Bitcoin Forum

I just set up my new 7970 for hashing, but can only push up the clock speed to 1150Mhz (of a maximum of 1500Mhz), before my drivers start crashing.
This is keeping my hash rate for this card down to 500-620Mh/s.
Also, at this speed, my fan is running at 84% and GPU temperature is a constant 76c
I am running it purely for hashing, while I use my 7850 for standard use, AND hashing.

I have 3 x 16x PCIe slots, but if I put both cards together, the 7970 heats up the 7850 terrifically so I've had to put the 7970 in the bottom slot.

Also, has anyone else found that the last version of catalyst dropped hash rates? I used to get 305 from my 7850, now, since updating my drivers, I only get 280.

Does anyone have any tips on how to get the best out of this set up? All help is appreciated.

Same rig. Quite surprised to be getting only 230 Mhash/s on a 7850 despite tweaking with all cgminer parameters. Driver is 12.100.17.0 of 3/20/2013, Catalyst version 13.3

My bfgminer shows 222 Mh/s average for 7850 and I can't make it better. I keep GPU running 1000 MHZ and memory 900. Temperature is 70-71C and fan 41% load.
Some miners post 320MH/s for this card, but I suspicious that it is not the average, or they abuse the card keeping temperature above 80C.

Well, I honestly used to get a constant(ish) 306MH/s on my 7850, with no real fan noise, and a low temperature, until I updated my drivers - and that was with it mining 24/7, for weeks on end.
Now, with Catalyst 13.4, Drivers 12.104 - I get only 280MH/s
I have re-installed the OpenCL part of the drivers, but that hasn't fixed it at all.

I am loathe to delete the drivers, and re-install the old ones, as I tried before, and although everything seemed fine, GUIMiner would crash on startup, and I wasn't able to fix it. until I did a system restore.

I have resolved some of the 7970 issues, getting a good 660MH/s, and about 73c temp, fan speed 80% (by clocking the GPU at 1160MHz - which seems to be the fastest I can get with stability, and the memory speed down to 1000MHz) - but now the GPU usage keeps dropping from 99% to as low as 65%, and I have no idea why.
I have -v 1 -w 128 in the flags window, and the card is in my 3rd PCIe x16 (running at x8) slot, due to temperature issues if I put it right next to my other card.

I wish I had that extra 10% back, though, as I would be getting a decent 720MH/s from my 7970, and over 300MH/s from my 7850 if I did - or just over 1GH/s :(

Q: Which AMD SDK is the best for cgminer?
A: At the moment, versions 2.4 and 2.5 work the best for R5xxx and R6xxx GPUS.
SDK 2.6 or 2.7 works best for R7xxx. SDK 2.8 is known to have many problems.
If you are need to use the 2.6+ SDK or R7xxx or later, the phatk kernel will
perform poorly, while the diablo or my custom modified poclbm kernel are
optimised for it.

Q: Which AMD driver is the best?
A: Unfortunately AMD has a history of having quite a few releases with issues
when it comes to mining, either in terms of breaking mining, increasing CPU
usage or very low hashrates. Only experimentation can tell you for sure, but
some good releases were 11.6, 11.12, 12.4 and 12.8. Note that older cards may
not work with the newer drivers.

https://en.bitcoin.it/wiki/Mining_hardware_comparison

Is it possible to run the older AMD SDK with the newer version of Catalyst, or do I have to roll back drivers too?

set the gpu core to 1040 on your 7970.

Will dropping my core speed increase my hashrate? It seems a backward step.

I use

Driver Packaging Version         12.104-130328a-155980C-ATI
Catalyst Version            13.4
OpenGL Version               6.14.10.12217

With intensity 10, GPU clock 1000, and memory clock 900 I have average 222MH/S only.
I noticed that thread 0 has 90.4 MH/S and thread 1 has 131.4 MH/S. It would be nice to have them both 131.
Changing intensity did not change this proportion. Intensity of 11 did not make speed better, but increased CPU load from 1-4% to 45-85%.
I use bfgminer 3.0.0.

oroboras, can you please post your mining settings for 7850?

Thanks in advance

Hi!
660 Mhash for ~ 1150 Mhz seems about right.
I think with most of the cards it is hard to reach 1200 Mhz without water cooling. I have 7970 Ghz edition and the most I can get is 1170 at 80 deg C (in open case). With higher clock the card crashes and I think even at this frequency it is sometimes throttling down.
Most of the higher rates reported are problematic to run for long term, so I would accept 660 Mhash. If you need to invest money into cooling, then maybe it is better just to buy another card than max out the one you have.

Sure.

