Intel nabbed a star opensource graphics developer: https://www.phoronix.com/news/Alyssa-Rosenzweig-Joins-Intel
She reverse-engineered Apple M1/M2 GPU's inner workings and made their support possible in Asahi Linux.
She reverse-engineered Apple M1/M2 GPU's inner workings and made their support possible in Asahi Linux.
another one:
Einsteiger Workstation-Lösung: Intel Arc Pro B50 im Test - Hardwareluxx
Einsteiger Workstation-Lösung: Intel Arc Pro B50 im Test.
www.hardwareluxx.de
Some more time for certain features and driver readiness
They are using DirectML Geekbench not Cuda/OpenVino RIPanother one:
Einsteiger Workstation-Lösung: Intel Arc Pro B50 im Test - Hardwareluxx
Einsteiger Workstation-Lösung: Intel Arc Pro B50 im Test.
Some more time for certain features and driver readiness
A real surprise - Intel Arc Pro B50 in the test and big workstation duel under 450 euros | igor´sLAB
The Arc Pro B50, officially launched today, positions itself directly against AMD’s Radeon Pro W7500 and NVIDIA’s RTX A1000 in the entry-level segment of professional workstations. The focus is on…
www.igorslab.de
- Aug 22, 2001
- 31,952
- 32,290
- 146
More rumors of the B770 https://videocardz.com/newz/intel-reportedly-preparing-packaging-for-upcoming-arc-b770-gpus
After reading the Intel comment about how they may as well be selling their cards wrapped in money, does it even make sense to release it?
After reading the Intel comment about how they may as well be selling their cards wrapped in money, does it even make sense to release it?
I personally think there are hungry buyers out there for Intel cards. Speaking for myself, if you are looking for LLM workhorse that just works out of the box, that's Nvidia and Intel only. I wasted two hours following instructions trying to get LM Studio to detect my 9060 XT but had no choice but to furiously give up. My time is worth more than that. If I had purchased it purely for LLM use, I would've thrown it right back at Lisa's face. Their driver support for GPGPU sucks. Still!
If Intel plays their cards right, they can sell the G31 with 32GB GDDR6 by default and sell it for $700 as an AI GPU that happens to run games with performance similar to $450 GPUs from nvidia and AMD.
For games, they missed another brief window of opportunity to get beta testers. 5070 stock is too ubiquitous.
But if it isn't cancelled all that's left to try is VRAM spam and/or cheap SRIOV.
But if it isn't cancelled all that's left to try is VRAM spam and/or cheap SRIOV.
They keep sabotaging themselves, if the rumor about cancelling G31 and then restarting development causing 6 month delay is true. Summer launch would have not only been received very well but also be able to price it higher.
The real issue with Intel is not just struggling technically but infighting and politics that cause them to self destruct.
Regarding marketshare: Threat Interactive made a good point in one of his videos. Nvidia's goal is basically moving the graphics pipeline to their proprietary technologies one by one, and the competitors(AMD and Intel) forever forced to play catchup, thus maintaining monopoly. He is right in that neither vendor will take any significant marketshare from Nvidia when all they do is copy what Nvidia is doing.
Perhaps Intel should veer away from RT/ML focus and go all raster or something. Or bring new AA tech that actually advances on MSAA in quality. I got Expedition 33 a while ago and while overall it is beautiful, I was quite distracted by visual noise. Many years ago noone really cared about anti-aliasing, because it was a small benefit in visuals which aren't noticed in any action. Now the older tech like MSAA which were discarded by many are seen as some sort of a savior, and we have trash like upscaling and blur filters(TSAA) while still causing massive perf decreases.
Note that when MSAA first came out many of us saw as a downgrade in terms of supersampling in image quality. Now AA has downgraded so much that MSAA is seen as really, really good! Intel should look into partnering with Matrox and use their Edge AA, which was very good image quality and performance wise.
The real issue with Intel is not just struggling technically but infighting and politics that cause them to self destruct.
Regarding marketshare: Threat Interactive made a good point in one of his videos. Nvidia's goal is basically moving the graphics pipeline to their proprietary technologies one by one, and the competitors(AMD and Intel) forever forced to play catchup, thus maintaining monopoly. He is right in that neither vendor will take any significant marketshare from Nvidia when all they do is copy what Nvidia is doing.
