Ok, I got report for SubSpace, he was able to repair his HD6xxx installation and now able to run few tasks at once.
The reason was (just as we thought) bad drivers from AMD.
I'll try to translate his report:
"
1) installed Windows 7 + Cat 11.1 - all working OK.
2) installed _only_ video-driver from 11.2 - immediately driver restarts.
3) deleted 8.821 driver via Win7 "delete driver files" from device manager
4) re-installed Cat 11.1 video driver (8.812) - all working fine again.
"

So, AMD once more provided broken drivers in their driver update (it happened before, with Brook-based applications).
Looks like exactly this part of new Cat 11.2 suite causes problems, so better to either stay with Cat 11.1 or replace video-driver from 11.2 to 11.1 (8.821 to 8.812)

I saw new HD6990 driverpack on the AMD site.
http://support.amd.com/us/kbarticles/Pages/Catalyst114earlypreview.aspx
AMD Catalyst™ 11.4 early preview – “Mjölnir”
It claims;
· Delivers performance enhancements for the AMD Radeon™ HD 6000 Series & Windows 7.

They speak about"HD 6000 series"........mistake???
Or give it a try.....?

---

I saw new HD6990 driverpack on the AMD site.
http://support.amd.com/us/kbarticles/Pages/Catalyst114earlypreview.aspx
AMD Catalyst™ 11.4 early preview – “Mjölnir”
It claims;
· Delivers performance enhancements for the AMD Radeon™ HD 6000 Series & Windows 7.

They speak about"HD 6000 series"........mistake???
Or give it a try.....?

I've got it on my 5670 - does work but no major speed increases
It is a preview (read beta) driver so would be carefull with it

Edit: does seem to play nicer with my 465 though - times have dropped slightly for cuda tasks since I've updated to this beta driver, possibly resolved some resource conflicts???

I'm willing to give it a try. my 6970 has been a dog. I'll try this and the suggestions from the Lunatics site. It appears that the current drivers don't work well with the 6970 cards

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

I've got it on my 5670 - does work but no major speed increases
It is a preview (read beta) driver so would be carefull with it

Edit: does seem to play nicer with my 465 though - times have dropped slightly for cuda tasks since I've updated to this beta driver, possibly resolved some resource conflicts???

say what? how are ATI/AMD drivers improving the performance of your nVidia GPU?

---

say what? how are ATI/AMD drivers improving the performance of your nVidia GPU?

Was halfway through running a bunch of VHARs when switched over drivers, all the tasks were the same AR to 2dps ~1.18, before upgrading were running at 5 mins +/- 5 sec's per task (2 at a time) after the upgrade the times dropped to about 4 mins 50 +/- 5 sec's
Didn't say it was a big improvement, but somethings different with how the drivers are interacting as nothing else changed on my box

Edit: One of the first things I do when beta testing a new app @ Lunatics is to see how it affects the other apps on my host.
nVidia and ATI/AMD haven't historically been easy to get working together on the same host (or at all possible is some cases), even the slightest change can make a massive difference in compute times, for example I have to manually set the affinity of my cuda tasks to work from a seperate cpu core to my openCl tasks otherwise they both suffer in compute times.

I've got it on my 5670 - does work but no major speed increases
It is a preview (read beta) driver so would be carefull with it

Edit: does seem to play nicer with my 465 though - times have dropped slightly for cuda tasks since I've updated to this beta driver, possibly resolved some resource conflicts???

say what? how are ATI/AMD drivers improving the performance of your nVidia GPU?

I find heavily blanked OpenCL ATI Astropulse tasks influence and slow down Cuda computation on my GTX460, most noticeable on VHAR's (both of us have mixed Nvidia/ATI hosts)

Claggy

I find heavily blanked OpenCL ATI Astropulse tasks influence and slow down Cuda computation on my GTX460, most noticeable on VHAR's (both of us have mixed Nvidia/ATI hosts)

Claggy

ahh, i see. thanks for clearing that up guys. i'll have to keep this in mind, as i eventually plan on getting an nVidia CUDA GPU for Einstein@Home, and i'm undecided as to which rig it will go in - the rig w/out a discrete GPU in my master bedroom, or the H=home office rig that already has the 5870 GPU.

