Two Radeon cards should equate to double the hash rate right?

[michaeljh]

Long time lurker, first time poster.
Even though I've been mining for a (relatively) short amount of time, I still have much to learn. Yes I am a newb, but I'm not stupid. I've had two miners each with their own Sapphire Radeon 6970 (+20% voltage, 1450/950) running 24/7. Strike me as much as you want for running Windows but remotely controlling a Linux cluster will happen eventually. For now, It's Win7 + VNC.

Santa paid me a visit (who knew he drives a big square brown truck and dresses all in brown!) and now that I have a power supply with the adequate PCI-e power connectors, I can put these two cards in one machine instead of two. Sounds like the prudent thing to do right?

I get in the 425 - 435Mhashes/sec range with each of the cards. I connected both cards on one board (two PCI-e x16 slots) and my train of thought is that I should be in the 860Mhashes range right? Sadly, GPU0 in cgminer is only giving me 325 - 350. Both cards are the exact same Sapphire 6970 (no these are NOT unlocked 6950s) so I'm thinking performance should be the same across the board.

What I've done so far:

• I swapped both cards into the other PCI-e slot and the same thing happened: GPU0 was in the 325 - 350 and GPU1 giving 425 - 435.
• Each individual card can run at the full hash rate without any problem with the second card isn't plugged in.
• I've never had them running in Crossfire

This tells me that the problem isn't unique to one of the cards, it's whatever is in the first PCI-e slot.

Attached is the screen shot. Yes, GPU1 is a little hot but I'll be fixing that in the near future with PCI-e risers giving them more room to breathe. For now, I just want to see a consistent  850Mhashes range. If I can push it to the max at 855 - 860 holy shit I will forever be in your debt.

https://i.imgur.com/ydWrv.png

Possibilities I'm thinking in my head:

• What's the chance the bottle neck could be the hard drive? For mining, I used an IDE drive that I didn't care about but I do have some SATA 2 and SATA 3 drives laying around. Hell if I find out speed's the issue, I have a spare OCZ Vertex 3 that it could use
• I HIGHLY doubt it's the PCI buses but I'm so stumped, I'll questioning everything. This board's PCI-e slots are 1.0 with the secondary x16 slot running x8 speed if both are being used. That should be PLENTY of bandwidth for mining.
• just to make sure, cgminer can run both cards under one worker when reporting to the pool right? Currently it's one worker under this cgminer install and that's okay right? I don't need a second worker for the second GPU correct?

Any advice would be tremendously appreciated

mobo: http://www.newegg.com/Product/Product.aspx?Item=N82E16813136158
cpu: AMD FX-60

[KWH]

Looks like one GPU has a lower clock. Look at CCC and see for sure.

[michaeljh]

Looks like one GPU has a lower clock. Look at CCC and see for sure.

When looking at the specs at AMD Overdrive, it doesn't specify which GPU it's indicating. I would assume it would be both.

https://i.imgur.com/ShQej.png

[KWH]

That's for 1 card, is the other not listed, it doesn't look like it? You can also install Afterburner or use GPUZ to check.

[DeathAndTaxes]

something is "hosed up" in your drivers then.  You should have two entries in the catalyst control center on for each GPU.
In cgminer press [G] and then check the clocks for each GPU.

Windows sometimes get "weird" when you install a GPU, then switch them around, then install other one, etc.  It doesn't seem to drop the devices cleanly. 

[michaeljh]

Well one thing got fixed and another problem came up.

That's for 1 card, is the other not listed, it doesn't look like it? You can also install Afterburner or use GPUZ to check.

I'm not sure if Afterburner solved it but this is what happened: I installed Afterburner and also turned off auto-fan in cgminer. I simply put both fans on 80%. It booted right up and lo and behold the hash rates were both ideal! That's the good news. The bad news is that the secondary GPU has become extremely crash-happy.

something is "hosed up" in your drivers then.  You should have two entries in the catalyst control center on for each GPU.
In cgminer press [G] and then check the clocks for each GPU.

Windows sometimes get "weird" when you install a GPU, then switch them around, then install other one, etc.  It doesn't seem to drop the devices cleanly. 

Afterburner sees both GPUs successfully but when I change the overclock settings in CCC, I assume I'm only changing it for the first GPU. Drivers seem to be behaving in Windows. My educated guess is that since the second GPU is crashing anywhere between 30 - 90 seconds of mining, the first GPU is getting the extra overclocked voltage but the second one isn't. Any idea on how to change that?

https://i.imgur.com/1iA5E.png

[KWH]

Unlock the over-volting, restart afterburner and up the voltage in small increments. Watch the temps. Also, drop the Mem speeds by a lot, say to 800-900 Mhz.

