Intel's ARC Alchemist GPUs For Gaming PCs and Workstation Measured In Detail, Flagship Die Just As Big As NVIDIA's GA104
The pictures posted by Moore's Law is Dead in his latest leak regarding the Intel ARC Alchemist GPU lineup gave a clear view of the DG2-512 (SOC1) and DG2-128 (SOC2) dies. It looks like these dies give a real idea of the actual die size compared to the renders posted before. Despite that, Locuza was real close to the actual figures even with his previous numbers.
As per the details, it looks like the Intel DG2-512 flagship die is going to measure around 396.2mm2 which makes it larger than both NVIDIA's GA014 and AMD's Navi 22 chips. The DG2-512 powered GPUs are going to compete against NVIDIA's GeForce RTX 3070(Ti) and AMD Radeon RX 6700 XT.
NVIDIA packs in tensor cores and much bigger RT/FP32 cores in its chips while AMD RDNA 2 chips pack a single ray accelerator unit per CU and Infinity Cache. Intel will also have dedicated hardware onboard its Alchemist GPUs for Raytracing & AI-assisted super-sampling tech.
The same is the case with the DG2-128 die which measures around 147.7-154.9mm2 or slightly smaller than the render which gave out a die size of 156.4mm2. This is also much smaller than the 200mm2 TU117 die that the chip will be competing against. The GA107 die size isn't known yet but it is likely to be around 160-180mm2.
Compared to Intel's render image, the chip size appears a bit smaller, as it was the case on DG2-512.
I get 147.70mm² - 154.93mm² (real) vs. 156.40mm² (render). pic.twitter.com/jCfbYSm5hy
— Locuza (@Locuza_) December 11, 2021
As for the configurations for each Intel ARC Alchemist GPUs, they are mentioned below in full detail:
Intel Xe-HPG 512 EU GPUs Powered ARC Alchemist Gaming Graphics Lineup
The top Alchemist 512 EU (32 Xe Cores) variant is said to feature at least three configurations. The full die features 4096 cores, a 256-bit bus interface, and up to 16 GB GDDR6 memory featuring a 16 Gbps clock though 18 Gbps cannot be ruled out as per the rumor. Following are all the ARC 512 EU variants that we can expect based on the DG2-SOC1:
- 512 EU (4096 ALUs) / 16 GB @ Up To 18 Gbps / 256-bit / 225W (Desktops) & 120-150W (Laptops)
- 384 EU (3072 ALUs) / 12 GB @ Up To 16 Gbps / 192-bit / 150-200W (Desktops) & 80-120W (Laptops)
- 256 EU (2048 ALUs) / 8 GB @ Up To 16 Gbps / 128-bit / 60-80W (Laptops)
The Xe-HPG Alchemist 512 EU chip is suggested to feature clocks of around 2.2 - 2.5 GHz though we don't know if these are the average clocks or the maximum boost clocks. Let's assume that it's the max clock speed and in that case, the card would deliver up to 18.5 TFLOPs FP32 compute which is 40% more than the RX 6700 XT but 9% lower than the NVIDIA RTX 3070.
In terms of performance positioning, the top 512 EU variant is said to compete against the RTX 3070 / RTX 3070 Ti, the 384 EU variant is said to compete against the RTX 3060 / RTX 3060 Ti on desktops. On the laptop side, the 512 EU might be just as fast as the RTX 3080, 384 EU variant around RTX 3070 level and the 256 EU will end up against the RTX 3060.
Also, it is stated that Intel's initial TDP target was 225-250W but that's been upped to around 275W now. We can expect a 300W variant with dual 8-pin connectors too if Intel wants to push its clocks even further. In either case, we can expect the final model to rock an 8+6 pin connector config, The reference model is also going to look very much like the drone marketing shot Intel put out during the ARC branding reveal.
As for the launch, the SOC1 variants are planned for launch in Mid-February & while AIBs are waiting on final dies to test with their custom PCB designs, desktop variants may be launching first, followed by laptop and then workstation in late 2022.
Intel ARC ACM-G10 vs NVIDIA GA104 & AMD Navi 22 GPUs
|Graphics Card Name||Intel ARC A770||NVIDIA GeForce RTX 3070 Ti||AMD Radeon RX 6700 XT|
|GPU Name||ACM-G10||NVIDIA GA104||AMD Navi 22|
|Process Node||TSMC 6nm||Samsung 8nm||TSMC 7nm|
|Transistor Density||53.4M Transistors/mm2||44.4M Transistors/mm2||51.2M Transistors/mm2|
|FP32 Cores||32 Xe Cores||48 SM Units||40 Compute Units|
|Max Clock||2100 MHz||1770 MHz||2581 MHz|
|FP32 TFLOPs||17.2 TFLOPs||21.75 TFLOPs||13.21 TFLOPs|
|Memory Capacity||16 GB GDDR6||8 GB GDDR6X||12 GB GDDR6|
|Launch||Q3 2022||Q2 2021||Q1 2021|
Intel Xe-HPG 128 EU GPUs Powered ARC Alchemist Gaming Graphics Lineup
Then lastly, we have the Intel Xe-HPG Alchemist 128 EU (8 Xe Cores) parts. There are two configs that feature the full-fat SKU with 1024 cores, a 96-bit, and a 64-bit variant with 6 GB and 4 GB memory capacity, respectively.
