scheduling for a task really requires an understanding of how that task is coded ( how many threads are started? What do the threads do? How is data shared between threads? ). Currently there is no mechanism for developers to provide that information to an OS, so for modern asymmetric architectures an OS either observes the task behaviour and makes a best guess, or provides tools to the PC user to do it themselves.
Most PC users are not likely to want to micromanage all tasks, so OS like Windows and Linux will have to improve their automatic task management which is sometimes not optimal. Ideally on the 7950X3D, a game should have exclusive access to whichever CCX is more beneficial for it to run on, with everything else running on the second CCX. If a game benefits from more than one CCX, then it should use both, but I haven't seen this behaviour in any 7950X3D videos yet.
For most people using AMD CPUs, an SKU with a single CCX may be a more consistent option ( unless you REALLY benefit from the extra cores from a second CCX ).
RIP Gordon
Ehhh... the problem for many, including me is exactly the last part of the video. If I'm already having to deal with this, I'd like game process to run on 3D V-Cache CCD and rest of the stuff including browser, streaming, downloading and other 435 BS services in the background on the non 3D V-Cache CCD.
And the thing is just not doing it in a perfect sense - which is what I'd expect off 700 USD CPU! And what's worse, if hypothetically I was chilling in game menu that barely uses CPU and running off-game stuff second monitor. Would this mean my 3D V-Cache CCD got populated with that off-game stuff and when I decide to do some 20 mins game session while CPU intense work being done, it will be forced to go to CCD2?
That's basically what ticks people off including me. I'd wish they cut the crap and released 7990X3D with both CCDs getting equally same dedicated 3D V-Cache modules. Sure it'd mean that workstation tasks would probably run a smidge slower, but I wouldn't have to deal with anomalies like above, whether it's psychologically or actually.
Make it cost 900 bucks, I'm game.
And yes, I'm with you that these are mostly edge cases, but it is a pinnacle consumer product at top premium price. It should be a "no compromises" product.
I just spent 8:58 watching Gordon’s soul leaving him
Process Lasso is just great, I don't understand the fuzz.
Not optimal behavior. AMD needs to copy how Intel thread director works. If the app is not focus all the threads go to the e-cores, while the focus app get the p-cores. It will take a while to amd to fix that without thread director hardware.
There is no "thread director hardware"
What you described is done by windows scheduler that uses focus to manage threads.
Christ, watch some more videos so you can understand how this works. Putting the in-focus app on vcache cores will only result in better performance if that app is a game. If it's not, it'll likely to better on the non-vcache. So it SHOULDN'T be done like Intel does it.
@@Doughillman Not exactly the same. Intel has it way to do it with hardware putting services on the background on the e-Corea but it use it with thread director and windows 11. Amd will only use the other cores if the work load exist in mean while everything will be on the b cache cores. Amd needs to have something like thread director in which knows which cores to use base of the app. Currently they don’t have it. That is why it will take a while. At bare minimum they should have an option which you can indicate which cores to use on their software for the games you are looking. I think the 7950x3D is dead on arrival. The best option will be the 7800x3D on this generation. Also the review mark games with 1080p with níveos 4090
why cb 4 threads is running on the 3d-cache ccd, that makes no sense... its an application it should run on the 2nd ccd, did u test it with the newest chipset driver ?
That was with the driver provided for review. It appears to be a pretty 'course' mechanism that when in game mode, tries to usher all threads onto the stacked die and when not in game mode, does not park, but the driver favors the non-stacked die. But I think a deeper look would need to happen.
The 7950x3d made a lot of sense for me. I wanted the 16 cores for workloads. But I also wanted the gaming performance of the 3d v-cache. This CPU is meant for people like me. Not a scam at all. If you just want to game this CPU makes absolutely no sense over the 7800x3d. People just like to complain about things they don't understand.
Process lasso can solve it, if I want a regular app to run on the non vcache ones when im running a game at the same time just use a CPU set and put the regular app on the non vcache ones.
i have a 7700x and very happy with it matched with a 4080.
Love the RoboCop shirt!
I'll buy that for a dollar! And: How about we get you an SUX6000 and throw in a Blaupunkt?
@@FakeGordonMahUng Dick! You're fired!
@@gunnaraw Your move creep.
I would love some testing between the 7950X and the 7950X3D.
