TalkChess.com

With potentially double the code count of the 2990wx, I guess the new one will cost a lot more.

Dann Corbit wrote: ↑Sun Aug 18, 2019 6:25 am With potentially double the code count of the 2990wx, I guess the new one will cost a lot more.

What do you mean by "double the code count"? Ahhh...you mean double the "core" count! ( Yes the 64-core version will be more expensive if/when it arrives). I'm not thinking about that yet as I have no specs on it and see little point in speculation. The 32-core is a little more tangible and likely more affordable.

Why does everyone think that? The AMD EPYC 7502P (32-core single chip version) is listed at just $2300. The trade being made is less clock speed but double the I/O and double the memory channels compared to a 32-core TR. It's hard to imagine them charging more for a 32-core Threadripper than for EPYC. I think one of the primary differences is the I/O chip. On EPYC it's 14nm on Consumer models it's 12nm. I think the trade off is one of cost, speed, and extended I/O capabilities. There server chips have the slower I/O chips but they are much larger and support multi-CPU etc. While consumer get faster I/O chipsets but they have less features. I'll be curious to see if Threadripper comes with the 12nm or 14nm I/O chips or if it has a new design just for that series.

So, just because Intel has been screwing their customers by crippling chips so they can charge higher prices for them, doesn't mean AMD will take the same road.

In fact, they have as much as stated that their will NOT be crippling chips just to segment the market. Lisa Su as much as stated this during her last key note address. AND "fair" pricing was mentioned. So, I'm guessing the price will be comparable to the 2990WX. i.e. $1800 US +/-$200. In any case, we shouldn't have much longer to wait to find out. I was thinking about an EPYC system, single CPU 32-core P model. It's not that bad of a trade off if you want the extra Memory and I/O.



Regards,

Zenmastur

AMD 3rd Gen Ryzen Threadripper ‘Sharkstooth’ With 32 Zen 2 Cores Possibly Spotted in Geekbench – Up To 35% Faster Than Ryzen Threadripper 2990WX

https://wccftech.com/amd-ryzen-threadri ... u-spotted/

JP.

I'm seriously thinking about building a system with 2 Epyc 7302 or 7402, I guess that within two weeks or so these processors will be available. The only thing that holds me back is that I still don't know whether the BMI2 (PEXT/PDEP) instructions are fixed with Zen2.

@joost ,then maybe you are interested to read this review:

AMD EPYC 7002 Series Rome Delivers a Knockout
https://www.servethehome.com/amd-epyc-7 ... -knockout/

And on Page 8 you see asmFish benchmarks with it!

JP.

jpqy wrote: ↑Sun Aug 18, 2019 3:09 pm @joost ,then maybe you are interested to read this review:

AMD EPYC 7002 Series Rome Delivers a Knockout
https://www.servethehome.com/amd-epyc-7 ... -knockout/

And on Page 8 you see asmFish benchmarks with it!

JP.

Thanks!

I saw (the same?) review on the Phoronix website the day the 7002 series were anounced, Epyc 7002 delivers serious horsepower, but I still would like to know whether BMI2 has been fixed, pretty soon somebody will test it on a Ryzen 3000.

For approx. 3000 euro's I can build a system with 2x Epyc 7302, 2x 16 cores 3GHz. base clock, probably a lot faster than a Threadripper, It certainly looks interesting. Unfortunately I have to wait, Epyc Rome is not available yet in the Netherlands.

Ipman has benches with the new AMD Ryzen R9 3900X

and BMI2 is still clearly slower then pop version on AMD cpu's..

http://www.ipmanchess.yolasite.com/amd- ... -bench.php

JP.

jpqy wrote: ↑Sun Aug 18, 2019 8:04 pm Ipman has benches with the new AMD Ryzen R9 3900X

and BMI2 is still clearly slower then pop version on AMD cpu's..

http://www.ipmanchess.yolasite.com/amd- ... -bench.php

JP.

Indeed, it looks like they didn't fix it. I don't need it for computer chess per se, but I'm also doing cryptography and in a new chess engine I'm working on I make heavy use of PEXT() in the evaluation function.

AMD also has slow AVX2, no AVX512 and no BF16 like the upcoming Intel 'Cooper Lake' chips. If you just use the machine to run Stockfish or any other chess engine it probably doesn't matter, but if you like to program and experiment with new algorithms it could be a drawback.

Joost Buijs wrote: ↑Mon Aug 19, 2019 6:44 am

jpqy wrote: ↑Sun Aug 18, 2019 8:04 pm Ipman has benches with the new AMD Ryzen R9 3900X

and BMI2 is still clearly slower then pop version on AMD cpu's..

http://www.ipmanchess.yolasite.com/amd- ... -bench.php

JP.

Indeed, it looks like they didn't fix it. I don't need it for computer chess per se, but I'm also doing cryptography and in a new chess engine I'm working on I make heavy use of PEXT() in the evaluation function.

AMD also has slow AVX2, no AVX512 and no BF16 like the upcoming Intel 'Cooper Lake' chips. If you just use the machine to run Stockfish or any other chess engine it probably doesn't matter, but if you like to program and experiment with new algorithms it could be a drawback.

They have full AVX2 support and all data paths and EU's are now 256-bit. So, AVX2 shouldn't be "slow" on the new CPU's. AVX2 performance is on par with all intel CPU's EVEN when executing highly optimized Intel code. No AVX-512 support that I'm aware of.

Regards,

Zenmastur

Zenmastur wrote: ↑Mon Aug 19, 2019 12:03 pm

Joost Buijs wrote: ↑Mon Aug 19, 2019 6:44 am

jpqy wrote: ↑Sun Aug 18, 2019 8:04 pm Ipman has benches with the new AMD Ryzen R9 3900X

and BMI2 is still clearly slower then pop version on AMD cpu's..

http://www.ipmanchess.yolasite.com/amd- ... -bench.php

JP.

Indeed, it looks like they didn't fix it. I don't need it for computer chess per se, but I'm also doing cryptography and in a new chess engine I'm working on I make heavy use of PEXT() in the evaluation function.

AMD also has slow AVX2, no AVX512 and no BF16 like the upcoming Intel 'Cooper Lake' chips. If you just use the machine to run Stockfish or any other chess engine it probably doesn't matter, but if you like to program and experiment with new algorithms it could be a drawback.

They have full AVX2 support and all data paths and EU's are now 256-bit. So, AVX2 shouldn't be "slow" on the new CPU's. AVX2 performance is on par with all intel CPU's EVEN when executing highly optimized Intel code. No AVX-512 support that I'm aware of.

Regards,

Zenmastur

Benchmarks show the same AVX2 performance because they compare 56 Intel cores with 128 AMD cores, so that is not really on par. AMD gives you a lot of cores for the money, but Intel still has the higher performance when you do the comparison with an equal number of cores.