MAA Posted February 22, 2024 Posted February 22, 2024 Hello, 1) CPU Queen Why is the result of the newest Threadripper 7975WX worse than that 3970X (and even old 2990WX)? 2) FPU Mandel Why is the result of the newest Threadripper 7975WX worse than that 3970X? it should be twice as fast, as can be seen from the result in FP64 Raytrace. Quote
Fiery Posted March 18, 2024 Posted March 18, 2024 On 2/22/2024 at 7:44 PM, MAA said: Hello, 1) CPU Queen Why is the result of the newest Threadripper 7975WX worse than that 3970X (and even old 2990WX)? 2) FPU Mandel Why is the result of the newest Threadripper 7975WX worse than that 3970X? it should be twice as fast, as can be seen from the result in FP64 Raytrace. 1) The current Queen benchmark scales quite bad on many-core systems. It typically stops performing any better at around 30 cores. We're already working on a new Queen benchmark that would replace the current out-dated one in the next AIDA64 stable update. 2) Our suspicion is that there may be cooling issues about the 7975WX test system. We'll make sure to improve on that and re-measure the Mandel benchmark result. Quote
MAA Posted May 28, 2024 Author Posted May 28, 2024 >>We're already working on a new Queen benchmark that would replace the current out-dated one in the next AIDA64 stable update latest stable 7.30.6900 still has this issue Quote
Fiery Posted May 29, 2024 Posted May 29, 2024 10 hours ago, MAA said: >>We're already working on a new Queen benchmark that would replace the current out-dated one in the next AIDA64 stable update latest stable 7.30.6900 still has this issue I'm afraid the new Queen benchmark couldn't get finished before the v7.30 update. It is scheduled for the next update. Quote
MAA Posted July 25, 2024 Author Posted July 25, 2024 Why new 7.35.7000 CPU CheckMate benchmark do not have Threadripper 7975WX result? Quote
MAA Posted August 11, 2024 Author Posted August 11, 2024 Why in ZLib the result of Threadripper 7975WX is almost the same as the old 3970X? Maybe need to use next gen Zlib? https://github.com/zlib-ng/zlib-ng Quote
Fiery Posted August 16, 2024 Posted August 16, 2024 On 7/25/2024 at 7:10 PM, MAA said: Why new 7.35.7000 CPU CheckMate benchmark do not have Threadripper 7975WX result? We'll add it in the next AIDA64 beta update. Quote
Fiery Posted August 16, 2024 Posted August 16, 2024 On 8/11/2024 at 1:48 PM, MAA said: Why in ZLib the result of Threadripper 7975WX is almost the same as the old 3970X? Maybe need to use next gen Zlib? https://github.com/zlib-ng/zlib-ng 2 reasons: ZLib is a GPR benchmark and so it doesn't "jump" because of AVX-512 support. Other reason: the 7975WX system has SMT disabled, while the old 3970X has SMT enabled. Quote
Fiery Posted August 16, 2024 Posted August 16, 2024 5 minutes ago, MAA said: Why is SMT disabled? In certain workloads performance increases after disabling SMT. Quote
MAA Posted August 16, 2024 Author Posted August 16, 2024 in that case both results should be published Quote
MAA Posted October 22, 2024 Author Posted October 22, 2024 Why is Threadripper 7975WX's FPU Mandel score worse than 3970X again? Quote
Fiery Posted October 22, 2024 Posted October 22, 2024 2 hours ago, MAA said: Why is Threadripper 7975WX's FPU Mandel score worse than 3970X again? We now include SMT enabled results. In certain heavy AVX-512 workloads with SMT enabled 7975WX performs worse than with SMT disabled. While in other workloads it's the other way round Quote
MAA Posted October 22, 2024 Author Posted October 22, 2024 8 hours ago, Fiery said: In certain heavy AVX-512 workloads with SMT enabled 7975WX performs worse than with SMT disabled. in that case this code is not suitable for the reference benchmark Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.