Performance of llama.cpp with Vulkan #10879
Replies: 171 comments 274 replies
-
|
AMD FirePro W8100
|
Beta Was this translation helpful? Give feedback.
-
|
With the latest updates:
|
Beta Was this translation helpful? Give feedback.
-
|
With FA:
|
Beta Was this translation helpful? Give feedback.
-
|
AMD RX 470
|
Beta Was this translation helpful? Give feedback.
-
|
With the latest updates:
|
Beta Was this translation helpful? Give feedback.
-
|
With FA:
|
Beta Was this translation helpful? Give feedback.
-
|
i got the mining edition 8 gb ran slightly better
build: e288693 (6242) |
Beta Was this translation helpful? Give feedback.
-
|
ubuntu 24.04, vulkan and cuda installed from official APT packages.
build: 4da69d1 (4351) vs CUDA on the same build/setup
build: 4da69d1 (4351) |
Beta Was this translation helpful? Give feedback.
-
|
Macbook Air M2 on Asahi Linux ggml_vulkan: Found 1 Vulkan devices:
|
Beta Was this translation helpful? Give feedback.
-
|
For the record I think this is slow on the HoneyKrisp side rather than llama.cpp |
Beta Was this translation helpful? Give feedback.
-
|
Can you share how you got vulkan to build on Asahi? I can't seem to get cmake to notice it. |
Beta Was this translation helpful? Give feedback.
-
|
Spoke too soon, got it working! |
Beta Was this translation helpful? Give feedback.
-
|
Gentoo Linux on ROG Ally (2023) Ryzen Z1 Extreme ggml_vulkan: Found 1 Vulkan devices:
|
Beta Was this translation helpful? Give feedback.
-
|
ggml_vulkan: Found 4 Vulkan devices:
|
Beta Was this translation helpful? Give feedback.
-
|
Cool setup! Could you also post the result of 1, 2 and 3 7900 XTX GPUs? You can use only the first GPU with |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
For this multi GPU case getting Vulkan to support #6017 pipeline parallelism might help improve the prompt processing speed. |
Beta Was this translation helpful? Give feedback.
-
|
@netrunnereve I updated the commit id in all my result. |
Beta Was this translation helpful? Give feedback.
-
|
build: 0d52a69 (4439) NVIDIA GeForce RTX 3090 (NVIDIA)
AMD Radeon RX 6800 XT (RADV NAVI21) (radv)
AMD Radeon (TM) Pro VII (RADV VEGA20) (radv)
Intel(R) Arc(tm) A770 Graphics (DG2) (Intel open-source Mesa driver)
|
Beta Was this translation helpful? Give feedback.
-
|
@netrunnereve Some of the tg results here are a little low, I think they might be debug builds. The cmake step (at least on Linux) might require |
Beta Was this translation helpful? Give feedback.
-
|
I've added My release numbers for the RX 470 are basically identical to the ones I posted earlier without the flag.
|
Beta Was this translation helpful? Give feedback.
-
|
Maybe not in your case, but some other results are suspiciously low in tg (for example the RTX 3080) |
Beta Was this translation helpful? Give feedback.
-
|
Build: 8d59d91 (4450)
Lack of proper Xe coopmat support in the ANV driver is a setback honestly.
edit: retested both with the default batch size. |
Beta Was this translation helpful? Give feedback.
-
I put my A770 into a Windows PC and gave Intel GPA and vtune a shot: GPA just crashes most of the time, I couldn't get it to trace anything useful. vtune works, but does not support Vulkan. It just shows some high-level metrics in that case, not really useful sadly. |
Beta Was this translation helpful? Give feedback.
-
I did build it with cmake with build type Release. |
Beta Was this translation helpful? Give feedback.
-
|
In that case it's something else, cause it should be performing similarly to my A770. I suspect the mesa version, there was something in newer mesa versions that slowed down tg on Intel. |
Beta Was this translation helpful? Give feedback.
-
|
A750 has 448 CUs, A770 has 512 CUs I think. Personally, I am not worried about tg. I am worried about pp here. The gemm batch quickly saturates my GPU. |
Beta Was this translation helpful? Give feedback.
-
|
@0cc4m https://gitlab.freedesktop.org/mesa/mesa/-/issues/12585 |
Beta Was this translation helpful? Give feedback.