The cut-down variant will come with 96 EUs or 768 cores and a 4 GB GDDR6 memory featured across a 64-bit bus interface. The chip will also feature a clock speed of around 2.2 - 2.5 GHz and have a sub 75W power consumption which means we will be looking at connector-less graphics cards for the entry-level segment.
Following are all the ARC 128 EU variants that we can expect based on the DG2-SOC2:
- 128 EU (1024 ALUs) / 6 GB @ Up To 16 Gbps / 96-bit / ~75W (Desktops)
- 128 EU (1024 ALUs) / 4 GB @ Up To 16 Gbps / 64-bit / 35-30W (Laptops)
- 96 EU (768 ALUs) / 4 GB @ Up To 16 Gbps / 64-bit / ~35W (Laptops)
Performance is expected to land between the GeForce GTX 1650 and GTX 1650 SUPER but with raytracing capabilities. One big advantage that Intel could have over AMD and Intel is that with these cards, they might enter the sub-$250 US market which has been completely abandoned in the current generation of cards. The GeForce RTX 3050 series only got a laptop release so far with RTX 3060 serving the entry-level Ampere segment at $329 US while the RX 6600 is expected to be AMD's entry-level solution for around $300 US.
This GPU will be very similar to the DG1 GPU-based discrete SDV board however Alchemist will have a more improved architecture design and definitely more performance uplift over the first-gen Xe GPU architecture. This lineup is definitely going to be aimed at the entry-level desktop discrete market based on the specifications.
Intel ARC ACM-G11 vs NVIDIA GA106 & AMD Navi 24 GPUs
|Graphics Card Name||NVIDIA GeForce RTX 2050||AMD Radeon RX 6500 XT||Intel ARC A380||NVIDIA GeForce RTX 3050|
|GPU Name||NVIDIA GA107||AMD Navi 24||ACM-G11||NVIDIA GA106|
|Process Node||Samsung 8nm||TSMC 6nm||TSMC 6nm||Samsung 8nm|
|FP32 Cores||16 SM Units||16 Compute Units||8 Xe Cores||24 SM Units|
|Memory Capacity||4 GB GDDR6||4 GB GDDR6||6 GB GDDR6||8 GB GDDR6|
|Launch||Q1 2022||Q2 2022||Q2 2022||Q1 2022|
Based on the timeline, the Xe-HPG Alchemist lineup will compete against NVIDIA's Ampere & AMD RDNA 2 GPUs since both companies aren't expected to launch their next-gen parts by the very end of 2022. NVIDIA and AMD are expected to release refreshes in early 2022 so that might give Intel's new lineup some competition but based on current performance expectations, the refreshed may not bring drastic performance differences to the lineup. The Xe-HPG ARC GPUs will also be coming to the mobility platform too and will be featured in Alder Lake-P notebooks. By 2023, Intel will have a proper high-end graphics card lineup going up against NVIDIA's Ada Lovelace and AMD RDNA 3 chips in the form of ARC Battlemage.
Intel Xe-HPG Based Discrete Alchemist GPU Configurations:
|GPU Variant||Graphics Card Variant||GPU Die||Execution Units||Shading Units (Cores)||Memory Capacity||Memory Speed||Memory Bus||TGP|
|Xe-HPG 512EU||ARC A780?||Arc ACM-G10||512 EUs||4096||Up To 32/16 GB GDDR6||18 / 16 / 14 Gbps||256-bit||~225W (Desktops)
|Xe-HPG 384EU||ARC A580?||Arc ACM-G10||384 EUs||3072||Up To 12 GB GDDR6||16 / 14 Gbps||192-bit||150-200W (Desktops)
|Xe-HPG 256EU||ARC A550?||Arc ACM-G10||256 EUs||2048||Up To 8 GB GDDR6||16 / 14 Gbps||128-bit||60-80W (Laptops)|
|Xe-HPG 128EU||ARC A380?||Arc ACM-G11||128 EUs||1024||Up To 6 GB GDDR6||16 / 14 Gbps||96-bit||~75W (Desktops)|
|Xe-HPG 128EU||ARC A350?||Arc ACM-G11||128 EUs||1024||Up To 4 GB GDDR6||16 / 14 Gbps||64-bit||35-50W (Laptops)|
|Xe-HPG 96EU||ARC A330?||Arc ACM-G11||86 EUs||768||Up To 4 GB GDDR6||16 / 14 Gbps||64-bit||~35W (Laptops)|