Hammering The AMD Ryzen 7 1800X With An Intense, Threaded Workload

[CochainComplex]

I wished you could test the impact of a decent heatsink and cooling system on the performance of New cpus. Ryzen has auto over clock feature.

true this would be very interessting.

[keivan]

I wished you could test the impact of a decent heatsink and cooling system on the performance of New cpus. Ryzen has auto over clock feature.

[librin.so.1]

I wonder what's perf/power ratio here. After all, a 1800X has a TDP of 95W while the Core i7 5960Xhas a TDP of 140W, aye...

[funfunctor]

Michael if you want a good test workload, try the monster that is compiling LLVM. Probably the largest pain in the ass thing to compile!

[Degra]

Ugh, what does the Ryzen 7 1800X costs vs i7-5960X? it is not mentioned in the article...

1800X, 586 Euros vs 5960X, 1200 Euros
Yeah, the Intel one costs about double.

I hope AMD's problematic 16 thread performance can be patched via kernel updates.

Personally I'm not in a hurry so I'll wait probably for the next CPU generation. i5-4670K is quite good for games but more threads would be nice for actual work.

How come I never see you mention the cheaper AMD part? These one sided price comparisons don't bode well for the neutrality of a benchmarking site...

I agree it's somewhat unfair to not mention you're comparing so differently priced products, the comparison itself is legitimate though. You compare the very best Intel and AMD have to offer.

[faph]

These clock frequencies match that of the Ryzen 7 1800X with its base and turbo frequencies, but the E3-1270 v5 sells for around $300 USD while the 1800X is around $500.
...
Ryzen 7 1800X and Xeon E3 1270 v5 that are clocked the same, but with this Xeon E3 part retailing for almost $200 less for this quad-core + HT workstation CPU.

Ugh, what does the Ryzen 7 1800X costs vs i7-5960X? it is not mentioned in the article...

Michael
How come I never see you mention the cheaper AMD part? These one sided price comparisons don't bode well for the neutrality of a benchmarking site...

[xpue]

More powerful cpu demonstrates better performance -- they suggest gimping it while calling it overpriced. Seems legit.

[Del_]

5, code was compiled with -mtune=generic, so far from what one would use doing soem real work.

I believe the bench uses the tuning from OPM. You will find the logics here:
https://github.com/OPM/opm-common/bl...mization.cmake
In other words, -O2 -mtune=native -DNDEBUG
while -flto is turned off due to compiler bugs. Using more aggresive compiler switches might give you 10-20% better performance.

[bridgman]

explanation of 16 thread result is 4 channels of memory on 5960x

Yep, probably 4 and 8 thread too (unless Michael comes back and says the 5960 was reconfigured with 2 channels for these tests) plus the Ryzen system is probably still running memory at 2133 MHz for now until all the required BIOS updates are available.

What I like about this test is that it's probably one of the more bandwidth-limited benchmarks we have seen, so once memory speeds start to go up there should be some interesting impact on the results.

AIDA64 and Sandra benchmarks show 33-50% bandwidth improvement going from 2133 to 3200 MHz, and there *might* be some additional improvement from the data fabric running faster as well (I say "might" because the numbers don't suggest that inter-CCX communication is a big factor).

[pal666]

explanation of 16 thread result is 4 channels of memory on 5960x