-
|
Here's something exotic: An AMD FirePro S10000 dual GPU from 2012 with 2x 3GB GDDR5. build: 914a82d (4452)
|
Beta Was this translation helpful? Give feedback.
-
|
Very interesting, and looks like it's pretty close to the W8100 in tg despite being a dual GPU card. Your backend scales pretty well with layer splitting which is why I find it worthwhile to run my RX470 and W8100 together (I end up getting results that are close to the average of both cards).
|
Beta Was this translation helpful? Give feedback.
-
|
Latest arch with For the sake of consistency I run every bit in a script and also build every target from scratch (for some reason kill -STOP -1
timeout 240s $COMMAND
kill -CONT -1
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Iris(R) Xe Graphics (TGL GT2) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | warp size: 32 | matrix cores: none
build: ff3fcab (4459)
This bit seems to underutilise both GPU and CPU in real conditions based on
|
Beta Was this translation helpful? Give feedback.
-
|
Unless you reduce the number of GPU layers, threads and openblas/non-openblas is not gonna make any difference. Try it with ngl 0, then only prompt processing is accelerated using Vulkan, the rest runs on CPU. This is often a good setting for integrated GPUs. |
Beta Was this translation helpful? Give feedback.
-
|
That's something I didn't think about, with
build: ba8a1f9 (4460)
|
Beta Was this translation helpful? Give feedback.
-
|
It seems latest patches has improved the results a bit: ggml_vulkan: Found 1 Vulkan devices:
|
Beta Was this translation helpful? Give feedback.
-
|
A few months later and I get:
build: f3a4b16 (5568) I run it on Linux (Arch with lama.cpp-vulkan-git package compiled by GCC 15). From my tests, only Vulkan backend ( I'm curious why I cannot go over 6 t/s. Is this an issue with the newer llama.cpp version or with my OS configuration? |
Beta Was this translation helpful? Give feedback.
-
|
Intel ARC A770 on Windows:
build: ba8a1f9 (4460) |
Beta Was this translation helpful? Give feedback.
-
Single GPU VulkanRadeon Instinct MI25 ggml_vulkan: 0 = AMD Radeon Instinct MI25 (RADV VEGA10) (radv) | uma: 0 | fp16: 1 | warp size: 64 | matrix cores: none
build: 2739a71 (4461) Radeon PRO VII ggml_vulkan: 0 = AMD Radeon Pro VII (RADV VEGA20) (radv) | uma: 0 | fp16: 1 | warp size: 64 | matrix cores: none
build: 2739a71 (4461) Multi GPU Vulkanggml_vulkan: 0 = AMD Radeon Pro VII (RADV VEGA20) (radv) | uma: 0 | fp16: 1 | warp size: 64 | matrix cores: none
build: 2739a71 (4461) ggml_vulkan: 0 = AMD Radeon Pro VII (RADV VEGA20) (radv) | uma: 0 | fp16: 1 | warp size: 64 | matrix cores: none
build: 2739a71 (4461) Single GPU RocmDevice 0: AMD Radeon Instinct MI25, compute capability 9.0, VMM: no
build: 2739a71 (4461) Device 0: AMD Radeon Pro VII, compute capability 9.0, VMM: no
build: 2739a71 (4461) Multi GPU RocmDevice 0: AMD Radeon Pro VII, compute capability 9.0, VMM: no
build: 2739a71 (4461) Layer split
build: 2739a71 (4461) Row split
build: 2739a71 (4461) Single GPU speed is decent, but multi GPU trails Rocm by a wide margin, especially with large models due to the lack of row split. |
Beta Was this translation helpful? Give feedback.
-
|
What is the power profile for this MI25? Mine is 110W but its running slower than yours on git from today. |
Beta Was this translation helpful? Give feedback.
-
|
Mine defaults to 220w. |
Beta Was this translation helpful? Give feedback.
-
|
AMD Radeon RX 5700 XT on Arch using mesa-git and setting a higher GPU power limit compared to the stock card.
I also think it could be interesting adding the flash attention results to the scoreboard (even if the support for it still isn't as mature as CUDA's).
|
Beta Was this translation helpful? Give feedback.
