M17x R2 7970m CrossfireX Ultimate Installation and Tweaking Guide

so is there a chance this is only bios not compatibile or a 100% bad card lol ?

 

"During benchmark I can see the picture jumping periodically. As if it stutters."

Well that's interesting. And it sounds familiar. I assume each card on their own works just fine. Anyway, my replacing the crossfire cable improved the situation greatly, though it didn't remove it completely. You should look into that. You could also try extra-shielding the cable with electrical tape, maybe.

Stop being stubborn bro, I confirmed it like 3 times that imsolidstate's upgraded mod works 100%. :thumbsup:

HwInfo clock reports are broken. Look at Throttlestop for accuracy. I used to get 7GHz readings in HwInfo.

Actually, just install it in his MSI. It'll probably work plug-n-play. Very good to know, though, for future wisdom

Sorry bro, but... Atiwinflash. Use that app long enough and you can guarantee yourself a brick sooner or later.
Always use atiflash in dos mode.
I hope you can reflash it as a slave (you may need to force it).

I'd say you're looking at 90% chance of incompatible vbios.

 

i'd say i am looking at 90% bad card.....

 

Well, that would be a possibility too, but I can't manage to find a beep code table anywhere, so... just my opinion
I'm just saying that behavior is in no way unexpected, since 8970m working in the R2 is something I find quite unlikely if keeping in mind that recent 7970m models don't work either.

 

Was the 8970M purchased with stating that it was to be used in your friends MSI machine? If so, I don't think the card is bad, just incompatible with the R2 ie. vBIOS and was probably loaded with a vBIOS that works on his machine.

Does his machine support enduro enabled cards? I hope its not the card as that would be much more headache.

 

Hmm if I upgrade to a 9970M I'll be sure to offer up my early version 7970M to M17x R2 users! I didn't realize the R2 was picky with GPU's! Is this only with crossfire setups or in general?

 

Actually, afaik, anything AMD pretty much works, except the 7970m UEFI-vbios cards. I believe nVidia was a bit more restricted, but still, 680m and 780m work so...

 

The 780m never worked for the m17x r2

Sent from my DROID RAZR using Tapatalk 4

 

Hmmm... I was under the impression that flingin's had worked before going bad. Maybe I'm too tired at this hour.

 

ithink he had it working with only an external monitor. his laptop screen never worked.

 

dont worry. i am sure when you install it in the second slot and flash it with its old bios it wil work.

 

Hmmm works fine each, and right now i have it semi shielded in alu tape.
Clocks for me in HWinfo and throtlestop show the same.
I can run clocks higher than most, I get the frequencies, i can run 29 with OC in BIOS even.Not many can, but result speed is not impressing.

Fingers crossed i can re-flash it. Dont know when it stopped, all went black.
Dont know how to write the force flash command in atiflash.

On positive side, i can run normal voltage on GPU and still get to about 55deg slight OC.-
With higher voltage i probably can run more speed.

 

I have found the root cause !
My 8970 has been flashed with 7970m vbios by a mistake.

I will try to blind flash it with correct vbios, but i dunno if i will succed, anybody knows if blindflashing will work in this case ?

 

OMG!! who flashed it with the wrong bios??
they sent it that way??
try inserting it as a second card and flash it. or try blind flash

 

If it doesn't post (and it sounds like it doesn't, stick it into the slave spot and try to boot with a 7970m in the main slot and try to flash it like that.

 

plugged into secondary slot.in primary kept my working 5870m.
laptop will not post as well, tried to blind flash but nothing happens or beeps...lol

 

Have you tried booting with a card that you would work?

Sent from my DROID RAZR using Tapatalk 4

 

so you can boot with just the 5870 but not if you plug in the 8970?
ofcourse i suppose you didnt connect the crossfire cable. it may conflict if you did

 

of course i did not connect the cf cable lol bro.
i can see caps and scroll lock flashing and num lock stays on also

 

that hapened to my r2 when i had a bad bios in the card

 

Hey Trome71, this is the command I found on the web and used a while back to force flash thru ATIflash:
atiflash -f -p i bios.rom

I'm sure you know this but I will inform you anyway...
"i" is the gpu: 0 for primary, 1 for slave
"bios" is whatever name you have for the vbios.

I hope this helps you. Sorry to hear your misfortune. I know how it feels. I had a bad flash with my 1st 7970m using ATIflash in BIOS. I am not 100 % sure what caused it although I feel the vbios I was using may have been corrupted resulting in a bricked card.

Anyway, good luck.

 

last time i tried force flash in atiflash it complained about command error. This is why i tried atiwinflash.
Need to try again next week when i get back to my alien

 

Off-topic as hell, I'd like to thank you guys for the advice of adjusting the CPU heatsink screws while under load.
25/25/25/25 @ 100W TDP, fully loaded, maxes out at 90C. Stock voltage and bclk speed.
I'm having some troubles with .net framework and can't run 3dmark atm, but I'm dying to see that CPU score now.

