same cards different hashrates in miner

[wijnand1]

hey all,

I got 4 290x xfx core edition card running on a rig.

When i run the miner it seems that 2 pairs form. 2 card running at about 750 and the other 2 at 890. All on the same config:


"intensity" : "20",
"worksize" : "512",
"lookup-gap" : "2",
"vectors" : "1",
"thread-concurrency" : "32765",
"temp-cutoff" : "95",
"temp-overheat" : "85",
"temp-target" : "80",
"expiry" : "120",
"failover-only" : true,
"log" : "5",
"queue" : "1",
"scan-time" : "1",
"device" : "0,1,2,3",
"expiry" : "60",
"failover-only" : true,
"temp-hysteresis" : "3",
"gpu-fan" : "60-100",
"auto-fan" : true,
"gpu-engine" : "920",
"gpu-memory" : "1350",
"gpu-powertune" : "15",
"gpu-threads" : "1",
"no-submit-stale" : true
}


another strange thing is that in the miner the memory does not go over 1250?

anybody a clue?

[1714816204]

1714816204

[1714816204]

1714816204

[1714816204]

1714816204

[1714816204]

1714816204

[1714816204]

1714816204

[1714816204]

1714816204

[ryancpwalsh]

Not sure if relevant - but I had to put a 90 second delay in my startup script that launched cgminer such that when the machine boots into windows, all of my miner settings worked.

What I found was if I launched cgminer too soon after startup, I couldn't get it to utilize higher memory clock rates, and fan speed adjustments (it's as though the driver needs time to load or something similar prior to making all features accessible to cgminer).

While I can't back this up with a technical reason, it is my observation, each and every time I start cgminer too soon after startup.  The delay needs to be somewhere between 45 and 90 seconds for me, I use 90 to play it safe.

[wijnand1]

i run the miner manually since i am still testing, even when i restart the miner after it has been on for a while it does not make a difference. thank you for your input anyway

[wijnand1]

Nobody encountered this or has an idea to solve this??

[the joint]

Try kicking your powertune up to 20 or 30 and see if that helps.  Also, what PSU are you using? 

[wijnand1]

Yes i tried this and it only results in more watt usage. Even if i try to set the two lower performing cards on higher engine, memory or powertune they dont preform better than before.

I got a silverstone 1500W PSU. From the wall I only use about 1180W at these settings.

[howardb]

Yes i've seen similar problem on Asus R9-280x Direct CU, it can happen for a number of reasons:

a) Temperature variations, the R9's have a sort of semi autonomous behaviour which causes them to throttle down when internal temp sensors tell it to, regardless of your cgminer settings.

b) Not all memory is equal, you can get slightly lower quality memory in one card over another (even from same production batch), this can cause significant differences in hashing rate. I've also seen the best hash rate degrade over time for some cards.

You could try swapping your cards around, see if the problem stays with the slot or the card.

[wijnand1]

I suppose this is not something you can go back with to the shop?

I did swapped the cards and fisrt it was gpu 3 and 4, and now it is 1 and 4, so its seems to be connected to the card itself.

Thanks for your input

[howardb]

I suppose this is not something you can go back with to the shop?

I did swapped the cards and fisrt it was gpu 3 and 4, and now it is 1 and 4, so its seems to be connected to the card itself.

Thanks for your input

I know i'm probably stating the obvious, but have you checked the firmware version on each card?

[wijnand1]

actually I didn't because first of all im do not have experience with bios and because there is a bios switch on the cards, used for OC or silent function.

How should i check the bios version (running windows 7).

[howardb]

actually I didn't because first of all im do not have experience with bios and because there is a bios switch on the cards, used for OC or silent function.

How should i check the bios version (running windows 7).

Download one of the popular overclocking tools (I think XFX do their own), and it should show you the firmware version on each card. Find via http://www.tomsguide.com/us/pictures-story/445-graphics-overclocking-software-tips-tools.html

[wijnand1]

Thanks,

MSI afterburner shows both cards are this bios version: 015.039.000.007

[wijnand1]

Update:

Got a second rig being build, same specs. 2 cards are running at the moment and the same phenomenon appears, one at 900 other at 750. Same bios, same config

[wijnand1]

could this just be a problem in the miner itself, since the total averages of the two rigs are more or less the same

[dealio]

gpu-memclock not gpu-memory, that's your problem

also use these values instead of what  you have

"intensity" : "19",
"gpu-engine" : "1000",
"gpu-memorymemclock" : "1500",
"gpu-powertune" : "20",

[wijnand1]

thank you very much i will try this.

Will this also solve the difference between the cards?

[wijnand1]

Update:

Second rig running on 4 cards as well. The same problem occurs. But it does not seem to be an issue with the cards itself or the slots they are in.

When running on 2 cards the 1st and 2nd differ in speed. Adding the third card made the first two run at the same speed and the third one faster. When adding the forth card number 1 and 3 were slower than 2 and 4.

Any help would be very welcome,

thank you in advance.