I think where the confusion has come in and this still needs answers, let’s say for example you game and stream at the same time, and it only loads up one ccd while the other is parked because the scheduler does not see it beneficial. Does that mean it takes a performance hit, as compared to the 7950X which would not exhibit this behaviour. Would it actually make any real world difference??
The "parking" of the non-stack dies are only for a moment. As the game loads, so it picks the right cores.
Moments later, the rest of the cores are back to normal. Then other programs, like streaming, can use them.
@@hrayz This is not correct if you listen to what Gordon is saying in the video.
@@hrayz naw, it is shutting OFF the other ccd boss. (Gaming).
Edit- If it does use the other ccd at all, then performance will go down
@@KB-1976 need to look into the scheduler method of how it works. By parking cores when a game starts up it can lock it into the "correct" cores. Other programmes can and do use the other cores. The game is still on the correct ones.
Its beena while, but if I remember correctly, most if not all game engines cannot use all 32 cores, i dont think they could fully use 16 cores at once. the threads will hop around from core to core, so might as well have them use the vcache cores. thats why usually 8 cores and higher clock speeds do well in gaming and why people are waiting on the 7800x
AMD Ryzen™ 9 7950X3D
Amazon - Currently unavailable.
Newegg - OUT OF STOCK.
Best Buy - Sold Out
so is my chip broken if its not parking? i've never seen a parked core in resource monitor
so if im playing a game all my background apps are forced onto the same cores as the game? why would i not just buy an 8 core then?
Gordon and Adam in the same room like the old days. Good to see you two hanging out showing dummies how things work.
Don’t say dummies that is mean
I wonder how this CPU scales in more demanding dames in the future.
it wont, only 8 cores while gaming
How will this Amd 7950x3d cpu work as hypervisor like Xcp-ng. I expect that I will not be able to take advantage of v cache. But will all 16 cores and all 32 threads be available in the hypervisor.
Do you have any recommended amd cpu to use in a hypervisor server no gaming
Ok…..you are correct…….but “core parking” is a nightmare that doesn’t function properly………..even after all required settings have been set.
People are, for the most part, idiots. If anyone who could afford the CPU paid attention it was all there!
Do Intel P and E-cores behave the same way? Like for example if I were gaming and streaming at the same time, will the streaming try to put the load on the P-cores instead of the E-cores?
How was your AMD managers meeting?
Oh Gawd, you must be from Frame Chasers..... 🤦♂
because the cinebench was designed for older cpu architecture. it is mot about being scammed.
TY Gordon! Very helpful!🙌
Finally some reason!! cheers Gordon!
I wonder if AMD could use Ryzen Master software to instruct their CPUs to direct executables to certain core complex, parking cores to maximize PBO on the remaining ones etc, just like they optimize their GPU drivers.
No, that's up to the CPU scheduler in Windows. On Linux often Intel and AMD modify the scheduler. But on Windows - the source isn't available so they have to collaborate.
@@_nom_ I never had to worry about that many cores or heterogeneous multi core processors but I watched recently a video about this subject at Level1Tech and Wendell talked about core parking, the benefits of leaving energy plan in balanced, how you could micromanage core affinity using an utility called process lasso or SetThreadAffinityMask function, so I don't see any reason for AMD to develop a similar tool (process lasso) or game developers to use that function if they test their game will run smoothly bypassing Windows Scheduler until it get better at handling these newer kind of CPUs.
The big issue is that it's not better than a 7950X for non-gaming and it's not any better than how a 7800X3D would run for gaming.
Thanks ..this enlightened me ..the question that users should ask is :why there is only 32gb ram in the laptops market for a processor that can address more than tera byte of ram?
Anyone getting mad about core parking clearly doesn't ever use a laptop for anything. I have a i7 12800H for work, so that's 6P cores and 8E cores, and at any given moment doing normal office things, most of those cores are parked. That's just how they work.
It's unfortunate there's a lot of misunderstanding about this and makes it wonder if reviewers should have anticipated it in our original reviews.
@@FakeGordonMahUng use the "the best gaming cpu" marketing really didn't help since I feel many are like I don't need that extra die if I'm just gaming.
