Intel's open-source driver team has managed to improve the performance of Intel's iGPUs by 15% while increasing performance per watt by 43% at the same time (via Phoronix). The post was published by Francisco Jerez that leads the open-source driver team. Intel had previously promised to turn its attention to open-source drivers as well and it looks like its delivering on at least one deliverable.
New Linux patch yields 15% more performance and 43% higher efficiency in Ice Lake iGPU
In a single patch, the Intel developers were able to improve performance per watt by 43% (which is a very significant gain I might add) while increasing absolute performance by 15%. The patch was tested on an Intel Ice Lake iGPU but should be valid for pretty much all Intel iGPUs that support this p-state. That said, you might experience a different level of performance improvement based on the exact iGPU. Before we go any further, here is an extract of the relevant portion from the patch note:
"This is my second take on improving the energy efficiency of the
intel_pstate driver under IO-bound conditions. The problem and
approach to solve it are roughly the same as in my previous series [1]
at a high level:In IO-bound scenarios (by definition) the throughput of the system
doesn't improve with increasing CPU frequency beyond the threshold
value at which the IO device becomes the bottleneck, however with the
current governors (whether HWP is in use or not) the CPU frequency
tends to oscillate with the load, often with an amplitude far into the
turbo range, leading to severely reduced energy efficiency, which is
particularly problematic when a limited TDP budget is shared among a
number of cores running some multithreaded workload, or among a CPU
core and an integrated GPU.Improving the energy efficiency of the CPU improves the throughput of
the system in such TDP-limited conditions. See [4] for some
preliminary benchmark results from a Razer Blade Stealth 13 Late
2019/LY320 laptop with an Intel ICL processor and integrated graphics,
including throughput results that range up to a ~15% improvement and
performance-per-watt results up to a ~43% improvement (estimated via
RAPL). Particularly the throughput results may vary substantially
from one platform to another depending on the TDP budget and the
balance of load between CPU and GPU."
Intel is preparing their Tiger Lake Graphics for launch soon and this is something that will be as powerful as a PS4 and should turn on the heat for entry-level graphics. Linux support for TGL will be very essential if the company wants to establish a lasting footprint in the market.
What we know about Intel's upcoming TGL graphics so far:
The Intel TGL-U processors will have a massive iGPU with 96 EUs each. If the ratio of EUs to SPs (or whatever you want to call them) remains the same, you are looking at roughly 768 cores. Clocked at 1.2 GHz, these cores will be able to output 1.84 TFLOPs of compute. Interestingly, this is exactly the same level of graphics power that the original Sony Playstation 4 packed. Intel would be seriously missing out if someone fails to point this out in the marketing! Without any further ado, here is the money shot of the leak taken from Sisoft's database by yours truly:
Intel's Tiger Lake mobility processor is essentially the DG1 in a mobility form factor considering the specs and architecture are exactly the same. The only difference is of the power draw and depending on whether this is a 25W part or a 15W part would make a huge difference. That said, I am fairly confident this GPU is going to beat the crap out of NVIDIA's MX 150 and probably even the MX250. In fact, to that effect, I am hearing that NVIDIA will be slashing prices of its entry-level lineup and is preparing the MX350 to go up against TGL.
That said, I do believe that unless you are going for a serious dedicated GPU like the GTX 1650 at the very least, Intel's Tiger Lake graphics will make the value proposition of any sub-par discrete graphics (ahem MX 250) a very hard sell for manufacturers. OEMs will probably not want to get into the complicated process of implementing a seperate cooling and power solution for a discrete chip when roughly the same power level can be housed on the package with Intel's TGL CPUs. The only concerning thing right now as far as I am concerned is Intel's ability to meet demand - which could end up being the saving grace for NVIDIA's MX lineup.
I was one of the first journos on the planet to call that the Zen architecture from AMD is one to look out for and I will say this now: Intel's TGL graphics are going to change the face of entry-level graphics in the mobility industry forever.
