Thursday, March 31st 2022
Intel Arc DG2-512 Built on TSMC 6nm, Has More Transistors than GA104 and Navi 22
Some interesting technical specifications of the elusive two GPUs behind the Intel Arc "Alchemist" series surfaced. The larger DG2-512 silicon in particular, which forms the base for the Arc 5 and Arc 7 series, is interesting, in that it is larger in every way than the performance-segment ASICs from both NVIDIA and AMD. The table below compares the physical specs of the DG2-512, with the NVIDIA GA104, and the AMD Navi 22. This segment of GPUs has fairly powerful use-cases, including native 1440p gameplay, or playing at 4K with a performance enhancement—something Intel has, in the form of the XeSS.
The DG2-512 is built on the 6 nm TSMC N6 foundry node, the most advanced node among the three GPUs in this class. It has the highest transistor density of 53.4 mTr/mm², and the largest die-area of 406 mm², and the highest transistor-count of 21.7 billion. The Xe-HPG graphics architecture is designed for full DirectX 12 Ultimate feature support, and the DG2-512 dedicated hardware for ray tracing, as well as AI acceleration. The Arc A770M is the fastest product based on this silicon, however, it is a mobile GPU with aggressive power-management characteristic to the form-factor it serves. Here, the DG2-512 has an FP32 throughput of 13.5 TFLOPs, compared to 13.2 TFLOPs of the Navi 22 on the Radeon RX 6700 XT desktop graphics card, and the 21.7 TFLOPs of the GA104 that's maxed out on the GeForce RTX 3070 Ti desktop graphics card.
Sources:
Hardware Unboxed (YouTube), VideoCardz
The DG2-512 is built on the 6 nm TSMC N6 foundry node, the most advanced node among the three GPUs in this class. It has the highest transistor density of 53.4 mTr/mm², and the largest die-area of 406 mm², and the highest transistor-count of 21.7 billion. The Xe-HPG graphics architecture is designed for full DirectX 12 Ultimate feature support, and the DG2-512 dedicated hardware for ray tracing, as well as AI acceleration. The Arc A770M is the fastest product based on this silicon, however, it is a mobile GPU with aggressive power-management characteristic to the form-factor it serves. Here, the DG2-512 has an FP32 throughput of 13.5 TFLOPs, compared to 13.2 TFLOPs of the Navi 22 on the Radeon RX 6700 XT desktop graphics card, and the 21.7 TFLOPs of the GA104 that's maxed out on the GeForce RTX 3070 Ti desktop graphics card.
27 Comments on Intel Arc DG2-512 Built on TSMC 6nm, Has More Transistors than GA104 and Navi 22
Why would that bum you out? You want to have naive optimism in the face of all available information go ahead, just dont be surprised when everyone around you says "I told you so" when ARC finally comes out and has drivers that would make late 2000s AMD blush.
Obviously there might be some driver glitches that is inevitable for a first release. But saying a driver issue now is the cause why the product has been postponed or delayed? It's like saying, next year it will be raining and when it happens, brag to everyone that you were right.
Im all for hating on Intel but they do deserve a vote of confidence here, the fact that they have not released something half cooked means a lot imo because they could have just to milk the market
videocardz.com/newz/intel-flagship-arc-gpu-features-21-7-billion-transistors-more-than-navi-22-and-ga104
If the report is correct, then either they based the 4MB DG2-128 figure based on the mobile part which has 64bit bus (so 6MB in total L2 on the full 96bit DG2-128 design with 2MB disabled on the mobile 64bit part) or the L2 is not correlated with the memory bus (like RDNA2) but with the render slice (2MB L2/render slice )
EDIT: if the report is correct and the L2 is so small then probably its main reason should be the raytracer, not to help alleviate bandwidth needs regarding classic raster (although it should help also) so then it makes more sense to be correlated with render slices instead of memory controller/bus.
ark.intel.com/content/www/us/en/ark/products/130409/intel-core-i78809g-processor-with-radeon-rx-vega-m-gh-graphics-8m-cache-up-to-4-20-ghz.html
What makes any of you think these drivers won't continue to be shit for years? People are willing put-up with AMD's shitty new product launches because they know over the long term that the architecture is superior, but who among us is seriously going to wait years for ARC (an inferior architecture, or else it would have been competitive with ancient Vega 8 in Cezanne),to top off this pile?
Well, to describe it this way isn't truly fair...
Wait... Isn't it the same Raja working on it? lollllll
That beauty as you said was responsability of the intel we all know so there was no surprise as to how that venture went down. ARC gpus should be under a different umbrella that is probably from a safe distance of intel's rain.