News Ryzen 9 9950X runs 16% faster on an Intel-optimized Linux distro

Status
Not open for further replies.
By the way, Phoronix went thru an analysis of the new AVX512 uarch of Zen 5, which is absolutely breathtaking. It just proves how today tech reviewers are nothing more than a bunch of Influencers.

"When taking the geometric mean of the 90 benchmarks used for this AVX-512 on/off comparison, the Zen 5 AVX-512 implementation with the Ryzen 9 9950X saw its performance go up by 56% while the Ryzen 9 7950X Zen 4 with its "double pumped" AVX-512 implementation saw its performance go up by 41%. Zen 4's AVX-512 was great and now with Zen 5 it's even better with the SKUs having the full 512-bit data-path. This is especially good news for Ryzen 9000 series (EPYC 4005?), EPYC Turin, etc."
 
Last edited:
Interesting to see how the Linux distro affects performance. I would also be interested to see these benchmarks run on the 7950X to get an even bigger comparison.
Go check that article, I think you'll find what you're looking for : performance against 7950X and against Windows 11 - if you extrapolate, running Clear Linux with a 9950X VS. Windows 11 gives you close to 25% extra performance with a 9950X.
 
Last edited:
  • Like
Reactions: artk2219
By the way, Phoronix went thru an analysis of the new AVX512 uarch of Zen 5, which is absolutely breathtaking. It just proves how today tech reviewers are nothing more than a bunch of Influencers.

"When taking the geometric mean of the 90 benchmarks used for this AVX-512 on/off comparison, the Zen 5 AVX-512 implementation with the Ryzen 9 9950X saw its performance go up by 56% while the Ryzen 9 7950X Zen 4 with its "double pumped" AVX-512 implementation saw its performance go up by 41%. Zen 4's AVX-512 was great and now with Zen 5 it's even better with the SKUs having the full 512-bit data-path. This is especially good news for Ryzen 9000 series (EPYC 4005?), EPYC Turin, etc."
How is this being an influencer? Did they state anything that wasn't factual? No.
 
What? No Linux Mint? Considering how popular Mint is its a glaring omission.
He did benchmark a bunch of distros recently. Apart from Ubuntu 24.04 that was quite a bit slower than all the other ones (gcc 13), and Clear Linux that's a cut above, all distros run within a 2% margin of error from each other. As long as you're using the same compiler with the same options (gcc 14 with -O2), frankly, there's no real point.
 
  • Like
Reactions: jlake3 and dwd999
These improvements will also be the case for CPUs going back 16 years..

A lot of these improvements are due to compiler flag changes for the packages. And clear Linux and cacheos uses other pathes and different scheduler.

But since zen 5 is already faster on Linux than windoze, and has resulted in some games performing better, using these might make sense for someone to game with the best perf
 
How is this being an influencer? Did they state anything that wasn't factual? No.
Doing a cinebench test and adding a couple of gaming benchmarks exclusively in a windows environment is not a CPU review.

Phoronik made a geomeans with over 400 benchmarks and demonstrated an IPC above AMD's owns number. Nobody can deny it.

9950x.jpg
 
  • Like
Reactions: NinoPino
Doing a cinebench test and adding a couple of gaming benchmarks exclusively in a windows environment is not a CPU review.

Phoronik made a geomeans with over 400 benchmarks and demonstrated an IPC above AMD's owns number. Nobody can deny it.
By that logic testing using built in accelerators an instruction set which has minimal real world client use and exclusively testing on Linux isn't a CPU review either.
 
Last edited:
Michael tested the CPUs as they are, of what "built in accelerators" are you talking about ?
AVX512 (and I should have said instruction set not accelerator)

This would be like testing Adobe programs that can leverage Quicksync and implying that's indicative of overall CPU performance. It's important information to have, but isn't applicable to overall general performance.
 
Last edited:
Status
Not open for further replies.