OK, everyone with an AMD ATI Radeon HD 4700/4800 (RV740/RV770) or similar low level OpenCL 1.0 only GPU, who managed to get any work validated, hold up your hand. Tell what OS you use, which drivers.

I don't believe any work around here for the HD4xx0 cards will validate, when put against anything else. I looked around a bit and see that all ATIOpenCL will not validate against other ATIOpenCL (not even when the same model GPUs are used), and not against ATIOpenCLLion, CUDA32, CUDA32OSX, CUDA32NV270, CUDA32FULLCPU, or CPUSSE.

Since the developers here have fallen very silent on this subject, or are spending their time on fixing the upload and download servers, I thought I'd ask.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

Windows 7 - 64bit
HD4850 - 1GB
Catalysts 11.11

17933 was done yesterday, it will not validate as it's paired against a CUDA32.
17971 did not validate as it was paired against a CUDA32.
17950 is what I am running now, I've never seen what it does when paired against a CUDA32OSX, but somehow I doubt it makes any difference.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

I am normaly running on ATI 5850

Now attached a second ATI card 4850

Now running on 2 cards.

The 5850 performs almost normal.... Now it uses 100% CPU to feed the GPU.
It hasn't used that before when I ran it alone.

The 4850 tasks all ends up "computation error" and the messages box says "output file abscent".

I will now remove the 4850 from the computor.

I hope to see that the CPU% gets lower when feeding the 5850 alone. 1 card in the computor.

---

Yep.

As I thought. Running only 1 card then the CPU goes back to normal 15-25% to feed the GPU...

Sorry about adding a note about this when the thread as about 48xx cards.

---

17933 was done yesterday, it will not validate as it's paired against a CUDA32.

As said, it did not pair against that CUDA32 and is sent to another CUDA32.

17971 did not validate as it was paired against a CUDA32.

It was sent to another CUDA32 and did not pair against that one either, so it's sent to yet another CUDA32 to see what that one makes of it. The suspense.... Oh.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

ATI HD 4850 1GB
Drivers 11.8
BOINC 7.0.2
Windows 7 Ult x64

WU's are inconclusive. And I think that they will be invalid =(
Computer is 1393
http://albert.phys.uwm.edu/workunit.php?wuid=18378
and
http://albert.phys.uwm.edu/workunit.php?wuid=18184

But thanks anyway for your work. =)

17933 was done yesterday, it will not validate as it's paired against a CUDA32.

As said, it did not pair against that CUDA32 and is sent to another CUDA32.

It's now Completed, marked as invalid for me and another CUDA32, the other two CUDA32 have walked away with the money.

17950 is what I am running now, I've never seen what it does when paired against a CUDA32OSX, but somehow I doubt it makes any difference.

It finished, Completed, validation inconclusive.

Seeing how the other 3 tasks I had were all paired against CUDA32 tasks, I aborted them already. There's in my opinion no use in burning electrons when they will never validate anyway. I've gotten 6 new tasks, in 3 of which I am paired against CUDA32 (which I'll abort again), 2 against other ATIOpenCL but higher denominations GPUs (a 58xx and 6xxx) and one undecided as at this time it's unsent.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

Seeing how the other 3 tasks I had were all paired against CUDA32 tasks, I aborted them already. There's in my opinion no use in burning electrons when they will never validate anyway. I've gotten 6 new tasks, in 3 of which I am paired against CUDA32 (which I'll abort again), 2 against other ATIOpenCL but higher denominations GPUs (a 58xx and 6xxx) and one undecided as at this time it's unsent.

The Unsent was resent eventually and of course to a CUDA32. So I aborted that one as well. Will only run the two paired against other ATIOpenCL ones now, but don't have much hope they'll do much of anything. It'll be more like we both will end up with nothing, since all that work will be resent twice to two CUDA32 who in the end will get the work validated.

We know already that the lower end ATI GPUs don't validate against the rest of the world or each other. Is this information useful for the project, or not?

