AMD Ryzen 9 7950X - Benchmarks

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

User avatar
Ras
Posts: 2720
Joined: Tue Aug 30, 2016 8:19 pm
Full name: Rasmus Althoff

Re: AMD Ryzen 9 7950X - Benchmarks

Post by Ras »

Jouni wrote: Thu Oct 06, 2022 9:53 amAnyone know where i can buy a 7000 series desktop?
Don't buy a pre-built PC, in particular not in the upper price range when you expect more than garbage. Build it yourself. Also, I'd wait for B650 mobos because they should still be perfectly fine, but quite a bit cheaper than X670(E) ones. You just don't really need e.g. PCIe 5.
Rasmus Althoff
https://www.ct800.net
Werewolf
Posts: 2058
Joined: Thu Sep 18, 2008 10:24 pm

Re: AMD Ryzen 9 7950X - Benchmarks

Post by Werewolf »

Jouni wrote: Tue Sep 27, 2022 8:53 am Stockfish 9 is from 2018 :lol: . Bench with SF 15 please? Seriously they say you can't beat anymore 8 core Stockfish from equal opening. Only benefit from fast processor is, that you solve (artificial) testpositions faster.
This is a serious point which needs discussing.

My contribution is this: greater depth would allow for a more accurate evaluation of the position which is useful for an analyst.
perejaslav
Posts: 240
Joined: Sat Mar 18, 2006 4:01 am
Location: Cold

Re: AMD Ryzen 9 7950X - Benchmarks

Post by perejaslav »

What about 4090 and Lc0?
Anyone has tested them?
Jouni
Posts: 3758
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: AMD Ryzen 9 7950X - Benchmarks

Post by Jouni »

Yes CSS forum:

backenbench 3090=Benchmark batch size 224 with inference average time 14.3069ms - throughput 15656.8 nps.
backendbench 4090=Benchmark batch size 222 with inference average time 6.63731ms - throughput 33447.3 nps.

630 Watt bei Vollast :!: :!:
Jouni
Werewolf
Posts: 2058
Joined: Thu Sep 18, 2008 10:24 pm

Re: AMD Ryzen 9 7950X - Benchmarks

Post by Werewolf »

Jouni wrote: Sun Oct 16, 2022 9:07 pm Yes CSS forum:

backenbench 3090=Benchmark batch size 224 with inference average time 14.3069ms - throughput 15656.8 nps.
backendbench 4090=Benchmark batch size 222 with inference average time 6.63731ms - throughput 33447.3 nps.

630 Watt bei Vollast :!: :!:
Over double the speed, not bad!
CornfedForever
Posts: 650
Joined: Mon Jun 20, 2022 4:08 am
Full name: Brian D. Smith

Re: AMD Ryzen 9 7950X - Benchmarks

Post by CornfedForever »

Has anyone seen NPS specific benchmarks for any of the new Intel processors? I'm coming up empty...
jpqy
Posts: 556
Joined: Thu Apr 24, 2008 9:31 am
Location: Belgium

Re: AMD Ryzen 9 7950X - Benchmarks

Post by jpqy »

Modern Times
Posts: 3781
Joined: Thu Jun 07, 2012 11:02 pm

Re: AMD Ryzen 9 7950X - Benchmarks

Post by Modern Times »

Many more reviews out now. Dropping to 105W ECO mode seems the most sensible thing for most users, unless time is money and every second counts. The extra performance between 105W and normal spec is achieved with big increases in power and heat, the law of diminishing returns is very much in play. AMD have done the same thing as Intel here, chased every last bit of performance for bragging rights, but it just isn't worth it for most people.
CornfedForever
Posts: 650
Joined: Mon Jun 20, 2022 4:08 am
Full name: Brian D. Smith

Re: AMD Ryzen 9 7950X - Benchmarks

Post by CornfedForever »

jpqy wrote: Tue Nov 08, 2022 4:43 pm Intel Core i9 13900K & i5 13600K

https://ipmanchess.yolasite.com/amd---i ... -bench.php
Thanks guys.

I have to say...there are just too many numbers on the ipman site. I can't make much of it because of all the 'cores'.
For your average 'chess person', interested in some analyis, you might use lets say 8 threads or 12 cores (?) for analysis.

Trying to compare, lets say the Ryzen 9 7950 or 7900 against the i9 13900 or i7 13700 looks impossible. :cry:

Even then...'generic' multi-threaded performance when you consider so many of the Intel processors are largely 'efficiency' cores running at a much slower pace than the 'performance' cores, it's hard (for my little mind) to square how well the top intel processors do.

And then again, as Modern Times mentions - dropping to 105W ECO mode, could be sensible for overnight analysis...but much does it bring nps down?
CornfedForever
Posts: 650
Joined: Mon Jun 20, 2022 4:08 am
Full name: Brian D. Smith

Re: AMD Ryzen 9 7950X - Benchmarks

Post by CornfedForever »

Long story short: I guess when it comes to multi-threaded engine analysis (lets say 5 min/move or overnight even), the improvements and MORE efficiency cores added (even at a slightly lower clock speed) results in a higher NPS for the new Intel 13th gen processors compared to the new (or last gen) AMD Zen 4 processors?

I remember the previous Zen 3 knocking the socks off the 12th gen Intel processors in NPS...not anymore.