-
|
There is no Vulkan flash attention support (except with coopmat2 on very new nvidia drivers). What you're measuring here is a CPU fallback. |
Beta Was this translation helpful? Give feedback.
-
|
I see, I was sure about the CPU fallback but didn't know there was no flash attention support at all. |
Beta Was this translation helpful? Give feedback.
-
|
I tried but there's nothing after 1 hrs , ok, might be 40 mins... Anyway I run the llama_cli for a sample eval...
Meanwhile OpenBLAS |
Beta Was this translation helpful? Give feedback.
-
|
Even at below 1t/s |
Beta Was this translation helpful? Give feedback.
-
|
Truth is ... (0.79 tokens per second), 3788.63 ms / 3 tokens So it's not even...it just slower... |
Beta Was this translation helpful? Give feedback.
-
|
Intel Arc B570 ggml_vulkan: Found 1 Vulkan devices:
build: 7f76692 (6527) |
Beta Was this translation helpful? Give feedback.
-
|
Ryzen Al Max+ 395 (128GB memory)
build: 7f76692 |
Beta Was this translation helpful? Give feedback.
-
|
xeon 2699v3 default 190W
updated to 228W
build: 5bb4a3e (6528) SYCL (llama.cpp)
build: 1eeb523 (6529) SYCL (ipex-llm)
build: d2c8ed1 (1) |
Beta Was this translation helpful? Give feedback.
-
|
We need this performance in Vulkan 😭 
|
Beta Was this translation helpful? Give feedback.
-
|
It would be nice. The Intel Vulkan performance started out really bad, and only through a lot of trial and error have I gotten it to where it is now. There is basically no information on how to optimize and much of the optimization would need to happen in the driver. |
Beta Was this translation helpful? Give feedback.
-
Thank you very much! If you need any tests (on 4xA770, windows/ubuntu) - just ping me. |
Beta Was this translation helpful? Give feedback.
-
|
xeon 2699v3
build: 1eeb523 (6529) |
Beta Was this translation helpful? Give feedback.
-
|
~/AIAgents/llama.cpp/build/bin$ ./llama-bench -m ../../models/llama-2-7b.Q4_0.gguf -ngl 100 -fa 0,1
build: 3b15924 (6403) |
Beta Was this translation helpful? Give feedback.
-
|
AMD Ryzen AI 9 HX 370 w/ Radeon 890M ggml_vulkan: Found 1 Vulkan devices:
build: 7f76692 (6527) |
Beta Was this translation helpful? Give feedback.
-
|
I get a slight increase in performance with a newer build: ggml_vulkan: Found 1 Vulkan devices:
build: 85e7227 (6553) |
Beta Was this translation helpful? Give feedback.
-
|
AMD Radeon RX 5700 XT Old performance data (January 2025): #10879 (comment) New performance data: ggml_vulkan: Found 1 Vulkan devices:
build: 0889589 (6559) |
Beta Was this translation helpful? Give feedback.
-
|
I want to add that, differently from my first submission of data, this new performance data has been obtained with the stock power limits (the first submission had power limits unlocked). If I unlock the power limits here too the PP with fa 0 goes up to 520 t/s. |
Beta Was this translation helpful? Give feedback.
-
|
Does Navi 10 support int DP? I forgot... |
Beta Was this translation helpful? Give feedback.
-
|
It does not |
Beta Was this translation helpful? Give feedback.
-
|
Thanks... |
Beta Was this translation helpful? Give feedback.
-
|
AMD Radeon RX 7800 XT Old performance data (May 2025): #10879 (comment) New performance data (I still include the AMDVLK driver, despite it being discontinued, as it still has a massive lead in coopmat performance): COOPMAT RADV
ggml_vulkan: Found 1 Vulkan devices:
build: 0889589 (6559) COOPMAT AMDVLK
ggml_vulkan: Found 1 Vulkan devices:
build: 0889589 (6559) RADV INT_DOT
ggml_vulkan: Found 1 Vulkan devices:
build: 0889589 (6559) AMDVLK INT_DOT
ggml_vulkan: Found 1 Vulkan devices:
build: 0889589 (6559) RADV SCALAR
ggml_vulkan: Found 1 Vulkan devices:
build: 0889589 (6559) AMDVLK SCALAR
ggml_vulkan: Found 1 Vulkan devices:
build: 0889589 (6559) |
Beta Was this translation helpful? Give feedback.
