VideoCardz has discovered a new Intel Xe DG2 model in the 100.9126 GPU driver. There are two variants, one of which has 128 EUs and was spotted not too long ago on Geekbench with a relatively underwhelming score of 9311 points. The new entry with 512 EUs could be a game changer, though. PC Gamer compared it with AMD’s Radeon RX 6000 Series GPUs, stating that it could deliver close to the same amount of raw graphics shading performance as NAVI 21. That would be a bit behind NVIDIA Ampere’s performance, but it’s still an accomplishment.
Intel’s Xe DG2 will reportedly be made on a 6 nm node, which would imply higher clock speeds than its processor. If performance manages to get around 17 to 20 TFLOPS, the DG2 would be within reach of the Radeon RX 6900 XT.
Intel continues to focus its new discrete GPU line in the mobile market. The Intel Xe DG2 is expected to be aimed at gaming laptops, so desktop users will have to wait a bit longer. In the meantime, it is exciting to see it make such gains.
Recent News
Nice that they are scaling up. Will be interesting to see where they perform in real tests.
If we could just kill off GeekBench, 3DMark and AotS once and for all and only get relevant tests, that would be nice.
I’m all for it if they can get competitive. I just don’t see Intel doing that at this point
6nm node? Who’s manufacturing that? Is it just CPU architecture that intel can’t move off of 14nm with good yields?
They were outsourcing their GPUs to the usual suspects
A 6 nanometer process node is actually interesting. Seems everyone else has skipped six.
[QUOTE=”Terminus, post: 26888, member: 204″]
6nm node? Who’s manufacturing that? Is it just CPU architecture that intel can’t move off of 14nm with good yields?
[/QUOTE]
Rumor is TSMC
I have to also give their team credit for focusing on the mobile markets first instead of taking on AMD and NVIDIA on their established desktop turfs. The mobile market has a much larger level of diversity which could give them a greater potential foothold to serve as a launching for more projects down the road.
[QUOTE=”Peter_Brosdahl, post: 26909, member: 87″]
I have to also give their team credit for focusing on the mobile markets first instead of taking on AMD and NVIDIA on their established desktop turfs. The mobile market has a much larger level of diversity which could give them a greater potential foothold to serve as a launching for more projects down the road.
[/QUOTE]
I think it also plays to intels strength with OEMs – they aren’t concentrating on retail markets, but putting these into the hands of OEMs will get much larger penetration faster.
They still have to convince the consumer it’s decent enough to buy and is worth something over just using their own IGP, but that same strategy has worked with their IGP and let’s Intel still claim to be the largest GPU producer by volume.
[QUOTE=”Brian_B, post: 26920, member: 96″]
…larger penetration faster.
[/QUOTE]
[ATTACH type=”full”]738[/ATTACH]
[QUOTE=”Grimlakin, post: 26900, member: 215″]
A 6 nanometer process node is actually interesting. Seems everyone else has skipped six.
[/QUOTE]
The “numbers” for a process node have been pure marketing-metaphor for a long-ass time now
[MEDIA=youtube]uEMDkbF3hu0[/MEDIA]
A “6 nm” process is merely (at least, I think so?) Intel’s 7 nm, fixed and refined, and instead of slapping ++++++++ as on 14 nm+++++++, it is instead going away from the tainted “7 nm” name that is associated to things going all the way back to Cannon Lake with its legendary <0.1% yields.
[QUOTE=”Terminus, post: 26888, member: 204″]
6nm node? Who’s manufacturing that? Is it just CPU architecture that intel can’t move off of 14nm with good yields?
[/QUOTE]
TSMC
[URL]https://www.anandtech.com/show/14228/tsmc-reveals-6-nm-process-technology-7-nm-with-higher-transistor-density[/URL]
[QUOTE=”Bloax, post: 26968, member: 255″]
The “numbers” for a process node have been pure marketing-metaphor for a long-*** time now
[MEDIA=youtube]uEMDkbF3hu0[/MEDIA]
A “6 nm” process is merely (at least, I think so?) Intel’s 7 nm, fixed and refined, and instead of slapping ++++++++ as on 14 nm+++++++, it is instead going away from the tainted “7 nm” name that is associated to things going all the way back to Cannon Lake with its legendary <0.1% yields.
[/QUOTE]
Intel won't have their 7nm ready until at least 2022.