What inel is doing is considerably more complex and can also cause more issues since the cores they have are actually different where as here it's just the cache and speed but otherwise no difference.
i'll be happy if/when in the future the behavior is improved to isolate Processes better. that's what the Chips need to be doing. so that People that like, idk, run Visual Studio, run Databases, run Graphics Software, or whatever - but also might launch a game at the same time, can get about the best peformance the CPU can offer without doing a lot of hand work to force it that way.
i'd be surprised if they didn't get to that soon, but until they do, multi-tasking WILL be a little rocky.
That’s up the operating system’s scheduler and the programmer of the app.
@@ryanslab302
"not my fault, but my problem" applies. it's Zen4s' problem if it doesn't work well, doesn't matter whos' fault it is.
Exactly. I run Visual Studio, VMware and some other stuff in the background and game at the same time regularly. Their intention with 7950x3d was combining gaming and workstation/productivity together which is great in my case, but at the current state it looks like a lose-lose situation when it comes to multi-tasking. If the regular ccd is parked, I won't have the performance that this cpu offers for my non-gaming tasks. And if both ccds are active, both the game and non-gaming threads may be scheduled on the wrong ccd which once again results in less than desirable performance. So it only kind of achieves what it's supposed to do, yet.
@@Life-cb5dz This is why I bought the 7950x. I need to run a bunch of VMs but I also like games at 4k. I made a decision to get 7950x because I figured the my GPU will be the bottleneck anyway. From some benchmarks, it looks like I was wrong. Anyway, I have enough issues with running VMs on this thing. I don’t want to add split die features.
So it’s a tinkerer’s chip just like intel’s arc gpu?
It’s much higher end than arc. And I wouldn’t really call setting ccd affinity for different apps tinkering.
More like Intel's P/E core design, it needs optimizations in software to yield its best performance
@@smokeyninja9920 it’s like first gen intel PE design back in alder lake era. Intel and Microsoft do optimized the thread director in raptor lake gen so that’s why intel showed off last year exactly how while gaming and productivity, P used gaming and E used productivity.
@@smokeyninja9920 P and E core scheduling is also done by hardware. No driver patches can fix the 7950X3D's multitasking for a no compromise gaming experience
There's really no user intervention required--and in fact--some users want more intervention. However, is it as elegant a design as Intel's chips with Thread Director? Arguably no.
Windows doesn’t handle X3D CCD affinity well
How about virtualization? VirtualBox, VMware Player and HyperVisor thing. I'm using VMWare player and I would like to know is it good to go with 7950x3d and updating to 192GB memory later but starting with cheap 32GB for install phase. Personally I like the idea of being my personal Thread Director to some extent i.e run Mama and like 5 virtual machines. This processor could take me to next level with those because with my 2600K I can run no more that one VM (Win11 for emails). I got carried away and tried 2 vms and then got knocked out very quickly. That Z77 motherboard is going bad and I'm on market for update. I would like the idea of working on web server code on a Win vm and run Linux web server as vm all the times just to be able to move it as readily compatible to some hosting provider's machine.
Its insane. Every comment on youtube that is calling it a scam are subscribed to Frame Chasers channel. Man he is very manipulative.
He's a master of getting clicks and engagement! His normal videos are pretty good.
so its better to fo prioritize cache in the bios when you are mainly gaming?
Gaming and production performance numbers using them at the time?
You can via task manager processor affinity
Some people just seem determined to misunderstand things.
Would the regular 7950x be a better option for me as I game on triple high resolution monitors + a upper 4th? sim racing
Also eyeing this for an upgrade to my sim racing rig. VR uses a ton of CPU performance and my 8700k doesn't keep up.
Wait what games can even utilize 8 cores or more?
Really not any
Very rarely, I think battlefield is the only one I can think of. Games are trying to scale with more but consoles have 8 cores so that’s the standard.
We probably won't see a game actually needing more than 8 cores for at least another generation.
@@Savitarax Cyberpunk 2077, Battlefield V (i think), Battlefield 2042, Portal 2 (i get way higher fps there than on CSGO and i have an i7-12700K), PS3 emulator, Troy a Total War Saga, the list goes on and on...
Yes, that's why the i9-11900K is considered as a garbage flagship and as a high end gaming chip.
@@saricubra2867 noone is saying not to go intel if u play those specific games. but fact is 79050x3d atm is the top dog in general.