7850 info:

I am using GUIminer, with no extra flags at all.
GPU clock speed is maximum (1050MHz), and memory at 1220MHz. Fan speed is set on auto (runs at around 64%) with CPU temp at 67-68c

It is a Sapphire HD 7850 OC (2Gb mem) model, and I'm running all 3 of my monitors (ok, 2 monitors and a TV) through it, with an active link cable, and using MSI Afterburner for overclocking.

7970 info:

I am using GUIminer, with the following flags; -v 1 -w 128
GPU clock speed is 1160MHz, and memory is at 1000MHz. Fan speed is set on auto (runs at around 85%) with CPU temp at 77c

It is a Sapphire HD 7970 (3Gb mem) model, and I have no monitors plugged in, and it is on an 8x PCIe slot.
Driver Info:

Driver Packaging Version   12.104-130328a-155980C-ATI
Catalyst Version                   13.4
OpenGL Version                   6.14.10.12217

I hope that helps you out. :)

no problem just when installing new drivers, select custom install and make sure the amd app crap is not checked.

Well, uninstalled all my drivers, reinstalled the older driver and the 2.7 SDK, and hey presto! No difference....

I have no idea why it won't give me that extra 10% back. The big meanie!

If your using CGMiner you need to delete your .bin files and let CGMiner create new ones when you change drivers/SDK.

I'm using GUIminer, but cheers for the advice. I'll remember it for when I change over to CGMiner - when my ASIC finally arrives, as I don't think GUIminer is compatible.

oroboras,

Thank you for the info. Without intensity setting my card was only under 60-70% load. With intensity 10 load is 98-99%.
GUIminer is not really a miner but only a front end for the miner (it can work with bfgminer and some other). I downclocked memory speed to 900, as I heard that keeps temperature lower and not important for bitcoin mining. For litecoin mining it is important though.

Thank you again and keep a good mining!

Oddly, I discovered that this evening - I set up CGMiner miners in GUIMiner - and my 7970 is now hashing at a lovely 690MH/s (or approx. 30MH/s more), but conversely, my 7850 loses about 40MH/s - but I expect that is as it is running my monitors too.
I'm happy to run one through CGMiner, and the other straight in the CLMiner ability of GUIMiner - I am now shy of 1GH/s by about 40MH/s

I will carry on tweaking it as I find out more

I have several 7970s and was running about 730 MH/s when I was mining BTC, I switched to scrypt mining and I'm running 730 MH/s per card. I didn't have much luck with the newest catalyst drivers though. 13.1 seems to be working for me so I stick with it.

please try it.

ok, tried it - it dropped my hashrate from 680MH/s to 580MH/s.
It's back to 1160 now. My hashrate has returned to 680MH/s

Don't forget that running one card on one slot (100% dedication to bus-speed/calls/power), is not the same as running two cards on two slots. It will NEVER scale-up to 2x, what any single card will get, at any time.

The "controller" for the bus uses an internal "thread", to distribute calls that the CPU is sending all at once. There is a slight buffer in there, when more than one card is installed. (It has to send data through x-dedicated-lines, and y-shared lines, to "simulate" full 16x support.)

Also note... Though you may have 3(16x) slots... rarely will you ever have 16x/16x/16x dedicated speeds. At the most, with one cheap controller, the slots reconfigure for...
(16x/8x/4x) or (16x/16x/4x) or (16x/8x/8x) or (8x/8x/8x) or (8x/8x/4x) etc... after adding a card.
Prior to adding three cards, slots may be like this...
(16x/xx/xx) -> (16x/16x/xx) -> (16x/8x/8x)
or
(16x/xx/xx) -> (16x/8x/xx) -> (16x/4x/4x)
or if only one card, in any slot...
(16x/xx/xx) or (xx/16x/xx) or (xx/xx/16x) They do this so you are not LIMITED to only stuffing the card into the first slot, like AGP. Not so that all slots will run at 16x speeds, all at once.

If all slots are 16x, they can still be shared-16x... (Effectively being only 16x for the most demanding card, on all slots, but the total bandwidth will not be 16x*3, ever.)

Thus, one card has that full capacity, one runs slightly less, one runs even less. But in a auto-balanced-load, all cards run the same speed, >4x or >8x and <16x.

To get more than 4(16x) true slots, you would need two controllers, and two CPU's. Controllers are physically limited to only 7-slots max, for PCIe. Even with load-balance and buffering, and sharing.

NOTE: Using all 1x connections actually yields higher numbers. There is NO buffering, and no load-balanced-shared-lines. Each has dedicated busses directly to the controller. (Thus, no "stalling" from that issue. Also no "cross-talk" noise on the datalines.)

