Intel IvyBridge/Haswell/Broadwell/Skylake OpenGL & Vulkan Benchmarks On Linux 4.10 + Mesa 13.1

[indepe]

...and also it might be specific to my hardware.

And I suppose it is: I used

watch grep \"cpu MHz\" /proc/cpuinfo

to monitor CPU frequency, and although the displayed frequency for one of the CPUs will quickly more than double when I run my test, it will have little effect on the execution time. I suppose the dynamic changes of "powersave" and "ondemand" are not getting through to the hardware correctly. However setting different fixed frequencies with "userspace" has the expected effect, for example.

[indepe]

I guess you can't use the intel_pstate frequency governor with the AMD chip, nevertheless I feel like pointing out that on my machine (with kernel 4.8) the "acpi-cpufreq ondemand" appears to be about 1.6x slower for pure CPU code, compared to "intel_pstate powersave".

That's for a test that runs more than 10 secs, and remains the same if the test is run repeatedly.

"acpi-cpufreq performance" is more than 2x faster, and just slightly slower than "intel_pstate performance", in my test.

("intel_pstate performance" compared to "intel_pstate powersave" is about 1.5x faster.)

The performance modes of each also have the least variation in execution times, so to me they seem best suited for performance comparisons. However I haven't tested this as throughly as would be necessary, and it may have changed for kernel 4.10 since 4.8, and also it might be specific to my hardware.

[LinuxID10T]

Right. Any idea btw why the i7-5775C / GT3e performs so poorly with Vulkan?

I'm going to guess it has something to do with the eDRAM. Otherwise Broadwell is pretty solid with Vulkan. At least for me. Have a HD 5300 which is again a GT2 part.

[Michael]

FWIW, the i7 5775C is a Broadwell GT3e, while the others are GT2 level parts.

Right. Any idea btw why the i7-5775C / GT3e performs so poorly with Vulkan?

[Jumbotron]

I long for the day the Zen APU arrives and perhaps Michael feeling embarrassed for Intel will stop these useless iGPU benchoffs.

Bugger off troll. None of Michael's work is useless here. Particularly since YOU are not doing this kind of work. You're just reading, not doing. For the majority of the world's PC users iGPUs are ALL they have. So that fact right there negates your useless trolling.

Secondly.....the majority of PC users do not play games above and beyond what comes on the OS or web based games the like you see on phones. So an iGPU is all they will ever need.

Thirdly, some of us like to build less expensive but still performant systems. iGPUs allow us to do that without the added expense of an expensive stand alone GPU. Thusly, the "useless iGPU benchoffs" are anything but.

Fourthly, these "useless iGPU benchoffs" show that it is not always the newest, shiniest object that Intel wants to market at an inflated price that is the most performant for a certain need. That, also, this fact alone makes these "useless iGPU benchoffs" anything but.

Once again....bugger off troll.

[Marc Driftmeyer]

I long for the day the Zen APU arrives and perhaps Michael feeling embarrassed for Intel will stop these useless iGPU benchoffs.

[Kayden]

FWIW, the i7 5775C is a Broadwell GT3e, while the others are GT2 level parts.

[hxfhjkl]

Following many comments in this forum, I was under the impression that AMD's integrated graphics were *much* better than intel's. These benchmarks show otherwise: performance seems to be rather similar (with the i7557C performing much better due to eRAM, AFAIU).

Is this correct, or is there any major caveat that I'm missing?

Anyway, looking forward to Ryzen.

My x4 860k paired with an rx 470, or r7 360 will only show any decent activity in games like Metro, Borderlands 2 etc. Most other games will get similar fps on both gpu's and are getting much, much worse performance than the same gpu's paired with something like i5 6600k. The memory controller seems to be an issue, since overcloking the northbridge by 300mhz nets much better results than actually overclocking the main frequency by 700mhz... Since the a10-7850 is the same cpu as the x4 860k, but with an integrated gpu, i'm pretty sure it's also getting wrecked in the cpu department, while the gpu is in idle mode half of the time. Even worse is that the cpu has to throttle down more compared to the x4 860k, because it's sharing the same current with the igpu.

[franglais125]

All the systems were using their maximum-supported stock frequency and maximum supported channels. For i7-5775C it shows it differently since it runs as root rather than 'phoronix' user. Unfortunately under Linux AFAIK there is no way to get the memory clock / vendor / DIMM topology except when running as root... PTS supports reading such through dmidecode but unfortunately only works when run as root sadly due to Linux and that even still sometimes the reported DIMM information through dmidecode is wrong.

Thanks a lot for the clarification!

[Michael]

Thanks for the benchmarks! I was waiting for something like this.

Michael did all systems get a similar setting for RAM?
- I did notice that some have 8GB and some 16GB of RAM, I don't know if the total amount can have an impact on benchmark results for the ones presented, 8GB sounds like it should be enough, i.e. not a bottleneck
- But it leaves me wondering if all systems get 2 channels of RAM. The i7577C is the only one shown as 2x(something) for RAM. All the others have single channel RAM?

All the systems were using their maximum-supported stock frequency and maximum supported channels. For i7-5775C it shows it differently since it runs as root rather than 'phoronix' user. Unfortunately under Linux AFAIK there is no way to get the memory clock / vendor / DIMM topology except when running as root... PTS supports reading such through dmidecode but unfortunately only works when run as root sadly due to Linux and that even still sometimes the reported DIMM information through dmidecode is wrong.