[RyNinDaCleM]

In Afterburner, in the settings menu, there is a drop down menu to choose which card you are tweaking, then click ok. Set the values for the other card.
Alternatively, there is also a check box for "synchronize settings for similar graphics processors" you can check that and clock both cards with one set of settings.
When using MSI AB you will want to close CCC, since they have conflicts with each other (or at least they used to and might still).

[michaeljh]

What would be the ideal gpu-threads setting? It seems to be slightly more stable (I'm not referring to the hash rates I'm referring to the computer not crashing) at 1 per GPU. Google couldn't turn up a good explanation of gpu-threads in cgminer. I'm curious as to if it'll affect whether GPUs like to crash.

Unlock the over-volting, restart afterburner and up the voltage in small increments. Also, drop the Mem speeds by a lot, say to 800-900 Mhz.

I changed the voltage in AB and the second GPU is in fact crashing drastically less. However, cgminer is reporting it to be 1.175. What gives? One of them is lying and I'm not sure which or why.

In Afterburner, in the settings menu, there is a drop down menu to choose which card you are tweaking, then click ok. Set the values for the other card. When using MSI AB you will want to close CCC, since they have conflicts with each other (or at least they used to and might still).

I have CCC deselected on startup. AB is running on bootup. They each have their own setting so as to figure out why the second GPU crashes but the first one hasn't crashed once.

I'm still conducting more testing on getting it to not crash. However, how do I even know if changing the parameters in AB are having any effect if cgminer is saying otherwise?

https://i.imgur.com/ShiHN.png

[KWH]

Intensity on one is 9 the other is 12. I'll wager GPU0 is crashing.

[michaeljh]

Quite the contrary. If/when a gpu is crashing, it's in fact the second one. I lowered the intensity so as to lower temperature. The one with I 12 hasn't once given any problems

[KWH]

Set them both the same. Also, when I see usage lines like this, it needs a bit more Mem speed. 84c, lotta heat. If this doesn't work, I would lower the Core clocks. Are you doing anything else with this box or mining only?

Some GPU's just won't work properly with mining, maybe it's nearly done or something else. I have 2-6950's and 2-5830's that won't overclock worth spit but run fine at stock speeds, minus Mem speed.

[michaeljh]

Here's the current setup:

https://i.imgur.com/yJYXq.png

Yes I know those temperatures/fan speeds are far from ideal. However I'll leave them be for a few day until more PCI-e risers come in the mail. Once each GPU have enough adequate breathing room, I'll crank them up both on max. I'm content with the 830Mhashes for now and anticipate 860+ after rearranging them.

My concern right now is Afterburner configured at one setting but cgminer reports at most 1.175. What gives? This is a headless dedicated mining rig responsible for nothing else. Win7 + VNC

[RyNinDaCleM]

Set them both the same. Also, when I see usage lines like this, it needs a bit more Mem speed. 84c, lotta heat. If this doesn't work, I would lower the Core clocks. Are you doing anything else with this box or mining only?

Are you referring to GPU %usage? That can be remedied by going into task manager->processes tab, right-click on cgminer and change affinity to a single core. This will also help with general usability while mining.

[KWH]

At 91c, you'd better turn something down or you'll have a fried GPU. Run GPUZ and post what it shows.

[michaeljh]

I should have just run GPUz first. I already routinely use it so I guess I was just herping and derping when cgminer reported one value and afterburner reported another. Thanks to GPUz, I now have an always-accurate measurement of voltage. Let the record state that for cgminer 2.10.3, you can't always trust the voltage and the --gpu-vddc parameter is useless for my setup. All my questions have been addressed. Thank you all.

[RyNinDaCleM]

I should have just run GPUz first. I already routinely use it so I guess I was just herping and derping when cgminer reported one value and afterburner reported another. Thanks to GPUz, I now have an always-accurate measurement of voltage. Let the record state that for cgminer 2.10.3, you can't always trust the voltage and the --gpu-vddc parameter is useless for my setup. All my questions have been addressed. Thank you all.

Glad to hear you got it working!
That's why I would never use cgminer to OC, verify clocks, voltages or temps. It's mining software not OC'ing software. Only MSI AB for me!