AMD Ryzen 7 CPUs Shipping 2 March, Pre-Order Today

[pal666]

windows 10 works, only heard windows 7 has issue's.....

so you don't need any kernel patches for ryzen to work. kernel patches are needed for some advanced features

[pal666]

I don't care about Windows at all.

but maybe you care about reasoning?

[extraymond]

http://openbenchmarking.org/test/pts/tensorflow

Nice!!! Hope bridgman can get you some free samples haha.

[qsmcomp]

so you don't need any kernel patches for ryzen to work. kernel patches are needed for some advanced features

e.g. detection of advanced instruction sets, errata workarounds

[hugo8621]

You really want to believe, but their chart is very clear:

i7 7700K (4 cores, 8 threads): 967
Ryzen 7 1700 (8 cores, 16 threads): 1410 (+46%)

I'm not expecting reality to be better than the pre release numbers.

comparing also a 95W super high clocked 4core CPU vs. a 65W lower clocked 8core... so putting that in a perf/watt perspective, the i7 sucks quite a lot I'd say, no?

I'm really curious about Ryzen AVX performance vs. Intels Kaby Lake

[Ehvis]

idiot, why did you ignore benchmark of 8/16 ryzen vs 8/16 intel with ryzen winning?
you have only +46% here because this load does not scale linearly

The idiot is the person that doesn't read. I said higher end ones looked better.

[pal666]

The idiot is the person that doesn't read. I said higher end ones looked better.

the idiot is the one who thinks that 8/16 should be twice as fast as 4/8
if you are interested in single thread spead, you run single threaded benchmark. and when you run single threaded benchmark on 4 core ryzen, you compare results against similarly priced intel, not against one more expensive than 8/16 ryzen

[efikkan]

Michael:
Remember to also include scores with XFR disabled for the X models, since the XFR score will not be reproducible from build to build.

Ryzen has 52% IPC improvement over Excavator? Holy cow, that blows their 40% target out of the water! Zen is going to kick some serious butt. If you don't already own AMD stock, get on it, AMD is going to take back a LOT of marketshare from intel.

Even the Bulldozer revisions claimed significant IPC gains that never materialized, and also Intel claims ~10% IPC gains for every "generation" since Sandy Bridge. Both are referring to edge cases, not general performance. The total gains from Sandy Bridge to Skylake is about 6-15%.

For the past couple of weeks the hysteria is claiming that Ryzen will have higher IPC than Skylake/Kaby Lake. People always forget that pre-release benchmarks are always cherry-picked, and these CPUs also feature a new feature called XFR, which will be an extra turbo which scales way beyond it's turbo target. XFR will be dependent on cooling and sample quality. So when we see these "leaks" of Ryzen at 3.6/4.0, it may in reality be running anywhere from 4.0 to 5.0 GHz due to XFR. So anyone basing IPC claims on this not understanding what IPC really is. To actually measure IPC, we need to lock the clock frequency and disable any turbo/boost, someone will test this soon.

The time to invest in stocks is when it's low, not when it's nearing it's peak. The smart investors bought stocks a year ago and will dump them just before release day, not because Ryzen is bad, but because it can't live up to the expectations.

[Ehvis]

the idiot is the one who thinks that 8/16 should be twice as fast as 4/8
if you are interested in single thread spead, you run single threaded benchmark. and when you run single threaded benchmark on 4 core ryzen, you compare results against similarly priced intel, not against one more expensive than 8/16 ryzen

I'll leave you to your fanboyism. I'll wait till the real world catches up with some actual data.

[phoronix_anon]

Michael, I'd be very interested in NAMD benchmarks on the consumer Zen CPUs. I'd also like to see how a pair of 1800X's perform in tandem with a pair of GTX 1080s on CUDA-enabled NAMD 2.12.

At a bigger level it might give some additional insight into what we should expect from the Naples server versions of these chips.