ALSO NOTE: SATA and USB and any onboard-sound or WiFi devices may ALSO be using that same controller as a directly wired PCIe connection. (Some use PCI, but PCI is all on that 8th slot that is unseen.) That is more "call events" and buffering.

Call-event -> send-data -> process data -> return-data. (If you have more time processing harder data, you have less calls. Thus, you see less "stalling". That is the draw-back of having "small hash-difficulty"... You are processing faster than the call-times. Longer processing, and the call-times are like 1%, not 50% of the time. Using arrays and pre-sending while processing, into the buffer, reduces this variance. Sending each call one at a time, is taxing. But that is the programmers issue, and the pool-operators issue, using low difficulties for shares. Solo-mine, and you never see this issue as bad.)

In short... the cards are just too damn fast for the program. :)

Thanks for the information.

I am using my 1st (16x) slot, and my 3rd (8x) slot - to combat heat issues that occur if I use slot 1 & 2.

I am happy with my speed I am getting now, at just under 1GH/s over all.

I can verify that my 7850 CAN DO 330MH/s+ hash rates, STABLE 24/7. I can post screenshots if needed, but here are my settings:
ASUS Radeon 7850
-Fan speed @ 67% - Temps never go above 60C
Core clock: 1175 MHz
Memory Clock: ~1000MHz I believe
Voltage: 1.175V

Using latest version of GUIMiner, that is the ONLY program running on the computer, other than GPU-z and Sapphire Trixx, with flags -v -w -128 (or the one for the highest hash rates, can be shown by hovering over the flags box in GUIMiner)

This system is powered by an i7 2600k @ 4.4 GHz (don't ask why, I just had the 2600k lying around), 6 GB XMP RAM, 750W PSU, 1 TB HDD.

If you need any more info, or want screenshots then reply and I can get some up for you.

How did you get your 7850 core to 1175MHz? MSi Afterburner will only allow mine up to 1050MHz. It's totally stable 24/7 at this speed, and only 68c in temperature, with the fan at 74%.

I'd love to push it higher though, somehow.

Sapphire Trixx has unlocked overclocking. It voids warranty, but allows for much higher overclocks and voltage adjustments. I'm also using the newest AMD drivers I think for that GPU.

Thanks for this, it clears some issues I have been having. Very informative!

If you disable anything you are not using in windows HARDWARE (Not just uninstalling drivers and software)... That helps a little, potentially.

Thus, disable the onboard sound, one or two of the USB busses not being used. (Leave one, just incase you pick the wrong one, and need to move your mouse/keyboard.) Disable onboard network-controller if you are using wifi, or vice-versa. Disable on-board graphics, if you have that, not being used. That frees-up memory, calls, and bus-width. Not to mention, it saves power, since they should all be "sleeping".

Oh, and IDE if you are using SATA for your HD. (Caution, disabling IDE may be linked to SATA. It will not shut-down if it is, and will re-boot back to normal if that is the case. Try that separately.)

@oroboras

680M#/s at 1160MHz sounds about right for 7970

I have 2 7970s which are Asus (1000 stock) running 1175 / giving 705 & Sapphire (925 stock) running 1075 / giving 645 - the M#/MHz rate looks the same for both

Sapphire is working in PCI-E x1 slot connected through x1-x16 riser. I don't feel any difference while i am using this Sapphire on x16 or x1

please try -v 1 -w 64, turn on --gpu-threads 2 & try to use -I from 7 to 9 (i.e. 7,7 for both cards)

@ISAWHIM

I always thought that the thing you explained here matters in gaming, not mining


Maybe the answer for all of that is "mine on Linux" ? :) As far as I remember, my first 7970 was doing a little worse on Windows but I didn't use 2 GPUs on it

regards / swr
-----
1F3xsZb1mPePE6gS9eK5kZ1esRd6bHHr8J

Quick question. So currently you are using 7970 AND 7850, installed at the same time? Is it a crossfirex mode or simply just on dedicated mode?

Because I have 7850 at the moment, and i wish i could still use it when i buy and install the new 7970 next month

No, not X-fired at all. I am using the 7850 for all my monitors too, as the 7970 seems to really drop hash rate when running monitors on it at the same time.
I'd advise you to seperate the cards if you can (if your mobo has space) as having them together REALLY ramped up the heat issues, on my machine.

Plus - if you have Sapphire cards with Dual X coolers, they are deeper than dual slot, I discovered.

I'm looking to buy a 7970.

The motherboard that comes with the rig has 1x 16x PCIe but if secondary card is pushed it it goes down to 8x/8x. What impact will this have? I'm not looking to have a SLI cause I don't really need it for scrypt mining?