imho, bigger is doubtful but we should see the same delta between 3d/normal
So it may not alleviate the issue of upgrade cost over performance jump then. For gaming its not worth going from a fast 5*** non3d chip to a 7*** series, certainly when MOBO and RAM prices are added.
So we may see the same dilemma for anybody with a 5800x3d or good clocking non 3D chip, certainly when it comes to 4k gaming.
That's a nice spread of CPUs. But no 6 core with 3DV cache. That is a shame, as it would probably be the bang for buck CPU for gamers.
My concern is the TDP. The 5800X3D only has a TFP of 105W, but still, when pushed hard it goes to 95ºC and starts to throttle.
So these new CPUs with 170W of TDP might throttle even harder.
A PBO offset does wonders for the 5800X3D. But for these 7000X3D, it will be essential.
Hmm, I'm a bit concerned about that after reading this thread, 7800X3D would be my next planned CPU which I hope to use with an NH-D14 Noctua air cooler (I'm assuming Noctua are still releasing "free attachment kits" for new sockets & CPUs?). My NH-D14 can keep my 1.4V 6700K below 65 degC in the summer when outputting 120-140W (it's delidded with liquid metal). So I kind of trusted that it would be able to handle a stock 7800X3D.....at least I expected it would keep the 5800X3D cool enough when I read about it re it's quite low TDP.....I just hope it's enough for a 7800X3D (I have good case air flow on top of that). I'm just hoping to invest in new motherboard & RAM (& of course that CPU) and leave everything else as is!
I would absolutely love to have a 3D cache version of a 16 core variant to upgrade from my 5950X. Would be the only way to make me take the expense of replacing the CPU as well as an expensive motherboard and having to replace 32GB of high-end Samsung DDR4 with some DDR5 version of it.
Also, it would have to beat Intels best 13900K in single threaded performance - right now that CPU is beating the crap out of AMD - my 4080 is completely bottle-necked by CPU in a certain 3D game benchmark (only running 44% or max 70% due to single threaded physics calculations in Unity engine) compared to the Intel 13900K
I would absolutely love to have a 3D cache version of a 16 core variant to upgrade from my 5950X. Would be the only way to make me take the expense of replacing the CPU as well as an expensive motherboard and having to replace 32GB of high-end Samsung DDR4 with some DDR5 version of it.
Also, it would have to beat Intels best 13900K in single threaded performance - right now that CPU is beating the crap out of AMD - my 4080 is completely bottle-necked by CPU in a certain 3D game benchmark (only running 44% or max 70% due to single threaded physics calculations in Unity engine) compared to the Intel 13900K
its still possible that they may launch something like that,(5950x3d) but it will almost certainly be after the 7000x3d chips launch, because if they were to launch them now, it would cannibalize their own product stack
I would absolutely love to have a 3D cache version of a 16 core variant to upgrade from my 5950X. Would be the only way to make me take the expense of replacing the CPU as well as an expensive motherboard and having to replace 32GB of high-end Samsung DDR4 with some DDR5 version of it.
Also, it would have to beat Intels best 13900K in single threaded performance - right now that CPU is beating the crap out of AMD - my 4080 is completely bottle-necked by CPU in a certain 3D game benchmark (only running 44% or max 70% due to single threaded physics calculations in Unity engine) compared to the Intel 13900K
What games and settings is that?
I know that a 6900xt is a good deal slower then a 4080, but I am GPU limited or monitor limited (1440p 165hz) 90% of the time with a 5800x at 4900Mhz single core boost.
Is it the Nvidia driver CPU overhead that is using more CPU power then my 6900xt?
The IHS for the 7000 series is too thick, so they should just make it thinner in the 3D models, to make room for the 3D cache on top of the chiplets. I'm sure there are technical considerations, but the IHS is still too thick, as made evident by the insane gains from delidding.
