nV News Forums

 
 

nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   NVIDIA Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=14)
-   -   Performance regressions with 256.25 driver? (http://www.nvnews.net/vbulletin/showthread.php?t=151376)

davelaser 05-26-10 10:38 AM

Performance regressions with 256.25 driver?
 
Phoronix have published some interesting benchmarks here:

http://www.phoronix.com/scan.php?pag...256_slow&num=1

Now sometimes the phoronix benchmarks are a bit unreliable but these look like quite serious regressions. Presumably the driver development team perform regression tests with the new drivers? Any explanation for the drop in performance?

Arup 05-26-10 12:35 PM

Re: Performance regressions with 256.25 driver?
 
Well this driver is still in beta stages and there are lots of new additions to it so it will take some time to iron out these deficiencies.

weltall 05-26-10 12:53 PM

Re: Performance regressions with 256.25 driver?
 
personally i see no regressions... actually i get better performance: just like this one points out http://www.phoronix.com/forums/showp...42&postcount=6

Licaon 05-26-10 04:32 PM

Re: Performance regressions with 256.25 driver?
 
Quote:

Originally Posted by weltall (Post 2258449)
personally i see no regressions... actually i get better performance: just like this one points out http://www.phoronix.com/forums/showp...42&postcount=6

yep, it's better for me (nana2)(nana2)

Arup 05-26-10 08:26 PM

Re: Performance regressions with 256.25 driver?
 
Same here, both GTK Perf and glxgears yield better numbers with latest drivers.

Dizzle7677 05-27-10 12:43 AM

Re: Performance regressions with 256.25 driver?
 
Keep in mind it's Phoronix and they need the Ad dollars.

mooninite 05-27-10 12:48 AM

Re: Performance regressions with 256.25 driver?
 
Quote:

Originally Posted by Dizzle7677 (Post 2258826)
Keep in mind it's Phoronix and they need the Ad dollars.

Actually it boils down to Michael's ignorance about benchmarking. He may have good intentions, but he needs to step back and re-evaluate how he benchmarks. He needs to involve others in benchmarking tasks, especially with regards to the kernel, before he posts them on his site so that they are run properly and the article is more accurate.

olifre 05-27-10 06:13 PM

Re: Performance regressions with 256.25 driver?
 
I am a regular user of compiz (in fact, I am living without Desktop Environment and use compiz standalone).
With the new driver, window-switching got a LOT faster! Finally, everything in compiz is completely fluent, if the card has not just gone to powersave-mode.

THANKS NVIDIA :)

wiresquire 05-27-10 06:22 PM

Re: Performance regressions with 256.25 driver?
 
Quote:

Originally Posted by mooninite (Post 2258830)
Actually it boils down to Michael's ignorance about benchmarking. He may have good intentions, but he needs to step back and re-evaluate how he benchmarks. He needs to involve others in benchmarking tasks, especially with regards to the kernel, before he posts them on his site so that they are run properly and the article is more accurate.

And what specifically do you see wrong with his methodology in this article/benchmark?

ws

Sean_W 05-27-10 06:31 PM

Re: Performance regressions with 256.25 driver?
 
Quote:

Originally Posted by Dizzle7677 (Post 2258826)
Keep in mind it's Phoronix and they need the Ad dollars.

Yep, it's very Ubuntu/GNOME biased. The memory benchmarks are wrong because anyone with an small amount of experience with Linux knows that using top and system monitor is for noobs and inaccurate.

I'd like to see the benchmarks not done on a mobile GPU for starters.

mooninite 05-27-10 06:47 PM

Re: Performance regressions with 256.25 driver?
 
Quote:

Originally Posted by wiresquire (Post 2259210)
And what specifically do you see wrong with his methodology in this article/benchmark?

ws

Example: his filesystem benchmark "showdown" between ext3/4, xfs, etc. was done in such a way that some file systems were not tuned properly or the benchmark itself was not performed correctly.

It's more than that one example, so please don't think I'm ratting out his benches over one article.

wiresquire 05-27-10 07:06 PM

Re: Performance regressions with 256.25 driver?
 
What has that got to do with THIS benchmark on an nvidia driver and nvidia card?

Do you see anything specific to this article/benchmark? If so, point it out. If not.....

ws


All times are GMT -5. The time now is 10:38 AM.

Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.