Intel ARC GPUs, Xe Architecture for dGPUs [2018-2022]
- Thread starter DavidGraham
- Start date
-
- Tags
- intel
- Status
Dayman1225
Newcomer
Looks like Intel is moving to compiler based scheduling with Gen 12/Xe graphics.
https://gitlab.freedesktop.org/mesa/mesa/merge_requests/1884
https://gitlab.freedesktop.org/mesa/mesa/merge_requests/1884
Dayman1225
Newcomer
I know this is a Gen12/Xe thread for DGPUs but we have a leaked Gen12/Xe iGPU appearing on CompuBench
96 Execution Units running at 1.1GHz.
Not sure how it compares to anything else as a lot of results seem to have missing scores.
96 Execution Units running at 1.1GHz.
Not sure how it compares to anything else as a lot of results seem to have missing scores.
Frenetic Pony
Veteran
Rumor that Xe will have hardware raytracing support, initial information could come as early as CES: https://www.notebookcheck.net/New-I...GPUs-to-get-ray-tracing-support.437548.0.html
A1xLLcqAgt0qc2RyMz0y
Veteran
Intel Xe GPU Based Discrete Graphics Card Rumors: Progress Not Going Well
https://wccftech.com/intel-xe-gpu-discrete-graphics-cards-progress-not-going-well
https://wccftech.com/intel-xe-gpu-discrete-graphics-cards-progress-not-going-well
I never thought it was at all feasible that intel would be any competition for AMD and Nvidia GPUs. Even in their area of expertise its been almost complete stagnation for nearly a decade. If anything performance seems to regress almost every month as new hardware vulnerabilities are found. My intel CPU is some 15-20% slower now than when i bought it. Will probably get worse.
I can't also refrain myself to think "ah, what a simplistic and sensationalizing view over the complex engineering process of processor design some people have".
WIth today's Intel , given their manufacturing troubles , it's also hard to think Xe wasn't impacted to some degree by the many missed targets there
D
Deleted member 13524
Guest
Story of Raja's life.
Or at least per the last half decade or so.
Raja gets a lot of shit. Maybe I'm just too kind on the guy, but was/is he really that bad? He was hailed as a prodigy up until the release of Vega, and he's obviously in high demand or else he wouldn't have been hired by Intel "despite" Vega. I guess I don't have the insight to judge, but it feels like convenient scapegoating to me.
There's a lot I (we?) don't know about his time at AMD. He was obviously unhappy towards the end of his tenure though. With rumours of resource relocation, budget issues, manufacturing issues, etc. The public letter he wrote implored his former team to follow the roadmap they'd set up together as he stepped down. So how much of RDNA was ultimately influenced by Raja? Will we ever know?
With Xe I don't see how it'd be terribly surprising to anyone that a first release is comparatively timid Either. Though bombastic marketing (also echoing Vega...) doesn't exactly help. Developing a new architecture in record time from scratch isn't exactly a cake walk for any company. And so far this is a rumour. A rumour about a company which currently struggles to get anything below 14nm working as intended (Ice Lake clock speeds come to mind...). It doesn't seem farfetched to assume there just might be other aspects involved than architecture choices not turning out as efficient in practice as they seemed in theory.
Obviously I don't have any insider info and may be too out of the loop. But this negging on Raja just rubs me the wrong way.
DavidGraham
Veteran
Intel has only one choice, 14nm. Their 10nm is not yet reliable to build large CPUs, let alone complex high frequency GPUs. So their efficiencey will suffer greatly at 14nm compared to solutions from NVIDIA and AMD.
On the last earnings Intel CEO commented on their 7nm process debuting with a "data center GPU" in Q4 2021. They got plenty of time to fix 10nm to yield GPUs.
// As for Raja, he was the face of both Polaris and Vega marketing disasters. Polaris hyped with 2.8x perf/W anyone?
// As for Raja, he was the face of both Polaris and Vega marketing disasters. Polaris hyped with 2.8x perf/W anyone?
Frenetic Pony
Veteran
Based on consumer sales of Vega vs Navi he definitely shouldn't be in charge of PR, at least that much is clear (Navi has already done much better sans him). Don't know how reliable any of these other rumors are though.
See that's what I'm getting from all this. A bunch of unsubstantiated rumours and a record of being a less than stellar PR frontman who had to eat crow. How that warrants a ride on the internet hobbling wheel for the rest of his professional life is beyond me.
D
Deleted member 13524
Guest
Raja is the head of Intel's Core and Visual Computing Group (i.e. Intel's GPUs), just like he was head of Radeon Technology Group (i.e. AMD's GPUs), so him showing face as representative of those groups is just part of his job.
DavidGraham
Veteran
2022 is too late for anything in my opinion.
The disconnect between the hype and the reality was indeed striking. This was explained to me by someone who worked at AMD in the Radeon Technology Group at the time as a failure from engineers to communicate the actual performance and efficiency of the designs, and the gap that existed between the targets and the actual designs. This failure to communicate was ascribed to the pressure under which engineers were put to meet objectives they knew to be unrealistic given the limited resources they were given.
The degree to which Raja may or may not have been responsible for this, however, I do not know. The fact that (to my knowledge) no quantitative claims were made about Navi prior to its launch suggests that at least some lessons were learned.
- Status