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

14141 had me paired against a HD69x0 GPU with Catalysts 11.11; Completed, validation inconclusive.

Is this of any use here, or are we just burning electrons because we can?

I've aborted the last one as well and set BOINC to ignore my GPU at start-up. There's at this time no use to run work here, other than to add to an electricity bill because we can. I'll be back when very strict changes have been done to the validator. It's been over a month now that I first reported that ATIOpenCL 1.0 does not validate against itself or 1.1, I've just done it again. Maybe that this time someone gives a damn.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

Hi!

Nevertheless, thanks for your input and tenacity. I think it does make sense to retest this whenever a) a new driver version comes out or b) a new app version (I haven't read anything that excludes a driver problem as a factor here, since it's the same app, and it's doing reasonably well for openCL 1.1 cards).

I think unless either a new app version or driver version comes out, it really makes little sense to try this particular card any further here.

HBE

---

A retest only makes sense when the validator has been adjusted, or a new app comes out.

Perhaps it's wiser though that the project should just throw their hands in the air and give up on trying to force a support for OpenCL 1.0 cards, make the requirement OpenCL 1.1 for ATI cards and be done with it. Exclude the 43xx, 45xx, 46xx, 47xx and 48xx GPUs, perhaps all the others with Beta support as well.

Want to do work with those GPUs? Go to the competition projects or upgrade. ATI cards are cheap, very cheap when compared to hard drive prices.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

I was running a 4870 and had all the same issues Ageless had. Then, about 3-4 weeks ago, I found a killer deal on a 5870 and have since upgraded. I didn't even bother to put the 4870 in tandem with the 5870. Next on the agenda is a new motherboard, RAM (possibly CPU, waiting on Piledriver) and another 5870. The point is, I can't see ever going back to the 4xxx series.

Like Ageless said, maybe it's altogether simpler to deprecate support for 4xxx ?

---

Hmm... 4850 doesn't work with albert.

But I successfully run 4850 with PEOM@Home OpenCl app and it uses boinc 7.x.x as well.

I wonder what the difference might be?

---

As I wrote earlier, the computations for the BRP app have to be rearranged/resized to fit the hardware of the 4xxx series cards (you'll see some messages indicting this in the log output). Either this resizing process has a bug or it's causing a loss in precision. So there is still some hope that this can be made to work. Something the devs have to look into after the holiday season.

HB

---

As I wrote earlier, the computations for the BRP app have to be rearranged/resized to fit the hardware of the 4xxx series cards (you'll see some messages indicting this in the log output). Either this resizing process has a bug or it's causing a loss in precision. So there is still some hope that this can be made to work. Something the devs have to look into after the holiday season.

HB

Thank you for your answer.
I must have missed the earliar post.

---

I have ordered a Sapphire Radeon HD6850 with 2GB of GDDR5, it'll arrive tomorrow. We'll see what that one can do around here, although I'll probably be breaking it in exclusively by playing more Skyrim 4GB.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

My Radeon HD6850 just finished its first task, it got validated against a CUDA32. How refreshing.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

My Radeon HD6850 just finished its first task, it got validated against a CUDA32. How refreshing.

6850 is a nice card.

I'll just have to wait until the program supports 48xx cards.

Is Boinc version 7.07 scheduling any better then 7.03?

---

I am using 7.0.7 because it's needed for the VBoxwrapper test project.

My other two OpenCL tasks went a lot quicker, 4,800 and 5,600 seconds.
But then this one didn't validate against the CUDA32. Unknown at this time if this is because of me or him.

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

I am using 7.0.7 because it's needed for the VBoxwrapper test project.

My other two OpenCL tasks went a lot quicker, 4,800 and 5,600 seconds.
But then this one didn't validate against the CUDA32. Unknown at this time if this is because of me or him.

Oh, ok.

What will the purpose be of Vboxwrapper?
Will VirtualBox be replaced at T4T?

---