-
|
There's probably performance to be gained from optimizing the coopmat implementation, it should not be as close to integer dot as it is for you. |
Beta Was this translation helpful? Give feedback.
-
|
I'm quite sure the mesa driver still needs much more optimization for RDNA3, that's why I kept this issue open (https://gitlab.freedesktop.org/mesa/mesa/-/issues/13181). Now that AMDVLK is officially discontinued and with all the focus from AMD switching towards mesa, it's possible that this performance issue will finally be fixed. |
Beta Was this translation helpful? Give feedback.
-
|
Yeah, but even amdvlk is too close to int dot. Or maybe that's just RDNA3's coopmat hardware implementation not being that good. |
Beta Was this translation helpful? Give feedback.
-
|
For comparison, on Nvidia Ampere each step nearly doubles performance, from scalar to integer dot to coopmat. Something like 1000 t/s scalar to 2000 t/s integer dot to 3800 t/s coopmat1 to something like 4300 t/s coopmat2. |
Beta Was this translation helpful? Give feedback.
-
|
I didn't really know how much the performance improves on Nvidia with coopmat, that's some massive gains! |
Beta Was this translation helpful? Give feedback.
-
|
using build b6550, aur script returns unknown error on the build number ggml_vulkan: 0 = AMD Radeon RX 9070 XT (RADV GFX1201) (radv) | uma: 0 | fp16: 1 | bf16: 1 | warp size: 64 | shared memory: 65536 | int dot: 1 | matrix cores: KHR_coopmat
AMDVLK
|
Beta Was this translation helpful? Give feedback.
-
|
I don't know that is happening, but my RX 7900 XT is really slow:
The MI50 has a pp512 higher than the reference post.
EDIT: Using rocm-smi, the GPU of my RX 7900XT is jumping between 70 to 100%, then stuck with 100%. |
Beta Was this translation helpful? Give feedback.
-
Can you open radeontop or an equvalent during your llama bench and check the gtt use as well as the memory and shader clock speeds? |
Beta Was this translation helpful? Give feedback.
-
|
Driver Version: 580.82.09 RTX 4070 Ti Super
RTX A5000
build: 4ae88d0 (6570) |
Beta Was this translation helpful? Give feedback.
-
|
Driver Version: 580.82.09 Got much higher results with 4090 compared to the table in the OP. RTX 4090
RTX 3090
build: 4ae88d0 (6570) |
Beta Was this translation helpful? Give feedback.
-
|
Running on a Minisforum UM780XTX (AMD 7840HS / 780M) with 32 GiB DDR5-5600
build: 835b2b9 (6586) |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
Maybe it's due to 803dac2? |
Beta Was this translation helpful? Give feedback.
-
|
Please use git bisect to figure out the cause. |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
Please stop posting threads here, this is off-topic. Open an issue about it instead. |
Beta Was this translation helpful? Give feedback.
and others
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
This is similar to the Apple Silicon benchmark thread, but for Vulkan! Many improvements have been made to the Vulkan backend and I think it's good to consolidate and discuss our results here.
We'll be testing the Llama 2 7B model like the other thread to keep things consistent, and use Q4_0 as it's simple to compute and small enough to fit on a 4GB GPU. You can download it here.
Instructions
Either run the commands below or download one of our Vulkan releases. If you have multiple GPUs please run the test on a single GPU using
-sm none -mg YOUR_GPU_NUMBERunless the model is too big to fit in VRAM.Share your llama-bench results along with the git hash and Vulkan info string in the comments. Feel free to try other models and compare backends, but only valid runs will be placed on the scoreboard.
If multiple entries are posted for the same setup I'll prioritize newer commits with substantial Vulkan updates, otherwise I'll pick the one with the highest overall score at my discretion. Performance may vary depending on driver, operating system, board manufacturer, etc. even if the chip is the same. For integrated graphics note that the memory speed and number of channels will greatly affect your inference speed!
Vulkan Scoreboard for Llama 2 7B, Q4_0 (no FA)
Vulkan Scoreboard for Llama 2 7B, Q4_0 (with FA)
Beta Was this translation helpful? Give feedback.
All reactions