Intel HD 4000 Ivy Bridge Graphics On Linux

[phoronix]

Phoronix: Intel HD 4000 Ivy Bridge Graphics On Linux

Now having looked at the processor performance of the brand new Intel Core i7 3770K "Ivy Bridge" CPU, up now is our first look at the Intel HD 4000 "Gen7" graphics performance for the Ivy Bridge processors under Linux. Building upon what's turned into a huge success for Intel with their Sandy Bridge graphics with admirable performance and stable open-source Linux drivers, Ivy Bridge volleys Intel's Linux graphics capabilities into a whole new realm for those concerned about open-source graphics drivers.

http://www.phoronix.com/vr.php?view=17284

[Kano]

Usually the hd 4000 core is set to a turbo frequency of 1150 while hd 3000 is running at 1350 turbo by default. It should be easy to increase that with a good board, maybe the kernel module could be patched as well. Win OC tools allow on the fly changes so it should be possible I did not see any hardware sites with oc hd 4000 results, but it scaled pretty well. Using ddr3-1600 setting was also faster than ddr3-1333, so faster ram helps as well (tomshardware compared impact of ram speed btw).

[jakubo]

do AMD CPUs/APUs and motherboards finally make fully use of ddr3 RAM? id guess that if not it could be some grave impact on performance compared to the intel units. Im really eager to see trinity results

[Luke_Wolf]

do AMD CPUs/APUs and motherboards finally make fully use of ddr3 RAM? id guess that if not it could be some grave impact on performance compared to the intel units. Im really eager to see trinity results

Well the other thing to keep in mind about these results is what was posted in earlier threads that for the APUs the GPU isn't set to it's highest performance level by default so these tests are very flawed that way

[RussianNeuroMancer]

In "A8-3870K - Gallium3D" case, default power mode instead of high, right?

[elmo23x]

VDrift illustrated another nice improvement for the Intel Core i7 3770K. Ignore the Catalyst results here since the Catalyst driver renders the GLSL shaders correctly for VDrift where as all the Mesa/Gallium3D drivers handle them incorrectly, which is why the open-source drivers have much higher results than the proprietary AMD / NVIDIA drivers.

Isn't this a very stupid comment? So we should ignore the driver that DOES rendering correctly? Shouldn't it be better to eliminate from the test the ones that aren't able to render scenes correctly?

I will program a black driver, that will render the scenes in full black, but that will give you 1000+ fps in any resolution for any game!!! How great is that?

[AlbertP]

If you want to compare a faulty driver you should compare it with drivers that are faulty as well, comparing with Catalyst is like comparing apples and oranges. That's what he meant.

[elmo23x]

If you want to compare a faulty driver you should compare it with drivers that are faulty as well, comparing with Catalyst is like comparing apples and oranges. That's what he meant.

And what is the point of comparing two oranges that do not taste as oranges?

First meet the minimum quality, and then you can compare.

[Kivada]

Looks pretty underwhelming as it can't even beat the current generation APUs, whats gonna happen when AMD drops the A10 5800K w/ HD7660D APU?

do AMD CPUs/APUs and motherboards finally make fully use of ddr3 RAM? id guess that if not it could be some grave impact on performance compared to the intel units. Im really eager to see trinity results

AMD's APUs are the ones that actually get increased performance from faster DDR3, last I checked a near linear improvement all the way up to DDR3 2500. IIRC increasing the ram speed has almost no effect at all on Intel's 2000/3000 GPUs for some strange reason.

[Kivada]

In "A8-3870K - Gallium3D" case, default power mode instead of high, right?

This is Larabel, so it's always on low/default with no performance enhancements like S3TC...