M17x R2 7970m CrossfireX Ultimate Installation and Tweaking Guide

As sangemaru mentioned you should try a lower voltage vBIOS to reduce the heat but honestly that is like putting a band aid on a broken finger. He's provided some really solid advice in his post so try to troubleshoot with that in mind.

13.8 Beta2's are out. I'm going to give them a try tonight. They only mention an AA fix for FC3 but I'm hoping there's other fixes for the game in there too.

 

13.8 first beta was pretty much the best AMD driver I've ever seen. I hope beta2 is just as stable

EDIT: So my own 330W PSU just arrived in my arms. Total cost: 48$, paid at the moment of receiving the package. Sweet sweet capitalism.

 

im still getting crashes from the new beta driver. and i also flashed the .975v vbios.

 

Have you tried to run AMD Uninstaller, and then fresh install the first 13.8 beta?

 

Hello fellow m17x enthusiasts,

I recently sold my old m15x.
And purchased an m17x r2.
I instantly upgraded the single 4870m to the 7970m.
I'm currently in the progress of setting up a crossfire configuration (second 7970m is shipped)
Already purchased a heatsink and fan.
Last thing I need is a crossfire cable.
Is this the correct part number?
b2725050g00004. (Found on eBay)
Greetings

 

Yea, that's the one.

 

Exactly what Sangermaru said. Thesd cards should work at 80-85 max temps in hot environments (ambient temps of 30c). . If it gets over 85 it shuts down as did my cards because of overvolting 1.1 V

Also guys sorry I didn't post recently but I was working on clubs near the beach so I didn't have my Alienware with me.


Ps. I will gladly buy a moded psu from you Sangermaru! If you manage to keep the second one you received and are willing to mod it for me. although I am really ok without a 330w psu. I never have a psu shutdown whatever I throw at these cards and to my 920 mx... So looks like I am really lucky with my psu. And it doesn't even buzz... Like yours does

 

Welcome back bro. We'll see how it goes with the PSU's. At the moment I was trying to take the 330W psu for a trial run in Stealth mode (since stealth mode can be set to stock clocks lol ).
Unfortunately, light on the brick turns on, but the light on the connector doesn't... and when I plug it in the laptop takes no notice of it... ALLMYHATE.

My techie friend won't be back around until the 2nd of september... My luck is terribad.

 

I think you should re-install your cards before anything else. If the crashes are due to the temps hitting 90's you should fix the root cause as software can only do so much.

Congrats on the R2! You're going to love this machine

Welcome back man I actually am returning the second PSU. Well to be more accurate I'm not going to fetch it when it arrives so it gets returned to the sender.

That's weird... there's no light on the 330W connector though, only on the brick so that's normal. But the laptop should work with the 330W plugged in. It should beep on boot saying an unsupported adapter is connected.. unless you disabled the warning in the BIOS.

Just to be certain that the charger is working pull the battery and run the laptop with only the 330W plugged in. It will run slower as it doesn't operate at full power with unsupported adapters but you should be able to accomplish light weight tasks.

That's what I did to check and I was running in normal mode. I noticed that the CPU didn't throttle its use but the system did feel slower.

 

Ah, no more cable light? Good to know, that worried me.
Well, the thing is, I'm not getting that windows "Plugged in, not charging" message with this adapter. I'll try it more during my work break, to see if it boots without battery in.

 

I think your adapter is fine then Mine also didn't give me that message but it worked without battery. I think that ID chip or lack of, screws up more things than we realise

I'm glad the light on the connector is gone though. It clashes with the AlienFX lights if you set anything other than blue

 

Ok, so my PSU works just fine My problem now is that I don't remember what I did before to force the 2D clocks (not the idle clocks) higher from 400/300/0.900. If I could do that I could probably run this PSU full force without modding it...

 

How did you manage to trigger that state??? I've been trying everything to get another state to trigger but my cards only do normal 2D (idle) and 3D. Does it trigger when using the 330W and you load the GPU's?

If you manage to make your cards work at the higher clocks you still have to ground the ID line on the PSU somehow for it to unlock the full power.

 

I'm sure if we could edit the Performance states, we wouldn't need to mod the PSU at all. You can view them all using VBE7, but you can't modify them. There's many more performance states than just idle and load clocks The 400/300/0.900 P-state is for multimedia applications (UVD state).