 

That is good to hear bro !

Also, when you play games or whatever, do you hear strange buzzing from the card ?

i feel the buzzing is a bit niosier comparing to my previous card....or it is like this because i have my keyboard not screwed to the chassis.....just asking

 

I'll be honest, it's impossible to tell. We're talking 3 fans, an extra 3 fans from the notepal U3 from me, + various other electronic noise. I can't tell which is which.

 

Ok, i just want to inform people that are thinking to go 7970m CROSSFIRE that its not sure that they will need to do the psu mod to get the max out of their system .

For example my 240psu can handle 26x multies on the 920xm (+075V from bios) and 944gpu 1100 mem.




so go for it with your single psu first . it might work

 

I would suggest re-installing the card as another user also experienced buzzing when his card was not installed properly. It might also be the reason you're experiencing artifacting.

 

Ok another question, not only for those who won CF 7970m.

on what elements you have thermal pads applied ?

1.Ram chips only
2. ram + voltage regulators
3. all of above + mosfets (require minimum 4mm thermal pad )

Also. I have been playing arount today with thermal pads.

I have LOADS of them so i can experiment. But, it is a pain to get the thickness correctly because, R2 heatsink bends up the more pads you add.

Yes, bends up. So when i am trying to apply 1mm thermal pads, which usually are OK thick, and i apply them on RAM ONLY the heatsink bends only very little.

$^!@ starts to get real when i try to cover everything on the card. So when i put 0.5mm pads on Voltage regulators, pads will slide out because they are too thin. When i aply 1mm thermal pads in there THEN i have to put like 7mm Thermal pad ( Stack 2 Pads) on Mosfets.

The heatsink itself is starting to bend ridiciolusly, and RAM (1mm pads) contact gets compromized. What i mean by that ?

Let me explain. Look at the draw i made.



So when the heatsink bends up beginning from the Ram, i can see clearly that about ONLY 50% of the ram pad has a proper preassure applied, further away contact is very weak or none

This is how this looks from the side when i look on the Ram pad in magnification




Now Many of you may say that am as stupid as those drawings are, but i have a point.

My first 7970m Died, from no reason. I never OCd the card, and played Diablo 3 with vsync which resulted in card being utilized for about 60-70% only at all times.

Something failed over time because it was a ''bad card'' or because the pads did not do the job properly. And i want to exclude that second thing.

I am getting sick of those ones who put stock thermal pads from their 5870m's that almost felt apart over 5 years, then they say they do not have Pads anywhere else than on Ram, and they run Overvolt + Max stable clocks and their card NEVER dies.

.............Bum, Flingins card dies after a year, ''die virgin, forever alone..........''

 

i have 2 super stable 7970s and i use pads only on mems. no need to use on power regulators.

also my cards work 100% 24/7 as they mine litecoins...

 

Wait for it ....

BOOM

Anyway, I feel your pain, flingin I've had bad experiences with not using enough padding, as well as using too much. At one point, I tried to pad every surface on a 7970m. It wouldn't boot unless I removed padding from some of the chips.

Anyway, I'm using Option 3 from your list right now. I still haven't gotten around to installing the 14W/mk pads
Also, I'm using 1mm pads on the vertical ram lane, 5mm on the horizontal one and a mishmash of 0.5/1mm on the power regulators, and a full 5mm on the mosfets. I install the heatsink outside the machine, and then press the hell out of card/hs sandwich trying not to bend the pcb itself but push the heatsink into the card more. Since I'm still using the sticky old dell goo, they tend to get contact and "hold" the heatsink in place, lol.

In a completely off-topic way, I find it funny that at 25x multis @ 100W TDP, my card maxes out Prime95 @ 88C, but giving it the +75mV option increases this to like 98. So i can run this CPU lovely undervolted, but since 25x multi is as far as i can go stable, I can't go any higher while keeping this level of TDP without boiling again. And I had so many hopes

 

PHU i got the modda woring
Had to swap the cards, and did forceflash the bad secondary card. (used ATIwinflash since it was the secondary card)

Now i have 1050 and will try a slight OC on the GPU.
I have used 0.5 on RAM and regulators. I got the pads fastened enough to not beeing able to move them without tearing them in the corner.
using the pads i got with the card from eurocom with the primary card and my new Phobya on the secondary card, same fit, and i really see to it that i press them together when screwing.
I will check in on the temps of the cards. If the pads are tight but you still suspect not good enough connection a thin film of very good paste could help you instead of using to thick pads. (that can either bend the board/heatsink or give you not enough contact on the GPU. )

 

Good to hear buddy, congrats

 

I also think that the pads on the VRM components might be unnecessary but I've put them on anyway just as a precaution. I rolled the pads to about 0.1mm or maybe less, they're thin as paper now. I used those for the inductors and then stacked normal 0.5mm for the mosfets making sure they don't influence how the heatsink sits on the card.

