Intel Nehalem vs. Ice Lake Benchmarks - Including Clock + Power + Thermal Metrics

[phoronix]

Phoronix: Intel Nehalem vs. Ice Lake Benchmarks - Including Clock + Power + Thermal Metrics

As part of the exciting benchmark week and our ongoing tests of Intel Ice Lake on Linux, this next piece has been driven out of curiosity... While recently I posted new benchmark results of Intel Haswell to Ice Lake laptop performance, what about going further back like to the days of Nehalem? Here is that comparison of Core i7 Nehalem to Core i7 Ice Lake including power / performance-per-Watt data, thermal, and performance-per-MHz data too. Enjoy this fun comparison for how the Intel mobile performance on Ubuntu has evolved over the past decade.

http://www.phoronix.com/vr.php?view=28567

[Mario Junior]

Tests already start wrong when you have one device using high-speed storage while the other uses a regular hard drive.

[ms178]

A TDP-constrained environment plus modern software which uses AVX should be the worst case for the older architecture. I'd think that on desktop, let's say between Westmere @ 4.2 Ghz vs. Ice Lake @ 4.2 GHz (with all mitigations disabled) the differences shouldn't be as high (albeit with a similar trend). But at least in regards to gaming I wouldn't expect anything too drastic.

[Michael]

A TDP-constrained environment plus modern software which uses AVX should be the worst case for the older architecture. I'd think that on desktop, let's say between Westmere @ 4.2 Ghz vs. Ice Lake @ 4.2 GHz (with all mitigations disabled) the differences shouldn't be as high (albeit with a similar trend). But at least in regards to gaming I wouldn't expect anything too drastic.

Too bad there isn't any Ice Lake desktop at least not yet.

[Michael]

Tests already start wrong when you have one device using high-speed storage while the other uses a regular hard drive.

Good thing I wasn't running storage benchmarks.

[ms178]

Too bad there isn't any Ice Lake desktop at least not yet.

Haha, indeed. And it was pure scientific curiosity on my part. I guess people would and should spend their money with AMD on the desktop anyway at the moment.

[stormcrow]

Haha, indeed. And it was pure scientific curiosity on my part. I guess people would and should spend their money with AMD on the desktop anyway at the moment.

If you're doing anything other than ML algorithms, apparently so. At least if you're going high end with Threadripper. With gaming rigs and general desktop I doubt anyone would notice if Facebook loaded 0.3 seconds faster or if you're system would render $GAME faster ... except that it's locked at 60 FPS anyway thanks to the engine being optimized for consoles instead of PCs like most popular games.

[coder]

Too bad there isn't any Ice Lake desktop at least not yet.

True. Comet Lake is the next desktop line (also, high-end mobile) and will remain @ 14 nm. I'm not clear on the relationship of the uArchs, but we know that Comet Lake won't have AVX-512.

Anyway, Ice Lake will be coming to servers & featuring PCIe 4.0, in early 2020.

[coder]

Good thing I wasn't running storage benchmarks.

So, is it your belief that all the relevant data files used by the benchmarks was read from disk cache (at least, to the extent it could've significantly affected performance)?

Anyway, thanks for the benchies.

BTW, on the outside chance you used sort to order your performance comparisons, you can specify -n to do a numerical comparison instead of lexical. Otherwise, that was a nice chart.

[Michael]

So, is it your belief that all the relevant data files used by the benchmarks was read from disk cache (at least, to the extent it could've significantly affected performance)?

Anyway, thanks for the benchies.

BTW, on the outside chance you used sort to order your performance comparisons, you can specify -n to do a numerical comparison instead of lexical. Otherwise, that was a nice chart.

Aside from potentially the code compilation workloads, it's unlikely the HDD skewed the results by any substantive manner that would yield like a real measurable impact.

All of the graphs are generated by PTS/pts_Graph.