 

Yeah that what I specifically looking for when I edited the voltages from 0.975v to 1.00v on my "forced" vBIOS. I was a bit concerned that if for example when playing flash video another state would trigger and the voltage would jump to 0.975v with different clocks and if that were the case I've edited that entry to 1.05v. So it would run hotter which is unnecessary when playing video.

But I couldn't get that state to trigger, even when playing videos on YouTube which apparently is a common trigger. My cards only use 2D (idle) and 3D states. I remember getting a 450Mhz core clock once but that was throttling going on when I was testing an extreme overclock.

 

Like I said, download VBE7 and open the vbios you're using if you want to view all the P-states available, although I have absolutely no idea what exactly triggers each one

 

Dude!!! I thought you meant RBE and VBE7 was a typo! Checking it out now. Thanks!

 

I reinstalled the card and it runs cooler now. It fixed the crashes too. When I ran evga precision, the voltage is 1.050v. Last time I flashed the vbios. I used .975v

Sent from my DROID RAZR using Tapatalk 4

 

Did you find the source of the problem? Misplaced thermal pads or inadequate thermal paste. Glad you solved the issue. It can be really frustrating. I was having heat related shutdown issues while running the 1.075 overvolt. It was unnecessary for regular use and gaming. I re-flashed down to the stock vbios have not run into any issues while gaming (mostly BF3). Still waiting for my other card to arrive, but hopefully it should be here today.

 

That's great news pinoy_92 If I recall correctly you're running a Clevo card? Which undervolted vBIOS did you use?

Exciting times batfuse Can't wait for you to have it up and running!

 

all i did was reinstall the gpu like nospheratu said and spray some of the dust off. the crashes i was experiencing wasnt heat related, or at least i dont think so. only the games are crashing, not the whole laptop.

yes i am using a clevo card.

when i put everything back together and booted it up, the voltage level on the card is saying 1.050v. last time i flashed the vbios i used an undervolted vbios, .975v. its really wierd. anyways even with the stock voltage my gpu doesnt get hotter than 85 degrees.

 

I am so PISSED. My new heatsink has near-identical temps to my old one. Same core disparity. My 3dmark11 score increased by 30 points.
HOW are you guys keeping your CPU's cool? What pastes do you use? Any special pressure tricks? I don't get it.

 

Damn... I'm really sorry to hear that man. I was looking forward to your results

I'm not using the retention mod although I am considering doing it to run a higher TDP 24/7. Its just the standard heatsink install currently. I'm using Shin Esu X23D on my CPU atm but I don't think its the paste issue. Even bog standard paste should allow you to run default Dell 62W TDP.

Try testing the heatsinks in boiling water. Dip the GPU die contact end in boiling hot water and then immediately check the opposite end of the heatpipe that contacts the fins. There should be no delay in heat transfer. If it takes a while to heat up the heatpipes are probably damaged and the gas/liquid inside them have evaporated.

Also try checking if the heatsinks are warped, other than that I'm not sure what else you can test

Bear in mind that my XM hits low 90c when I run it at 25x on all cores with 90 TDP. I only use that setting for benches though. I game at stock Dell settings with my 240W.

edit: Also try placing the heatsink on and screwing the screws in the numbered order but only one 360 degree turn at a time (or even less try 45 degree turns). So turn screw 1 360, then screw 2 360, then screw 3 360, then screw 4 360, then screw 1 360.....

Also check that the bit of thermal pad on the heatsink for the inductors on the motherboard is not interfering with the CPU die contact.

 

Well, it's much better than before in that it appears I can now run 25/25/25/25 multis again with about 75W TDP to stay in the low 90's... but my 3dmark11 score is, as romanians would say it, varza. Virtually identical to before.
And it doesn't hold clocks very well, even at 85W TDP it will only be able to sustain about 23.4 multi out of 25. I'll recheck that thermal pad (replaced it with a phobya one)... the screwing process is the same one I'm using, pretty much... it pisses me off.

 

I normally use ThrottleStop's built TS Bench 32M test to test how well the XM holds the multi's at a given TDP. 85W TDP is sufficient for 24x across all cores but barely. The multi's will drop on the 1024M test.