Congrats Gordon, thanks for this video.
Is process affinity dead?
Thanks so much for this prompt response Gordon!! I've been struggling trying to explain to people that this is wrong assumption as well. If I could count how many times I've been told last few days that 7950x3d is a scam just wait for 7800x3d, and I knew this was intended way the processor is actually running.
The class action lawyers have already filed a lawsuit. I hope it goes down in flames
I wish Gordon would test it on Minesweeper to see if 8 or 16 cores were running.
I Tested mincraft dungeons. Insane jump. From 170-300fps it went up to 650-900 just by setting priority Cache CCD.
Thanks for this Gordon!
Such scheduling is not much different from the Intel Alder/Raptor Lake big small cores. This is not a scam. I personally don't like either, and would go for a 7800X3D for gaming, or 7950X for general everything (its a great gaming CPU anyway - just not the best).
What you are showing here, proves exactly why this is a scam. When IN GAME the Operating System pushed all new processes to that CCD which eliminates the benefits of having a High Core CPU. The fact that Cinebench forcibly used all 16 threads is not a real world scenario. When you opened new processes, these were forced into the CCD with cache because a game was already running. This does not allow the system to spread the different loads to the other CCD cores so any CPU Load Balancing advantage from having a lot of cores is lost !!!!!
xD just this one comment destroys the entire video.
Can anyone tell me if there are benifts for VFX or CG artists rendering CPU? Big scenes .,? Lot of geo?
People don't understand it = It's a scam
I'm not saying it's a scam, I'm saying when people are ignorant they are quick to judge and frequently make wrong conclusions.
I have literally had people saying "it's a scam" and that they are not getting 16-cores. Look, I don't want to call anyone stupid because I myself am stupid as all humans are in every area they don't understand. But it is not a scam or fraud. Is it disappointing to some? Too expensive for some? That's up to their view of it, but letting the accusations of "scam" spread without trying to rein them in would be wrong..
@@FakeGordonMahUng Stupid was a strong word, and I shouldn't have used it.
I should have said, when people are ignorant they are quick to judge and frequently make wrong conclusions.
In fact I'll change that...
What happens if you try to run 2 games at the same time?
It’s running tomb raider at 170 fps while doing cinebench on the other cores
The scam is that games don't use more than 8 cores. Thus the 7800X3D at a lower cost will be a better choice for gamers.
Yeah. But don't tell that to a certain group of people, they just live in their own little reality bubble. It's also why AMD releases the 7800X3D much later than the CPUs which are basically not really meant for gaming only.
SOME gamers. There's no way that AMD is going to let the 7800 outperform the 7950, so it's going to clock lower. It'll almost certainly be the better cost vs frames value, but it's not going to be the outright frames leader. Some people are willing to spend the extra money to have that small performance lead.
So what if i assign 30 threads to a threading optimized game that allows threadcount set in the game settings, how will this affect the game compared to 16 or running the 13900k/ks?
Well, we'd need to see this game and there might indeed be conditions that favor Intel's design over AMD's. But I think you need to realize that very few games can use all the resources of a 16-core or 24-core CPU. It is exceptionally rare in fact for the bulk of today's gaming experience.
Once again, great analysis by Gordon!
Scam? No... it's exactly what everyone asked for... it just isn't as good as hoped.
Why isn't it just all automatic like the ewaste cores on bintel cpus?
Because it is easier to determine which task would benefit from the P-Cores and which aren't. Also Intel put in actual effort in R&D and worked with Microsoft years ahead to make it work properly, stuff which AMD doesn't really care about. They're just rapid prototyping CPUs and selling them for higher and higher margins. Unfortunately there will be no competition from Intel in the coming years. So AMD will raise margins even further and become more and more anti consumer.
@@kyoudaiken it feels like a beta product again. I'm sure the 7800x3d will be the real deal as the 5800x3d has been proven to be good now
@@LawrenceTimme I also get that feeling that AMD's products are all very rushed. But hey, they have success with that and there are many many people who are happy with their products. I had nothing but issues with the TRX40 Threadripper platform and I wanted an upgrade to the snappiness of my system and some multicore speed bump. So I decided to get a 13900KS which is factory overclocked. I undervolted it and I am very happy. No weird USB issues, no weird freezes, no performance degradation with every Windows Update… Yes the X3D parts of AMD are stronger in some games, but does it matter much to me? No, since my monitor is 144 Hz and I don't play competitive FPS at all. Who cares if I got 10 FPS more in some games, double in some outliers? For my NAS and my HTPC I am using Ryzen 5000 and I am happy with it on Linux.
