• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Zen 7 speculation thread

Page 32 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.
Fjodor2001 said:
8 core N+1 gen is also typically faster than 8 core N gen. And:
16 core N gen is typically faster than 8 core N gen.
Click to expand...

Normalizing the clocks (16 core typically gets extra 200 MHz), there would be a lot of ties within the same generation, and N+1 8 core would get clearer wins over 16 core N generation if 8 core was not held back by ~200 MHz.
 
Last edited:
Joe NYC said:
Normalizing the clocks (16 core typically gets extra 200 MHz), there would be a lot of ties withing the same generation, and N+1 8 core would get clearer wins over 16 core N generation if 8 core was not held back by ~200 MHz.
Click to expand...
But the real-life clocks in the actual SKUs are not normalized.

You could also ”normalize” for X3D difference and conclude that 9950X and 9950X3D perform the same.
 
Fjodor2001 said:
Because no X3D.
Click to expand...
And because cross cc'd penalty is high. They are not gaming chips. Some people buy them but it's tiny as a proportion of chips sold. Boo me wants high core count CPUs (well, people think they do but the hobbyists want them but won't pay and the people who really do but epyc). Gaming will, forever, by the single CCD option with max cache.
 
511 said:
Ahh yes they messed up AVX-512 but they are the biggest reason for it's existence not to mention NVL give us APX/AVX 10.2 and Zen 6 will be stuck with ISA of server GLC (except AMX) so the same thing applies to AMD as well
Click to expand...
By the time people use those zen will have them, well, the bits people use.
 
These extensions are a non issue for now. Later maybe they will help, but it's gonna take a long while for the baseline to include them. You have to add them in to not miss out on the future, but for now it doesn't matter
 
inquiss said:
Boo me wants high core count CPUs (well, people think they do but the hobbyists want them but won't pay and the people who really do but epyc). Gaming will, forever, by the single CCD option
Click to expand...
…and the single CCD will get higher core count too in the next generations, whether you like it or not.
 
inquiss said:
Boo me wants high core count CPUs (well, people think they do but the hobbyists want them but won't pay and the people who really do but epyc).
Click to expand...
The hobbyists will get 52C with NVL-S about 1 year from now without having to pay HEDT prices. And they don’t need EPYC.
 
Fjodor2001 said:
The hobbyists will get 52C with NVL-S about 1 year from now without having to pay HEDT prices. And they don’t need EPYC.
Click to expand...
I am not a fan of such high core counts unless bottlenecks in other areas are addressed

I remember that high core count intel cores failed when performing UE5 decompression — the software was exactly designed to utilize the high core count of intel CPUs
 
marees said:
I am not a fan of such high core counts unless bottlenecks in other areas are addressed

I remember that high core count intel cores failed when performing UE5 decompression — the software was exactly designed to utilize the high core count of intel CPUs
Click to expand...
The memory speed will be increased. 8000 MT/s official, and potential for 10000 MT/s. Caches will be improved on NVL-S too.

So we don’t know if/what bottlenecks there will actually be per type of workload, if any. We have to wait for test results.
 
Last edited:
DrMrLordX said:
Nova Lake-S won't be an adequate replacement for EPYC/Threadripper.
Click to expand...
Depends on the workloads. Not everyone needs more cores than NVL-S.

It’ll grab a lot of market share currently covered by low-mid TR SKUs. Simply because NVL-S will have a lot better perf/$ in that MT perf range.
 
511 said:
It will be for people who want more core for cheap if it had more PCI lanes than it would have replaced thread ripper at lower end
Click to expand...
NVL-S will get more PCIe lanes too. Not as many as TR, but for a lot of use cases in the MT perf range that NVL-S covers you don’t need all those TR PCIe lanes anyway.
 
511 said:
it will beat Zen 6 24C CPU for Sure in CPU MT Task no way it will beat Zen 7.
Click to expand...

So what makes you so sure? Or is that a wishful thinking that Intel will finally deliver?

Unless we are talking CB24, then well who cares😉
 
DrMrLordX said:
What workload needs more than 48t but less than 64t?
Click to expand...
Idk where you got that range from. It's not like a workload usually needs an exact number of threads or cores. They'll just execute faster the more you have (until you run into some bottleneck). Also, not only T count, but C count matters. And perf/$ matters too. So you basically buy something that provides sufficient perf within your budget.

If you're looking for something in the range of MT perf that 52C NVL-S provides it'll be a good option. For the next step up you'll have to enter into the TR platform, which will be be substantially more expensive and thus usually not worth it if it's only a little faster than NVL-S. But if you're looking for higher end of TR, then NVL-S is not an option.
 
Last edited:
Status
Not open for further replies.

TRENDING THREADS

Back
Top