comparison between a stock 7950x vs stock 13900k cinebench R23 with the exact same cooling setup
2x 360x54mm + 1x 360x30mm corsair HydroX custom loop + EKWB block(s) with weak (but quiet) QL120 fans
what you see in the single core run at the begininning followed by 3 multi-core cinebenchR23
on the left it's the 7950x and you kind of see it but the temp goes down noticeably slower from <1s to 2s on the intel to 3,4,5s on the AMD 7950x
I had the same setup with a 3960x,9900k,5950x,12900k and it's the 1st time I see that it's 100% coming from the IHS
note that the single core run on amd is hotter probably because amd clocks are higher than intel on that gen
FULL STOCK NO AIB OC on both plateform
amd=5781/5505 all cores of a ccd have the same clocks
intel=5487/4290 with 2 cores only reaching 5786Mhz
I wish I wouldn't be able to do that comparison and that the 7950x worked fine but it didn't so I built a 13900k 2 weeks later
do not buy the MSI X670E meg ace most of the sold motherboards are defective and/or the bios has unacceptable issues..it freezes and crashes...the bios screen !
lots of complaints about memory with it on msi own forums and those people seem "casuals" who haven't tried all I did and seen that the problems are deeper than being forced to run your ddr5 at 4800(2)3600(4) my benchmarks are same or better than the 13900k but gaming is simply impossible rivatuner shows me a straight horizontal 120fps line...it looks like 12fps insane stuttering 😡 in fact it's so bad it made me physically sick kind of like VR sickness,m if that wasn't enough PBO doesn't work at all on my motherboard it does nothing or destroys my performance while increasing watts (32k cbr23@270w with pbo while activating eco mode gives me 32k@80w >< eco works at least )
the 13900k full stock goes to 350w and with msi "boost" (where all p-cores go 5886mhz) 383w so....air coolers and I would say even aio users need not apply
it's not as bad as when I tried to OC the 3960x (430+watts) but getting close and btw it's that threadripper who forced me to go full custom loop
its still possible that they may launch something like that,(5950x3d) but it will almost certainly be after the 7000x3d chips launch, because if they were to launch them now, it would cannibalize their own product stack
Sorry you misunderstood me, I wasn't expecting a 5950X3D - I meant if they launch a 7950X3D I might upgrade.
What games and settings is that?
I know that a 6900xt is a good deal slower then a 4080, but I a m GPU limited or monitor limited (1440p 165hz) 90% of the time with a 5800x at 4900Mhz single core boost.
Is it the Nvidia driver CPU overhead that is using more CPU power then my 6900xt?
In Virta-mate. Benchmark comparisons here: https://hub.virtamate.com/threads/benchmark-result-discussion.13131/
The one to look at is Baseline 3 result scores - this is a heavy single core physics bound test... see how much faster the 13900K is than the 5950X here. Physics time - 10.33ms vs 3.6ms resulting in 56fps vs 104fps. In this Baseline 3 test the 4080 does not go over 44% usage. In "Simpler Physics" test it does not go over 70%.
AMD will have to step up and start competing with Intel in this sort of heavy graphics combined with physics tests. Not sure if there is a 7950X comparison there.
EDIT: I did improve (double) Baseline 3 scores from above by using clean install of it.
In Virta-mate. Benchmark comparisons here: https://hub.virtamate.com/threads/benchmark-result-discussion.13131/
The one to look at is Baseline 3 result scores - this is a heavy single core physics bound test... see how much faster the 13900K is than the 5950X here. Physics time - 10.33ms vs 3.6ms resulting in 56fps vs 104fps. In this Baseline 3 test the 4080 does not go over 44% usage. In "Simpler Physics" test it does not go over 70%.
AMD will have to step up and start competing with Intel in this sort of heavy graphics combined with physics tests. Not sure if there is a 7950X comparison there.
I have no idea what this benchmark is and I do not feel like downloading it from a home page, with that high amount of nude Anime boobs and butts.
Few rumours circling 7950X3D is up to 30% faster than regular 7950x. Will make future 8950X potentially hard sale? if it's not good bit better than 7950X3D.
comparison between a stock 7950x vs stock 13900k cinebench R23 with the exact same cooling setup
2x 360x54mm + 1x 360x30mm corsair HydroX custom loop + EKWB block(s) with weak (but quiet) QL120 fans
(...)