@@kyoudaiken "lso Intel put in actual effort in R&D and worked with Microsoft years ahead to make it work properly"
"When not in focus push to ecores"
There is literally 1 line of code from Microsoft and this results in render tasks leaving P cores when you open Explorer or Chrome, lol.
Hypothetical question: if a game manages to max out the utilization of CCD1 (the 8 cores with 3D Vcache), will the other parked cores from CCD2 wake up and help out? Or is the game strictly limited to CCD1?
I don’t know of any game that can max out 8 cores, but maybe in the not so distant future it’ll happen.
@@pcworld Alright, thanks Adam! That should remove the slightest suspicions left out there that they’re being scammed 🙂
@@MasterKoala777 buy intel…or 7800x3d when it comes out
@@MasterKoala777 Though, the other application (say streaming) will try to get its fair share from the 3D Vcache as well, so you will get an performance hit, even if the other non 3D Vcache 8 cores come into play. You want the 3D Vcache cores to not share its cores with other applications.
@@muSPKwow yes, agreed. Tech Yes City did some further tests on this, and the gaming performance suffered in some cases.
The people who say the 7950X3D and 7900X3D are scams are 1)have no clue what they are talking about or 2)they're clickbaiting. The way these CPUs are designed is genius as 8 cores is plenty for games especially since not all games engines are not fully optimised to run across multiple cores so there is no way games will be using 16 cores any time soon. Plus games benefit more from cache too, so I hope we do get the ability to set which die applications run on as I'd be setting every single game to run on the stacked die.
Gordon, you should put a game load on that demands all the cores and demonstrate that all the cores will be utilized if there is a need. If a game like that even exists. Alternatively you could define something which demands all those threads as a game which will force the primary core selection and run that load. Saying the CPU is a fraud is like saying the RAM manufacturers are committing fraud because you have 64s Gig of RAM and your game never uses more than 32 Gigs.
I know someone who won't be Happy with this response.
Framechasers?
"As expected, Twitter comments or Twitter is full of hollow heads anyway." It would be smarter, if people paid attention to the catastrophically high gaming CPU/system consumption on Intel i9 processors vs Ryzen 9 7950X3D.
Multiple core cpus have to be tested in virtualization capabilities also. I can't find a good benchmark for the intel P+E cores solution, same i don't see a benchmark on v-cache cores on amd cpus :( ...
So games can shoved into the slow bit when all cores are being used generally?
From what it looks like, it's a pretty course or blunt instrument. Gaming on and in foreground? Bias everything toward the stacked die. Running a render in the background? All cores in use plus gaming in the foreground and going to the stacked die.
The focused game will looks to always be pushed onto the stacked die. If you are doing something in the background where the second die is hot, it looks game will run on the stacked die, then the rest will 'overflow' it and onto the second die. Will the game run slower? Yes, but you are doing more things at the same time too.
Yes I see. Maybe it's for the better, I should not be working and gaming at the same time anyway.
So what about virtualization on windows? Could use the v cache cores for a vm?
Sure why not?
@Elliot Kaufman is it an easy configuration?
@@jeremiahk364 you would just assign the first 15 threads to a VM those are the v-cache ccd
I would have payed (a lot) more to get a full 2 CCD 7950X-3D.
I don’t believe the process has gotten to a place where they like it in terms of temperature.
Then you definitely would feel like scammed, as AMD didn't see performance increase in doing so... atleast for now
I would have too. It was extremely ambitious to try to do a hybrid layout without a hardware scheduler like 12/13th gen intel cpus have.
Could have lower performance though. The benefits of this configuration is that the chip can use its CCDs depending on workload. V-Cache CCD have lower clock, so if the task not cache sensitive it actually a disadvantage. This way the CPU can use normal CCD for higher clock task.
I love Gordan but this is a straw-man argument and not the reason why people are really upset.
@@pcworld I agree with what you just wrote I just don't agree with the video and how it talks about half the threads not being used is the reason in terms of being utilized, again love ya especially when I see the collabs
I have been seeing and responding to arguments that this is "scam" because "I don't get a 16-core CPU!" for days now.