Also TS Bench isn't as stressful as wPrime. Multi's will drop on wPrime even though it was stable on TS Bench. This is just from my experience though. I think it also also depends on the quality of the chip.

 

Actually, from what I've been able to observe so far, the better the cooling, the lesser the TDP required to reach a certain frequency. My old 920xm on my Medion Erazer could pull off 25/25/25/25 at a TDP of 75, full multis during TS 1024M test, max 95C. I'm sure chip quality does count for something, but the better you cool it, the lower TDP required.

 

Any good news @sangemaru?

 

Well, it'll be a week + 3 days until i can enable the 330W PSU full-force. And I'll also get some Tuniq TX-4 and try the retention clip mod on the CPU heatsink. We'll see then.
I suppose it's usable for now.

 

hallo guys. i just recovered from a crazy weekend djing on top of a mountain....

anyways.

i cant understand the problems sangermaru has.
look at that



i just tried it. ambient temps are 30C , even at ~70% in the 1024M test the temps domnt reach 90C
Also 24X is super stable at anything even gaming etc. At 25x multi in all cores i get random BSOD (i will soon try to see if i can solve this somehow although 26/25/24/24 multies 24/7 are super for me! ) . my tdp is 90

also as i said even with 24X 4 core multies and my gpus mining (running at 99%) or heavy gaming i am rock stable.

So sangermaru your temps seem liitle high but i dont know why. whats your ambient temp? Also i use arctic silver 5 . when i tried a generic thermal paste my CPU temps reached 100C+ so pastes do make a difference when they are noname

what do you mean multies drop?? Havent you checked the "clock modulation" and "chipset clock mod" on throttlestop? whith these checked the cpu is locked on max when the cores are used. for me whenever i run bences the multies stay locked on max!

 

I totally hate you, Alienhack
Regarding multis dropping, that's because the stress of the test at that clock generates more heat than dictated by the TDP. Increase TDP limit.

 

I hate you too! One of my cores hits 95C 20% into the 1024M test

The Clock Modulation and Chipset clock mod tickbox's are for machines that implement throttling using clock modulation. It can be seen when running our machines in Stealth mode or if we run an unsupported PSU. Generally if we are not in those modes our machines throttle according to TDP.

Your temps are incredible man. If you look at your max temps in TS two of your cores hit 93C and 91C but considering they are only there after 68% of the test is done that's a really good result :thumbsup:

 

.

they didnt get to 93C at 67% , they hit these temps when the test started until the fans started spinning up. after that the temps are constantly what you see until the end of the test, sub 90C
and just think that the ambient temp is 30C , at winter time when it drops to 17-20C i think they will drop even more.

I have these boxes in throttlestop ticked and i never see multies droping. maybe you should use them too. i knew what they do, they disable throttling , so i supose thats whats needed for multies not to drop.
when i used the noname thermal paste and hit 100C thats when i noticed thermal throttling, the multies would drop when a core reached 100C

But when i checked these ticks in throttlestop the multies were stable even at 102C
as i said my TDP is 90, so i dont know if that is TOO high or could be raised even more

 

Hey guys. Got my other card today. Long story short, it was actually my original card (a Dell version ) but has a reprogrammed vbios. I had a bad flash via atiflash in DOS a while back. Bad luck I guess. Anyway, I flashed it with .975 V; it boots fine. Although, I get a "This device cannot start" message in device manager. I have tried 3 different dell vbioses thus far to no avail. I read the first post stating to keep trying different vbios. WIll device manager eventually say it is compatible? Is it certain that if I install drivers I will get a freeze on boot?

I currently have it installed in the 2nd slot. Should I continue trying differrent vbios? I have still have not tried it in the primary slot. What would you guys advise?

Edit: the card was reprogrammed with a clevo vbios. Should I try Clevo vbios also?

 

Okay now I'm really jealous... 30C ambient and those temps?!?! Wow AlienHack that's an excellent paste/install job you've done there.

Hmm... that's strange about the multi's not dropping. Even with it ticked on my side multi's drop if my TDP is exceeded.