---

Hello,

the astropulse app is verry much faster than CPU ( 5,787.45 sec. )
But the multibeam app is only 4.5x faster than CPU app... is this normal or can i tune up this ?
I use a HD 5770 and the multibeam-units are running 40-70 minutes...

Greetz,
Chris

---

almost 24 hours now on the ati 11.4 drivers. No crashes no WU errors. Though I've noticed a major slow down in VLAR and AP WU's

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

I finally got around to getting the MB app on one of my machines at home with a 4870. I used the 11-2 release, upgrading from 9-6 or such, & everything seems to be going along fine at the moment. Tasks are validating & no errors in any of the pending tasks as of yet.

Seems to run typically 89-95% GPU usage according to CCC. With the temp staying low enough the fan only runs around 40% speed. So it is sufficiently quiet.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

I finally got around to getting the MB app on one of my machines at home with a 4870. I used the 11-2 release, upgrading from 9-6 or such, & everything seems to be going along fine at the moment. Tasks are validating & no errors in any of the pending tasks as of yet.

Seems to run typically 89-95% GPU usage according to CCC. With the temp staying low enough the fan only runs around 40% speed. So it is sufficiently quiet.

LoL, you mirroring my own setup then :D

I finally got around to getting the MB app on one of my machines at home with a 4870. I used the 11-2 release, upgrading from 9-6 or such, & everything seems to be going along fine at the moment. Tasks are validating & no errors in any of the pending tasks as of yet.

Seems to run typically 89-95% GPU usage according to CCC. With the temp staying low enough the fan only runs around 40% speed. So it is sufficiently quiet.

LoL, you mirroring my own setup then :D

Well I have had this setup for about 2 years now.... might be time to upgrade soon. My HTPC & notebook are both newer than my "main" computer now. :/

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

I have just swapped out a GTS 250 for a Radeon 4870

The GTS 250 did a task around every 19mins
The first task i finished on the 4870 took 60mins

Am i setup correct

i was expecting it to be about 30% faster than the GTS 250
not 3 times slower

---

I have just swapped out a GTS 250 for a Radeon 4870

The GTS 250 did a task around every 19mins
The first task i finished on the 4870 took 60mins

Am i setup correct

i was expecting it to be about 30% faster than the GTS 250
not 3 times slower

That sounds about reasonable - not to sure on the 4870 specs though

My GTX 465 will complete a mid-range AR task in around 17 mins (two at a time)
My 5670 will take around 45 mins to complete a similar AR task 1 at a time

That is correct, for now if you want best results, then you should stick with the Nvidia cards. For gaming purposes the ATI you chose can outperform that model Nvidia card.

---

Cheers

Im not sure tbh it looks like the shader performance on the GTS 250
smacks the 4870 up and down

My first result on the 4870
http://setiathome.berkeley.edu/result.php?resultid=1826735075
Second
http://setiathome.berkeley.edu/result.php?resultid=1826735069


GTS 250
multiProcs 16
clockRate = 1900000

Radeon 4870
Max compute units 10
Max clock frequency 850Mhz

From them numbers it looks like it wont get that much faster with newer builds

16 Processors at 1.9ghz vs 10 at 850mhz
sounds about right at 3 times longer per wu

My bad i should have picked up another Nvidia

---

The big disadvantage of HD4xxx GPUs - they have no local (shared) memory available for OpenCL app. It makes almost all kernels memory-bound. And being memory-bound computational performance of card per se means nothing. Memory controller performance only matters....
In some degree this solved in HD5 build for HD5xxx and up GPUs.

The main goal of released builds - to give ATi GPUs owners chance to use their GPUs for SETI, not to incline to buy ATi GPUs instead of NV ones :P

Its my own stupid fault

I needed another gfx card for a bedroom pc that i built out of spare parts

I waited for 2 weeks for a GTS 250/260 to go cheap on ebay and kept getting
pipped at the post

So gave up and got a Radeon 4870

I moved the GTS 250 to the bedroom pc and Fitted the Radeon 4870
into the living room pc

Being as they are slightly faster than the GTS 250 in games
I presumed they would be atleast close with seti

---

Normal & vlar's run around 50 min on my 4870, with vhar's running around 15 min. On the plus side an ATI card and run a vlar w/o choking on it.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[