@@FakeGordonMahUng I think people are upset because they don't think they are getting an 8 core gaming + 8 core MT when both are used at the same time.
@@FakeGordonMahUng I think the arguments right now are in such a spectrum that nobody is happy for a multitude of reasons, I think we can all agree the 5800X3d being released now would have vaporized this current dissent cycle.
@@marktackman2886 I'm very confused by all this because it looks like it's a pretty simple choice between them: Best gaming value, non-content creation focused, high-core count focused CPU goes to the 7800X3D. If you do indeed need a higher core count while valuing the goodness of the V-Cache do the 7950X3D. And yes, Intel is still very much in the conversation here because 8-core 7800X3D vs. its equivalent priced Intel chip isn't a walk in the park either in some workloads and games.
Why the 7800X3D is delayed by a month I don't know. We'll ask though.
It needs to isolated the background stuff to the frequency cores and kept the cache cores for the foreground application.
Depends on the application, really. The opposite assignment can be favorable too.
You can manually do that but that's far more complex, especially if you want to determine that automatically.
Bravo!!! GREAT video!
you're only getting 8 cores that have the V cache not the 16. But games don't use more than 8 cores anyway either way i'ts kinda misleading but still works fine.
Please add arc card to your test, i use for years intel+nvidia card, for 422 codecs. I dont know who dont use it as me
6:43 While I never said a "scam" (maybe I did and maybe it is) but I did say Streamers prefer one (1) PC that can both run the **GAME AND STREAMING** apps like OBS and all the other open apps that they require. It would be a very interesting comparison between Intel and AMD. There are a lot of people that look at these high-end CPUs, GPUs and high performance systems for their streaming. It's a big industry.
.. My point is if both the game and OBS are all stuck on eight of the 16 cores. In other words, half of the CPU then that is a absolutely a major problem!!!
@@DJaquithFL There is a comment by PC World here saying it did stick to the first CCD and only spilled over after the load became too much. Ideally that isn't what you want so AMD need to fix that. Ofcourse you can do it yourself manually but most people aren't going to do that.
@@griffin1366 .. This extremely popular Gaming + Streaming scenario needs to be tested ASAP. I beg the differ, I think a lot of the people who are buying these high-end CPUs are indeed Sreamers.
In the past a lot of people were purchasing two separate PCs, one to handle the streaming and one handling the gaming because the CPUs simply were struggling too much to do it all-in-one. So there needs to be not just a proof of concept but also benchmarking. At this point I wouldn't trust AMD to perform this task.
@@DJaquithFL I'm waiting for FrameChasers review.
@@griffin1366 .. Hopefully he'll run a test the same way streamers play on their PC, the game plus all the capture and chat run enough of one PC. Then compare it to Intel.
funniest detail here is that the game is running better on the 8 cores selected for the task than using all 16.
those people just don't understand their own $700 CPU
makes me wonder if they're also blowing money on getting the fastest RAM even though that has much less influence on performance compared to the non-X3D AMD CPUs
In the end of the video you have explained what this CPU can’t do and at the same time saying it’s not a scam. That’s literally the reason why everybody are saying it’s a scam.
People who are buying 7950X3D want their 3D CCD to run games and games only, while their other CCD is handling background processes, encoding or anything else.
Yes, you have paid for a 16 core CPU and you get a 16 core CPU. The problem is that it behaves like 8 core CPU.
Except you are literally getting a 16-core CPU. Believe me, an 8-core Ryzen 7000 CPU will not come even near the performance a 16-core Ryzen 7000 CPU. What you allude to is that in gaming it does behave closer to an 8-core CPU by trying to keep everything on one die--however--if you had the RGB unicorn game that indeed needed more than 8-core / 16-threads, the second die would go to work. Ideal? No, but it is a design decision that AMD took given the cards it had and the state of PC games where the overwhelming amount of games do not need more than 16 threads. And remember: If you bought single-ccd 7800X3D and you are indeed playing a game or running a task that needs 16+ threads, it's out of gas.
This is still early on for this product for both amd and windows, give it time they will figure it out.
100% scam and now they even lowered the clocks on the non X3D CPU's
Scam after scam