HD 5870 worth running...

Crunch3r
Crunch3r
Joined: 22 Jan 05
Posts: 90
Credit: 30237616
RAC: 0
Topic 198128

i've recently decided to give einstein a go and am wondering if i should replace a gtx 470 with a hd 5870.

Searching the forum has not given me any clear result of witch will be faster on PMPS XT...

Any advice would be appriciated. Thanks in advance.

MAGIC Quantum Mechanic
MAGIC Quantum M...
Joined: 18 Jan 05
Posts: 1695
Credit: 1042854322
RAC: 1379006

HD 5870 worth running...

YES the HD 5870 is worth running.

mikey
mikey
Joined: 22 Jan 05
Posts: 11889
Credit: 1828153331
RAC: 203683

RE: i've recently decided

Quote:

i've recently decided to give einstein a go and am wondering if i should replace a gtx 470 with a hd 5870.

Searching the forum has not given me any clear result of witch will be faster on PMPS XT...

Any advice would be appriciated. Thanks in advance.

I have one here and it is doing units like this:
8,031.34 1,361.28 3.66 4,400.00
8,059.18 1,368.30 3.67 4,400.00
8,070.30 1,398.83 3.76 4,400.00
8,024.82 1,379.44 3.70 4,400.00
8,154.17 1,459.03 3.92 4,400.00

It is running the Binary Radio Pulsar Search (Parkes PMPS XT) v1.52 (BRP6-opencl-ati) with zero optimizations at all, just stock. The gpu itself has been run hard and hung up wet ever since I got it, which seems like a LONG time ago now. So YES I too would say "it's worth it"!!

If you don't want it anymore I'm sure a teammate, or someone else closer to you would love it. Germany is probably a bit far for me to pay for the shipping. BUUUT my wife is going there later this year, contact me if you want to get rid of it.

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 536617661
RAC: 186379

Performance-wise that GPU is

Performance-wise that GPU is still good. I'd only want to run it eco-tuned, though. Lower the GPU voltage instead of overclocking the chip and push for high memory clocks (helps a lot here).

MrS

Scanning for our furry friends since Jan 2002

astro-marwil
astro-marwil
Joined: 28 May 05
Posts: 511
Credit: 402847498
RAC: 1051855

Hallo! The same results here,

Hallo!
The same results here, using 1 task at a time, and working continuesly without any problem.
When I was choosing it about 2 years ago, it had the highest GFLOPS/Watt in its class.

Kind reagrds and happy crunching.
Martin

Aedazan
Aedazan
Joined: 10 Jul 11
Posts: 4
Credit: 23859424
RAC: 0

I have this cute little

I have this cute little machine here crunching away with 5870's.
They are tough little cards.
All of them are recycled and are in various states, all of them are damaged to some degree and Ive rewritten the bios on 2 of them to fix issues.

https://einsteinathome.org/host/12011965

Chris S
Chris S
Joined: 27 Aug 05
Posts: 2469
Credit: 19550265
RAC: 0

I run 5850's quite well

I run 5850's quite well :-)

Quote:
push for high memory clocks (helps a lot here).

Interesting, I thought that Einstein didn't need high memory clocks? I usually wind mine back to the minimum to reduce heat output, and wind the core clock up for performance using MSI afterburner.

Waiting for Godot & salvation :-)

Why do doctors have to practice?
You'd think they'd have got it right by now

Daniels_Parents
Daniels_Parents
Joined: 9 Feb 05
Posts: 101
Credit: 1877689213
RAC: 0

RE: ... Lower the GPU

Quote:

... Lower the GPU voltage instead of overclocking the chip and push for high memory clocks (helps a lot here).

MrS


Is this a general "rule" or only a recommendation for ATIs or especially HD 5870 ?

I know I am a part of a story that starts long before I can remember and continues long beyond when anyone will remember me [Danny Hillis, Long Now]

Chris S
Chris S
Joined: 27 Aug 05
Posts: 2469
Credit: 19550265
RAC: 0

In all the Boinc projects

In all the Boinc projects that I have participated it, the general rule, at least for all ATI cards, is to wind the memory clock back and the core clock up. The exception I believe is/was Collatz that required high memory clock speeds because of the way in which the project was designed to crunch.

The answer I think is to run Einstein for a few units at max GPU memory speed, and note the time taken per unit, and the GPU temperature. Then wind the memory clock back, and take new readings. If the temperature drops quite a lot but the time taken doesn't increase by much, you have your answer!

My 5850 card runs nearly 10C cooler with low memory clock, and not much time difference on workunits. Of course anyone elses experiences may differ.

Experiment!

Waiting for Godot & salvation :-)

Why do doctors have to practice?
You'd think they'd have got it right by now

archae86
archae86
Joined: 6 Dec 05
Posts: 3145
Credit: 7024134931
RAC: 1805897

Chris S wrote:the general

Chris S wrote:
the general rule, at least for all ATI cards, is to wind the memory clock back and the core clock up.


Not good advice at all for Nvidia GTX 970 running Einstein Parkes PMPS. It benefits considerably from raising the P2 state memory clock speed from as-shipped. The available improvement from raising core clock is much less, though not zero on my card.

Quote:
Experiment!


Agreed.

Chris S
Chris S
Joined: 27 Aug 05
Posts: 2469
Credit: 19550265
RAC: 0

Each BOINC project has its

Each BOINC project has its own algorithms that it uses to process their workunits. Some projects favour AMD cards. others favour ATI cards. Some projects have differing types of workunits in their projects, e.g. Seti has Astropulse and Multibem. As archae86 says, experiment.

It would be nice if there was a definitive set of advice somewhere that listed all the popular BOINC projects, listed their type of workunits, and suggested which cards would get the best results and the optimum settings for the cards.

I'm afraid I'm an ATI man, I have no knowledge of AMD stuff, but there are many others that do. Never be afraid to ask around :-)

Waiting for Godot & salvation :-)

Why do doctors have to practice?
You'd think they'd have got it right by now

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.