Review: AMD Ryzen 9 7950X3D processor

that is true, 7950x3d is unusually low too. I meant more 13600k/13700k vs 7000 non-3d. I frankly didn't read about 7950x3d/13900K that much, not my price range.
now that you meantion a suspicious x3d score, I think they're just looking for more postitive reception of their review from early r7000 adopters.
you know how hub operate, conclusions come first, then you just gotta make the right charts for them.

 

look how their 7700x score is 96.2% but the other 12 sites average 84.8%. Same way, 13700K averages 98,8% on other sites but 94,9% on HUB. So 99% vs 85% in reality is 95% vs 96% on HUB.
that is some next level BS, catering for their YT fanbase.

 

They addressed that in a follow up video, as they were one of the few sites that tested with faster memory. Sites who also used 6000 MT/s memory got similar results.
They even made a memory scaling test with Zen4, using 5200 MT/s, 5600 MT/s and 6000 MT/s.

 

E-waste cores are what they are... and the fact that intel hasn't got into their heads that gamers do no benefit from or want E-waste cores, and made gaming specific cpu's like amd is doing with the 3d chips, is beyond me.

 

Indeed, memory makes a massive difference on ryzen, and a comparatively small difference on intel.

 

7950X3D actually lose in 1% low FPS at 4K LOL, perhaps 7950/7800X3D would be best CPU for esport gamers who play competitively at 1080p, for casual gamers, not so much

TPU test with 6000MT DDR5 for both Intel and AMD too

 

what a bunch of rubbish

https://www.purepc.pl/test-pamieci-...c-z-procesorem-intel-core-i5-13600k?page=0,28

 

TPU's reviews aren't a good indication either, they're just gpu locations at lower resolutions, just like G3D's. They are not cpu intensive places/scenes at all. and 1% is margin of error really, and even of it isn't it's unnoticeable, it's not like HUB making a 14% difference in favor of intel a 2% difference for amd.

 

I would rather look the avg and minimum FPS result at the resolution that I'm playing at (4K) rather than reading through 1080p reviews LOL.

Yes i know that 1080p is still the default way to bench CPU, but the best CPU at 1080p should also be the best at 4K too, but in the case of 7950/7800X3D they are definitely not.

Should be interesting when TPU release 50+ games benchmark between 13900k vs 7950X3D

 

Wow, I find this article/review very poorly setup and the testing results are NOT accurate for gaming. I have a 7950X3D and if you don't set the CPU correctly via EFI, you'll not get it's full gaming potential performance.

This is the Guru3D article by Hilbert using AIDA64 CPUID and these are the major issues:



1. SMT enabled, should NEVER enable SMT for gaming when you have 16 real cores, this will hurt performance significantly.
2. There is no BIOS/EFI 9922 ? ... there is a 0922 EFI for the Hero?
3. Why is Integrated Video enabled? Turn that off as it wastes resources and generates heat and consumes power.
4. What happened to the rest of your L3 Cache? 64MB?? Mine shows 192MB ... you have a serious issue here.
5. Core voltage at 1.376v is very high for just 3600Mhz. My 7950X3D is 1.119v at 5200Mhz.

Here is my AIDA64 CPUID:



Here is my Time Spy score:



My setup is 13.2% higher CPU score and 7.9% higher overall than the results you listed here. Was Dynamic Preferred Cores set to "Cache" in EFI?



So all the gaming scores listed in this article are all wrong/skewed.

I highly recommend Hilbert H. revisits his process in order to provide readers with a more accurate figure of 7950X3D gaming performance.

Rob.

 

@V8VENOM best to tag Hilbert ( @Hilbert Hagedoorn ) if you want his attention. He has in the past tested SMT enabled and disabled but found mixed results in his testing. But that was several years ago involving 8c/16t AMD CPUs and may have influenced his testing methods. I can understand 16c/32t would be a whole different ball game today and SMT disabled may indeed merit further investigation.

 

@Hilbert Hagedoorn ... please see above post. The 7950X3D is not like any prior CPU from AMD and needs some "adjustments" in order to get significant performance gains in games.

@alanm ... thanks, yes times have changed, but there are other odd issues with Hilbert's testing such as only 64MB L3 cache being report when it should be 192MB and the EFI version that doesn't exist ... and is VERY important to set CPPC Dynamic Preferred Cores to "Cache" otherwise you have to do some trickery with Micorsoft's Xbox game bar and that might NOT always work correctly ... so preferred option is to set CPPC via EFI.

There does seem to be a work around to the CCD0/1 core "mapping" that requires one to disable EXPO profiles first, then set core mappings (disable), then enable EXPO profiles ... but SMT disable and CPPC seems to work well from my testing.

Rob

 

nonsense, the difference is usually within error and the cases where big differences are observed are due to the application itself.

 

Yes, and games are inherently one of those "applications" ... hence why I qualified my statement with "gaming". All games operate on a main thread from which other threads are created, if they didn't their simulated environment would make no sense and be out of order that's why you always see one core at a much higher usage (the main thread manager). If you want to go down this road, that's fine, but I've been a software engineer for 35 years (yes, almost as many years as your age, take a look when I joined Guru3D), I know thread management, thread pools, worker threads and parallelism, especially when it comes to games. There are very few (maybe Ashes of the Singularity) just because the nature of the game worked well for DX12 and parallel computing. But the games listed in this Guru3D review where not of that "type of application".

One key element to having SMT/HT disable is less power consumption and less heat. For those that overclock, having SMT/HT off will provide for higher real core frequencies and overall stability improves which adds to it's advantage.

If one uses 3DSMax or Adobe Suite with lots of Rendering or CAD software then yes, SMT/HT enabled will benefit those applications because rendering is not order dependent.