I wish I wouldn't be able to do that comparison and that the 7950x worked fine but it didn't so I built a 13900k 2 weeks later
do not buy the MSI X670E meg ace most of the sold motherboards are defective and/or the bios has unacceptable issues..it freezes and crashes...the bios screen !
lots of complaints about memory with it on msi own forums and those people seem "casuals" who haven't tried all I did and seen that the problems are deeper than being forced to run your ddr5 at 4800(2)3600(4) my benchmarks are same or better than the 13900k but gaming is simply impossible rivatuner shows me a straight horizontal 120fps line...it looks like 12fps insane stuttering 😡 in fact it's so bad it made me physically sick kind of like VR sickness,m if that wasn't enough PBO doesn't work at all on my motherboard it does nothing or destroys my performance while increasing watts (32k cbr23@270w with pbo while activating eco mode gives me 32k@80w >< eco works at least )
the 13900k full stock goes to 350w and with msi "boost" (where all p-cores go 5886mhz) 383w so....air coolers and I would say even aio users need not apply
it's not as bad as when I tried to OC the 3960x (430+watts) but getting close and btw it's that threadripper who forced me to go full custom loop
Thank you so much for your post!
I was waiting for 7950X, but the TDP looked bad compared to 5950X (which my brother has). On top of that, some gaming benchmarks have shown 5800X3D to be faster despite 7950X having more total cache. Ok - waiting for 13900K... and its thermals are even worse. I live @ top floor, with no AC, so this is an issue for me. The low noise level is important too. On top of that I get something like motion sickness, similar to when being on a boat. It happens when the framerate drops to around 20 fps / less, or there's stuttering, uneven motion lag. Also, my brother has MSI X570 Unify, which we're satisfied with except for the early memory-related and CPU incompatibility issues. MSI X670E mobo would be the initial choice for me.
I wonder how big part of the stuttering you experienced was related to RAM ECC kicking in (J2C has shown GPU RAM ECC affected his 3DMark runs) and dynamic CPU scaling / throttling - like vs running them at constant lower speed (it made a huge difference in my bro's rig when BIOS was in the infant age).
And I started reading this thread to learn if I should get a computer case, where 360 AiO cooler would fit & work properly. You basically answered everything I might have asked for!
Well.. except for the ECC, dynamic throttling and if I should get 13900K, or wait for 7950X3D, and if custom loop is really a must now if you want acoustic comfort. I'd rather prefer to upgrade less frequently, have less maintenance, and avoid spending $$$ on something with diminishing returns. But if this a new reality, I'll have to accept it.
Not to big of a loss I see, it looks like they still need to work on the stability, a benchmark program should not double the score after a reinstall.
Not the benchmark that was reinstalled - it was entire VAM itself. They did say to only run on a clean install of VAM. But anyway, one of few things to truly push the latest GPUs. But of course no mainstream sites/youtubers would ever mention it - way too non-PC and R18 for them, so they stick to children-safe benchmarks on flat 2D monitors only - pretty boring.
what you see in the single core run at the begininning followed by 3 multi-core cinebenchR23 on the left it's the 7950x and you kind of see it but the temp goes down noticeably slower from <1s to 2s on the intel to 3,4,5s on the AMD 7950x I had the same setup with a 3960x,9900k,5950x,12900k and it's the 1st time I see that it's 100% coming from the IHS note that the single core run on amd is hotter probably because amd clocks are higher than intel on that gen FULL STOCK NO AIB OC on both plateform amd=5781/5505 all cores of a ccd have the same clocks intel=5487/4290 with 2 cores only reaching 5786Mhz
I wish I wouldn't be able to do that comparison and that the 7950x worked fine but it didn't so I built a 13900k 2 weeks later do not buy the MSI X670E meg ace most of the sold motherboards are defective and/or the bios has unacceptable issues..it freezes and crashes...the bios screen ! lots of complaints about memory with it on msi own forums and those people seem "casuals" who haven't tried all I did and seen that the problems are deeper than being forced to run your ddr5 at 4800(2)3600(4) my benchmarks are same or better than the 13900k but gaming is simply impossible rivatuner shows me a straight horizontal 120fps line...it looks like 12fps insane stuttering 😡 in fact it's so bad it made me physically sick kind of like VR sickness,m if that wasn't enough PBO doesn't work at all on my motherboard it does nothing or destroys my performance while increasing watts (32k cbr23@270w with pbo while activating eco mode gives me 32k@80w >< eco works at least ) the 13900k full stock goes to 350w and with msi "boost" (where all p-cores go 5886mhz) 383w so....air coolers and I would say even aio users need not apply it's not as bad as when I tried to OC the 3960x (430+watts) but getting close and btw it's that threadripper who forced me to go full custom loop