I've also placed two bits of rubber at each end of the heatsinks (directly above mosfets) so when I install the keyboard it presses slightly down on the heatsink to make better contact with the card and its the padded mosfets. I'm using Fujipoly pads and they're almost like a thin sheet of play dough and can be rolled/squeezed into shape.

On my CLEVO cards with the hand soldered capacitors its slightly more difficult to get the heatsink to sit flush as some of them stick out slightly (less than 0.1mm).

 

13.10 beta 2 came out yesterday.

AMD Catalyst

 

Direct Link

 

cool... going to download now

 

Waiting for my second card to arrive.....and pulling out my hairs, i hate to wait!...

Also flashed Vbios of my primary 7970m to undervolted 0.975V. Seems very stable.

 

I can feel your pain. Think im going to undervolt my cards too.
Right now im actually running my old PSU just for testing purposes and I cant seem to find anything different from the 330W one.
Will have to do more tests and perhaps get me a current/power measurement thingymyjiddon to verify that i actually get 330W or how much my system drains.

 

@Trome71: If you can run fullscreen crossfire with the CPU overclocked and with no shutdowns, then it means that:

1. Your 240W is of good quality (delivering probably more than 240W - you have to OC the GPU's as well to make sure)
and/or
2. Your 330W is properly modded, otherwise you'd get shutdown at 240W power drain.

 

Ok thx.
I did run OC on both GPU's and on CPU at the same time using both the 240 and the 330, so it should probably mean both are OK.
Should perhaps go back to the 330 that should have more overhead then.

 

Yea, no use stressing the backup

 

welcome to the 240W monster psus! :thumbsup: mine never shuts down from overcurent.

 

Nice.
Now i have to figure out why i cant seem to find all my ram until a few minutes after starting the computer. CPU-Z shows 4 Gig first time i open it after boot, and closing it and re-opening it makes it see the both RAMS. BIOS doesnt even show them both on boot.

Then i have to figure out what happens with my CPU since i have this odd behaviour with low speeds. I really have to OC to get to 9K on 3Dmark 11.

 

Sounds like you may have a defective stick. I'd run memtest if I were you.
And regarding the CPU, if you run 25/25/25/25 @ 100TDP, you should get something around the lines of 6800 physics score, although I've been scoring 6400 lately, and I think it's a 3dmark or net framework problem.

 

They do the same regardless of which one of them in slot 0 and 1. Slot 0 is only to be discoverd.

Your settings will give me in the 5000 area. Low 5000.

Right now i had to remove one of the GPU's. Have had problems starting the computer. Re-installed windows eventually after a mid game lock.
Had probs starting the computer again and then re-installed windows. Had no probs starting after that but it stopped again during login after I installed catalyst again. Removed one GPU and now it starts all the time.

Could still have a power problem...... It stalled very often right before initiallizing the screen.

 

Hmmmmm.... what are your temps on the CPU during the test? Have you tried increasing TDC? I keep mine at 75.
Do you have other apps running in the background?
Have you made sure that the second GPU is well seated? Is the machine stable with crossfire disabled in CCC?

 

Ive had TDC at 75 TDP at 120. No other apps, that would normally stop 3DMark11.

The card had been running great after resolving VBIOS issue, and was running good before that too in Blackops for hours.
Only thing i had installed was the beta2 and then problems started during gameplay with full freeze.
All settings according to your guide.

 

If each card on its own runs perfect (and you should test each card individually, then most likely it's a crossfire cable issue (either badly seated in the connectors, or deteriorated, or unshielded, etc.).
Sometime soon I'll open up my own machine to change the thermal pads to higher performance ones and wrap the crossfire cable in electric tape in order to shield it and I'll get back to you guys.

You still haven't told me about your CPU/GPU temps, by the way.
Also, what's your pad layout on the cards? And also, what voltage are you running them at?

 

I am using 0.5 on regulators and RAM. Also have pads on some of the other components as shown by eurocom.
I have the original coolingpads from buying 7970M kit for one card and the new better ones on the other card to compare. Get temps to below 60 even when gaming when i run on my chillpad as i did when it shut down today. I have fantastic cooling keeping it cool even when OC'ed. I have used 1050 V and OC on GPU and CPU (my slow CPU) and no thermal problem.
I used no OC today when i got problems, not on CPU and not on GPU. After that it didnt even start when totally chilled.

My xfire cable is brand new, and wrapped in alu tape.
Will try to seat secondary card in slot 1 again another day when i get the time for it. Just need to install everything again after the crash first.

 

That's quite the boggling issue you have there. I can't think of anything else off the top of my head. See if both of your cards run properly single-card mode.