• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Does Intel offers ROCm-like CUDA replacement that can be used out of the box?

Joined
Aug 29, 2024
Messages
57 (0.45/day)
Location
Shanghai
System Name UESTC_404
Processor AMD R9 7950x3d
Motherboard Gigabyte B650E Aorus Elite X Ax Ice
Cooling Deepcool AG620 digital
Memory Asgard ValkyrieII 6800c34->6400 16g*2
Video Card(s) Yeston 7900xt SAKURA SUGAR
Storage Zhitai TiPlus7100 1T
Case RANDOMLY PICKED TRASHBIN
Power Supply Thermalright TGFX850W
I've tried using ROCm on my 6750GRE12G in early 2024 for running ChatGLM3-6B, and it's totally a disaster:

The installing steps are annoying even for me, a CS graduate with some experience for every espect of CS.
I followed instructions provided by AMD, successfully installed ROCm 5.8(the latest version with torch support at that time) and torch(rocm ver), and then I found that nothing works fine, I need to setup a lot more (environment variables not mentioned by AMD's doc and not automatically setted by the install process). ROCm is complex to setup, especially on a rare GPU, I finally mixed the recommended settings of 6750xt from the document and the actual GPUID I found by myself. Then another problem comes out: I use 7700x which contains a iGPU, and ROCm needs to disable it at that time(maybe this issue is fixed later), if not doing so, everything'll break down without noticable error reports.


I noticed that Intel may cancel the Arc Pro series and release B580 24G later in 2025, so I'm wondering if Intel provides some software stacks like ROCm that I can use it OUT OF THE BOX? The performance is also not fine when using 6750GRE12G(I know 6000-series AMDGPU lacks AI performance), but it's acceptable, and 12GB vram makes it enough to run ChatGLM3-6B-int8——I don't care about performance&convenience a lot, so a 24GB version of B580 is attractive to me.

I already have three Arc GPUs: A380, A770-8G, B580(just give it away to my friend). If someone tell me that Intel's AI software stack is easier than ROCm to deploy, I may try it later
 
Joined
Jan 8, 2017
Messages
9,523 (3.26/day)
System Name Good enough
Processor AMD Ryzen R9 7900 - Alphacool Eisblock XPX Aurora Edge
Motherboard ASRock B650 Pro RS
Cooling 2x 360mm NexXxoS ST30 X-Flow, 1x 360mm NexXxoS ST30, 1x 240mm NexXxoS ST30
Memory 32GB - FURY Beast RGB 5600 Mhz
Video Card(s) Sapphire RX 7900 XT - Alphacool Eisblock Aurora
Storage 1x Kingston KC3000 1TB 1x Kingston A2000 1TB, 1x Samsung 850 EVO 250GB , 1x Samsung 860 EVO 500GB
Display(s) LG UltraGear 32GN650-B + 4K Samsung TV
Case Phanteks NV7
Power Supply GPS-750C
I don't know what you mean, Intel's equivalent to CUDA or ROCm is oneAPI but if you think you're going to have an easier time with it good luck.
 
Joined
Aug 20, 2007
Messages
21,579 (3.40/day)
System Name Pioneer
Processor Ryzen R9 9950X
Motherboard GIGABYTE Aorus Elite X670 AX
Cooling Noctua NH-D15 + A whole lotta Sunon and Corsair Maglev blower fans...
Memory 64GB (4x 16GB) G.Skill Flare X5 @ DDR5-6000 CL30
Video Card(s) XFX RX 7900 XTX Speedster Merc 310
Storage Intel 5800X Optane 800GB boot, +2x Crucial P5 Plus 2TB PCIe 4.0 NVMe SSDs
Display(s) 55" LG 55" B9 OLED 4K Display
Case Thermaltake Core X31
Audio Device(s) TOSLINK->Schiit Modi MB->Asgard 2 DAC Amp->AKG Pro K712 Headphones or HDMI->B9 OLED
Power Supply FSP Hydro Ti Pro 850W
Mouse Logitech G305 Lightspeed Wireless
Keyboard WASD Code v3 with Cherry Green keyswitches + PBT DS keycaps
Software Gentoo Linux x64 / Windows 11 Enterprise IoT 2024
yeah, it's not going to be easier with Intel. AMD is making strides but has a way to go. NVIDIA still has quite a lead in AI and ML.
 
Joined
May 10, 2023
Messages
409 (0.68/day)
Location
Brazil
Processor 5950x
Motherboard B550 ProArt
Cooling Fuma 2
Memory 4x32GB 3200MHz Corsair LPX
Video Card(s) 2x RTX 3090
Display(s) LG 42" C2 4k OLED
Power Supply XPG Core Reactor 850W
Software I use Arch btw
oneAPI is what you're looking for, as said above.
For some stuff it works better/easier than ROCm, but neither are close to CUDA still.

Also, are you trying it on windows or linux? That kind of stuff is often easier to get up and running in Linux.
 
Joined
Aug 29, 2024
Messages
57 (0.45/day)
Location
Shanghai
System Name UESTC_404
Processor AMD R9 7950x3d
Motherboard Gigabyte B650E Aorus Elite X Ax Ice
Cooling Deepcool AG620 digital
Memory Asgard ValkyrieII 6800c34->6400 16g*2
Video Card(s) Yeston 7900xt SAKURA SUGAR
Storage Zhitai TiPlus7100 1T
Case RANDOMLY PICKED TRASHBIN
Power Supply Thermalright TGFX850W
oneAPI is what you're looking for, as said above.
For some stuff it works better/easier than ROCm, but neither are close to CUDA still.

Also, are you trying it on windows or linux? That kind of stuff is often easier to get up and running in Linux.
I never tried it using windows/wsl, it's not possible for ROCm to run with pytorch on Windows though there's a windows version of ROCm.
All of these annoying steps are done in ubuntu 22.04
 

Solaris17

Super Dainty Moderator
Staff member
Joined
Aug 16, 2005
Messages
27,153 (3.84/day)
Location
Alabama
System Name RogueOne
Processor Xeon W9-3495x
Motherboard ASUS w790E Sage SE
Cooling SilverStone XE360-4677
Memory 128gb Gskill Zeta R5 DDR5 RDIMMs
Video Card(s) MSI SUPRIM Liquid X 4090
Storage 1x 2TB WD SN850X | 2x 8TB GAMMIX S70
Display(s) 49" Philips Evnia OLED (49M2C8900)
Case Thermaltake Core P3 Pro Snow
Audio Device(s) Moondrop S8's on schitt Gunnr
Power Supply Seasonic Prime TX-1600
Mouse Razer Viper mini signature edition (mercury white)
Keyboard Monsgeek M3 Lavender, Moondrop Luna lights
VR HMD Quest 3
Software Windows 11 Pro Workstation
Benchmark Scores I dont have time for that.
I never tried it using windows/wsl, it's not possible for ROCm to run with pytorch on Windows though there's a windows version of ROCm.
All of these annoying steps are done in ubuntu 22.04

If you want to try some models with some limited control over the settings you might take a look at Intels AIPlayground. Specifically the 2.0 preview.



alternative model guide:

 
Top