DAVID_RANDOM Telegram 455
David's random thoughts
llama.cpp用上SYCL之后终于展现出XMX的优势,B580的prefill性能赶上了规格比它大50%的7800 XT,不过decode的效率还是稍微低了点,带宽差距不应该这么大。 当然目前SYCL版本的flash attention kernel看起来性能还是不太好,性能会减半,目测有不少优化空间。提高batch size性能也会立刻锐减。
B580在Win下用最新驱动运行llama.cpp的Vulkan版本会获得比较正常的文本生成性能,与7800XT对比性能与显存带宽大致成比例,后者8B q8_0 >50 t/s。但是依然没有cooperative matrix所以prompt processing性能稀烂(vkpeak的matrix测试也跑不了)

SYCL则是全盘比Linux还差,flash attention也没什么好实现



tgoop.com/david_random/455
Create:
Last Update:

B580在Win下用最新驱动运行llama.cpp的Vulkan版本会获得比较正常的文本生成性能,与7800XT对比性能与显存带宽大致成比例,后者8B q8_0 >50 t/s。但是依然没有cooperative matrix所以prompt processing性能稀烂(vkpeak的matrix测试也跑不了)

SYCL则是全盘比Linux还差,flash attention也没什么好实现

BY David's random thoughts





Share with your friend now:
tgoop.com/david_random/455

View MORE
Open in Telegram


Telegram News

Date: |

Just at this time, Bitcoin and the broader crypto market have dropped to new 2022 lows. The Bitcoin price has tanked 10 percent dropping to $20,000. On the other hand, the altcoin space is witnessing even more brutal correction. Bitcoin has dropped nearly 60 percent year-to-date and more than 70 percent since its all-time high in November 2021. On Tuesday, some local media outlets included Sing Tao Daily cited sources as saying the Hong Kong government was considering restricting access to Telegram. Privacy Commissioner for Personal Data Ada Chung told to the Legislative Council on Monday that government officials, police and lawmakers remain the targets of “doxxing” despite a privacy law amendment last year that criminalised the malicious disclosure of personal information. Telegram Channels requirements & features Add the logo from your device. Adjust the visible area of your image. Congratulations! Now your Telegram channel has a face Click “Save”.! Hashtags are a fast way to find the correct information on social media. To put your content out there, be sure to add hashtags to each post. We have two intelligent tips to give you:
from us


Telegram David's random thoughts
FROM American