Intel Core i7-1065G7 Ice Lake Linux Performance Benchmarks

[phoronix]

Phoronix: Intel Core i7-1065G7 Ice Lake Linux Performance Benchmarks

Recently I picked up a Dell XPS 7390 Core i7 Ice Lake laptop for finally testing this Intel 10nm+ processor under Linux. I have delivered some results so far like the Windows vs. Linux OpenGL/Vulkan performance and the Spectre impact with Ice Lake while this article is the first of several really drilling down on the CPU performance. In this article are benchmarks showing how the Core i7-1065G7 compares in raw performance and performance-per-Watt to the earlier Core i7-8565U (Whiskey Lake) and Core i7-8550U (Kabylake-R) processors.

http://www.phoronix.com/vr.php?view=28394

[milkylainen]

No matter AMD vs Intel or Intel vs Intel setups. I still think some of the MKL-DNN tests can severely distort the picture of performance.
Nice work on hardware fixes though. I bet just the context switch performance increase is a helluva win.

[davidbepo]

impressive, i was expecting a smaller improvement over i7-8565U

[birdie]

The Timed Linux Kernel Compilation benchmark says it all - IceLake CPU-wise is barely faster than Whiskey Lake despite the new uArch and 10nm node. I suspect most other benchmarks were spiky/short-lived which meant the CPU could boost for their duration which is not possible for compiling the kernel which takes around four minutes to complete.

Ice Lake Iris GPU is twice as fast its Whiskey Lake counterpart but Michael is yet to publish GPU benchmarks.

In short, I'm not impressed, in fact I'm almost disappointed. We've had the SkyLake uArch since 2015(!) and now it's 2019 and Intel still has nothing tangibly faster to replace it.

[schmidtbag]

The Timed Linux Kernel Compilation benchmark says it all - IceLake CPU-wise is barely faster than Whiskey Lake despite the new uArch and 10nm node. I suspect most other benchmarks were spiky/short-lived which meant the CPU could boost for their duration which is not possible for compiling the kernel which takes around four minutes to complete.

Ice Lake Iris GPU is twice as fast its Whiskey Lake counterpart but Michael is yet to publish GPU benchmarks.

In short, I'm not impressed, in fact I'm almost disappointed. We've had the SkyLake uArch since 2015(!) and now it's 2019 and Intel still has nothing tangibly faster to replace it.

Considering nearly everything about this CPU is either equal or lower (clock speeds, transistor size, wattage, heat, etc), I'd say these results are actually pretty good. Compared to the 8565U, it has a 28% lower base clock with a 16% lower boost clock, yet it's 20% faster. I don't think that's bad at all for a 29% die shrink on an aging architecture.

[andrei_me]

Can't wait to AMD's next year APU to finally have some decent competition in the laptop market

[tildearrow]

Typo:

The socket activity performance is much better with the Stress-NG kernel micro-benchmark, thanks to Intel's continued work on hardware-based mitigations for the vairous speculative execution vulnerabilities.

[GruenSein]

Considering nearly everything about this CPU is either equal or lower (clock speeds, transistor size, wattage, heat, etc), I'd say these results are actually pretty good. Compared to the 8565U, it has a 28% lower base clock with a 16% lower boost clock, yet it's 20% faster. I don't think that's bad at all for a 29% die shrink on an aging architecture.

To actually evaluate the new architecture, knowing the average frequency would be nice. I mean, do any of the three CPUs actually spend much time at base clock or at max boost? Personally, I find the improvements modest to decent. Power draw is more or less similar and the "clean" performance advantage is around 11%. Considering that this is achieved using a new manufacturing process, this is... well... OK, I guess. I am more excited to see how AMD's 7nm APUs will measure up.

[schmidtbag]

To actually evaluate the new architecture, knowing the average frequency would be nice. I mean, do any of the three CPUs actually spend much time at base clock or at max boost? Personally, I find the improvements modest to decent. Power draw is more or less similar and the "clean" performance advantage is around 11%. Considering that this is achieved using a new manufacturing process, this is... well... OK, I guess. I am more excited to see how AMD's 7nm APUs will measure up.

Not sure about this CPU in particular, but a lot of Intel's CPUs have a sort of "time limit" on how long they'll remain boosted. Generally speaking, they will remain boosted for the duration of the average benchmark. These CPUs aren't powerful enough to be thermal throttling.
EDIT:
For the record, I agree the results aren't impressive and Intel is long overdue for a new architecture, but considering this is basically just a die shrink, the results are pretty good.

[M@GOid]

Not sure about this CPU in particular, but a lot of Intel's CPUs have a sort of "time limit" on how long they'll remain boosted. Generally speaking, they will remain boosted for the duration of the average benchmark. These CPUs aren't powerful enough to be thermal throttling.

There is a lot of benchmarks at Notebookcheck.net that disagree with your last remark. But is not exactly Intel's fault, but manufacturers releasing stupid Macbook Air clones that also mimic the poor Macbook's thermal performance.