Thanks for the feedback.
I'll provide the PCI ID as soon as the hardware arrives in early February so we can test the folding stability.
I'm also looking forward to seeing those official developer notes to confirm exactly how the Xe Vector Engine handles FP64 in hardware.
Whitelist Request: Intel Panther Lake Xe3 iGPU (Arc B390)
Moderators: Site Moderators, FAHC Science Team
Re: Whitelist Request: Intel Panther Lake Xe3 iGPU (Arc B390)
[/quote]
FP8 is so outdated. Everyone knows FP4 is most important
[/quote]
Haha, very true! While the AI PR teams are racing to 0-bit precision, we're still over here demanding those solid 64-bit floating points.
I'm eager to see if the Xe3 architecture finally breaks the 'iGPU curse' with its native FP64 support. I'll be posting the full PCI ID and some power-efficiency benchmarks as soon as the mini PC lands in February.
Let's see if we can get these Series 3 chips to actually move the needle on modern Work Units!
FP8 is so outdated. Everyone knows FP4 is most important
[/quote]
Haha, very true! While the AI PR teams are racing to 0-bit precision, we're still over here demanding those solid 64-bit floating points.
I'm eager to see if the Xe3 architecture finally breaks the 'iGPU curse' with its native FP64 support. I'll be posting the full PCI ID and some power-efficiency benchmarks as soon as the mini PC lands in February.
Let's see if we can get these Series 3 chips to actually move the needle on modern Work Units!
-
Joe_H
- Site Admin
- Posts: 8319
- Joined: Tue Apr 21, 2009 4:41 pm
- Hardware configuration: Mac Studio M1 Max 32 GB smp6
Mac Hack i7-7700K 48 GB smp4 - Location: W. MA
Re: Whitelist Request: Intel Panther Lake Xe3 iGPU (Arc B390)
Earlier Gen Intel iGPUs did have FP64 support and then for some unknown reason they stopped including it with the Gen12 Intel Xe iGPUs for Tiger Lake and later processors for mobile usage. That also removed FP64 hardware support from the Rocket Lake desktop processors.
The higher end Gen9 and Gen11 iGPUs were actually decent performers on smaller GPU projects when they came out, but the low end ones weren't. No idea what happened to Gen10, either it didn't get off the drawing board, or Intel skipped the number for some reason. Gen8 and earlier were not recommended by Intel for use as too weak.
The higher end Gen9 and Gen11 iGPUs were actually decent performers on smaller GPU projects when they came out, but the low end ones weren't. No idea what happened to Gen10, either it didn't get off the drawing board, or Intel skipped the number for some reason. Gen8 and earlier were not recommended by Intel for use as too weak.