Perhaps Intel should veer away from RT/ML focus and go all raster or something. Or bring new AA tech that actually advances on MSAA in quality. I got Expedition 33 a while ago and while overall it is beautiful, I was quite distracted by visual noise. Many years ago noone really cared about anti-aliasing, because it was a small benefit in visuals which aren't noticed in any action. Now the older tech like MSAA which were discarded by many are seen as some sort of a savior, and we have trash like upscaling and blur filters(TSAA) while still causing massive perf decreases.
Note that when MSAA first came out many of us saw as a downgrade in terms of supersampling in image quality. Now AA has downgraded so much that MSAA is seen as really, really good! Intel should look into partnering with Matrox and use their Edge AA, which was very good image quality and performance wise.
Last edited:
Intel Arc Pro B-Series GPUs and Xeon 6 Shine in MLPerf Inference v5.1
MLPerf v5.1 benchmarks showcase Intel Xeon 6 and Arc Pro B-Series GPUs delivering powerful, low-latency AI inference for workstations and edge systems.
newsroom.intel.com
Bryo4321
Member
- Dec 5, 2024
- 67
- 127
- 66
What really? It took me 2 seconds to get my 9070xt to work using Vulkan with llm studio.
Regarding MSAA being dead, you can thank deferred rendering pipelines in basically every modern engine. Performant MSAA just doesn’t really work with modern pipelines the way they did when forward rendering was common. I think DLSS and FSR4 really look quite good while increasing performance. They’ve rendered taa obsolete as far as I’m concerned. (And assuming you have supported hardware)
Last edited:
Thunder 57
Diamond Member
- Aug 19, 2007
- 3,992
- 6,670
- 136
Supersampling came at such a cost it was never a real option except for slightly older games and the best hardware. MSAA was a great comprimise.
Ding ding ding! We have a winner.Isn't the advantage of deferred rendering supposed to be with lighting? We seemed to do just fine without it. Now we sacrifce MSAA for garbage FXAA and the like. SMAA seems decent enough but maybe my eyes are playing tricks on me.
MSAA is good enough. My point is Intel needs to stop following Nvidia if they want real success. Currently, it's not feasible. But if they have their R300 moment and get an actual top tier GPU, it can introduce new features that deviate from Nvidia. If they can do that constantly, they can dethrone them. You'll never displace a monopoly by copying what they do.
Also, Nvidia's current strategy is same as what Intel did with their CPUs and vector extensions. The whole SSE and AVX strategy was to throw AMD off.
Neither Nvidia, AMD, or even Intel is offering you true value on their GPUs. It's mediocre vs bad, vs worse.
Also, Nvidia's current strategy is same as what Intel did with their CPUs and vector extensions. The whole SSE and AVX strategy was to throw AMD off.
The problem with ML rendering is now you are required to have semi-decent performance on $2000 hardware, when it was supposed to be a boon for low end hardware. The blurriness and noise is seriously extremely distracting.
This is the same strategy that streaming services have been doing. They give you initially a very good service, and add things like tiers and fake premium that sucks money out of users when they get a monopoly position. As soon as Ray Tracing becomes semi-playable on <$500 cards, they're going to introduce another thing that's questionable benefit but requires new hardware to use.
Neither Nvidia, AMD, or even Intel is offering you true value on their GPUs. It's mediocre vs bad, vs worse.
Bryo4321
Member
- Dec 5, 2024
- 67
- 127
- 66
under the settings, go to runtimes, and make sure the vulkan engine is selected. Your gpu should be detected automatically, mine was. Maybe since the 9060 is newer, lm studio needed an update? I don’t really know.
That's the problem. Vulkan is unavailable and no GPU is detected.
Maybe 9060 XT 8GB is unsupported? I don't see why it should be. It has enough VRAM to let some layers be offloaded onto it.
Maybe 9060 XT 8GB is unsupported? I don't see why it should be. It has enough VRAM to let some layers be offloaded onto it.
VirtualLarry
No Lifer
- Aug 25, 2001
- 56,581
- 10,220
- 126
LET'S WAIT FOR INTEL TO BRING BACK REAL SCAN- LINE INTERLEAVE RENDERING.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 22K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Facebook
Twitter