Intel Core i9 7900X Linux Benchmarks

[sverris]

I would be really interested in the temperatures and the fan noise during benchmarking with the air cooler.

You will surely need very good cooling... At golem.de ( in German ), they were recently testing power consumption of the i9-7900X, and it may become enormous: up to 80 W more than i7-6950X, when all cores run at maximum.

[lucasbekker]

Does anyone know which of these benchmarks make use of MKL, instead of openBLAS? The numpy benchmark for example might be MKL, but is there a way to check?

[Jabberwocky]

You will surely need very good cooling... At golem.de ( in German ), they were recently testing power consumption of the i9-7900X, and it may become enormous: up to 80 W more than i7-6950X, when all cores run at maximum.

Tom's hardware claims the Core i9-7900X went up to 364W when overclocked to 4.8GHz. The review also said never mind air, some water-cooling solutions were not good enough... and here I'm sitting wondering about the Core i9-7980XE Extreme Edition.

[Michael]

Tom's hardware claims the Core i9-7900X went up to 364W when overclocked to 4.8GHz. The review also said never mind air, some water-cooling solutions were not good enough... and here I'm sitting wondering about the Core i9-7980XE Extreme Edition.

I'll have air cooling results shortly.

[Michael]

Does anyone know which of these benchmarks make use of MKL, instead of openBLAS? The numpy benchmark for example might be MKL, but is there a way to check?

http://openbenchmarking.org/innhold/3ca0b23d9bf6abb4028a7168871714509f263da8

[sverris]

Tom's hardware claims the Core i9-7900X went up to 364W when overclocked to 4.8GHz. The review also said never mind air, some water-cooling solutions were not good enough... and here I'm sitting wondering about the Core i9-7980XE Extreme Edition.

Thanks. Golem did not go into overclocking. They claim that their whole system needed, when running at Max., 254 W (Blender). Temperature went to 79° Celsius, with a Noctua NH-U14S running on its limit.

[Michael]

Michael, may I ask you to run the compilation benchmark O:-)?

Yep, timed kernel compilation and such will be coming in a few articles and more this week. Just an oversight forgetting to add build-linux-kernel (thought I did, as my favorite compilation test) when starting this comparison.

[caligula]

Yep, timed kernel compilation and such will be coming in a few articles and more this week. Just an oversight forgetting to add build-linux-kernel (thought I did, as my favorite compilation test) when starting this comparison.

Does the kernel compilation run in tmpfs?

[arjan_intel]

Does anyone know which of these benchmarks make use of MKL, instead of openBLAS? The numpy benchmark for example might be MKL, but is there a way to check?

they use whatever ubuntu ships as default, e.g. not MKL.

[efikkan]

Any particular reason for the +50% perf on some tests? Is because of AVX 512? Optimized binary? Cache?

Higher IPC due to a completely redesigned cache hierarchy, with four times the L2 cache.

Nice results, but for US$ 1K, 140W, I prefer Ryzen 1800X

If you want to look at Ryzen 1800X, then compare it with i7-7800X, which has better overall performance at a great price.
i9-7900X will compete with larger threadrippers, due to Intel's much higher IPC, you should look at 12-16 core Threadripper when they arrive.

All I takeaway from this is that Intel has purposely been shelving their next iterations in order to maximize profits on current upgrades while the competition catches up.

No one is stupid enough to belive that.
Intel have not kept CPUs back.