The only thing stopping us from setting a TDP as high as desktop chips are temps. So if your temps are good I don't see why you cant raise them if you don't mind the temps the cores are reaching. Personally I try to stay below 90C.

Congrats on the second card!!

I also get the same error message on either one of my cards when no drivers are installed. Once the Catalysts are installed everything works as it should with no error messages.

sangemaru experienced the exact same thing if I am not mistaken. Try the latest 13.8 driver and see if it works out for you. If you do get a freeze on Windows load, Boot into safemode and uninstall the driver.

It's strange that it shipped with the Clevo vBIOS though. Unless the seller knows this revision of card has issues with Dell vBIOS's on the R2 which isn't the most ideal of situations.

 

Thanks, that's good news. I will give it a shot when I get back from work.

 

Yea, I experienced the same thing with the caveat that that second card never could start, for whatever reason. The replacement card I have now works great. Good luck batfuse
Are you using a single card or trying for crossfire?

 

Great news. 7970m crossfire config worked. I installed the 13.8 beta 2 and did a prelim benchmark on stock clocks with 3dmark11. Looks like its on par, maybe slightly below, what you guys have achieved. I'm super excited. Im currently using .975 V on both cards with the stock 240 W PSU. Temps on both cards while benching were 65 and 67 for GPU 0 and 1 respectively. I'm going to see how it does with a few games for stability, temps, and micro-stutter, as well as play with clocks to get a top stable bench.

One question I do have is regarding Regedit and EnableULPS. I only set "0" to EnableULPS only and not EnableULPS_NA. Should I also set 0 to EnableULPS_NA as well?

Thanks goes to you guys Sangemaru, AlienHack, and Nospheratu for taking the time to work out the kinks to get this mod to work in this beastly machine.

 

Excellent news batfuse!

For stock CPU/GPU's the results are perfect :thumbsup: Temps are exactly like my cards too.

You need only set the "EnableULPS" value to 0. Ignore the "EnableULPS_NA" key. Congratulations man, I'm really happy everything worked out

 

My advice is to NOT ignore EnableULPS_NA. I noticed that if i didn't change that value as well, resuming from hibernate/sleep would still lock my clocks.
Also, that score is quite good. To get it higher you need to really push the CPU into the 85+TDP, 24+ multi range with throttlestop and hope your cooling holds.

 

Thanks for that info sangemaru. If I remember correctly you use sleep/hibernate instead of shutdown to keep HWinfo fan control running? batfuse has Dell card's as well so I think he may prefer your method also. I'm always on high performance mode and the laptop is either on or off so I haven't experienced anything like that.

I hope AMD fixes these ULPS issues for CrossFire users though. Even desktop configs have issues with this buggy ULPS implementation. If they can manage to fix the micro stutter I'm sure they can manage it. Or at the very least give us an option to turn it off in CCC. One good thing about these 13.8's are that the second GPU running at constant load is a known issue. They don't provide a fix but we know that disabling ULPS fixes it. So hopefully the ULPS issues are resolved when fixing this bug and in the process it fixes our BSOD on alt-tab as well!

In other news the "donor" power cable should be here this week so I can finally mod the 330W

 

Thanks for the advice Sangemaru. So I tried BF3 and it played well with temps in the mid 60s. I did get a black screen/shut down after playing for 5-10 minutes which I attribute to insufficient power. The odd thing is that my power supply appeared to not power itself; the blue light was off and it would not charge the laptop. On another attempt, I found my CPU was throttling to 1000mhz after several minutes of BF3. I looked at the power supply and again it was "off" and not powered, switching the laptop to power saving mode on battery. This happened with the cpu multi on 24 and another time w/ multi at 21; not using throttle stop. Any one had this issue?

My temps for the Gpus and CPU are within normal limits (60s gpu/70-75 Cpu). This PSU is actually rather new, purchased in april this year. Although my battery is worn, with HWinfo showing ~30% wear last time I checked.

I am suspecting my power supply may be less fortunate than the ones you have. I do have my original PSU, and will give it a shot.

 

If you aren't using ThrottleStop that means your TDP is set to 62W which is exactly what I'm using for gaming with no shutdowns. If you aren't using ThrottleStop what are you using to change multi's?

It could well be your PSU not being able to deliver above 240W which is causing the problems you're experiencing. I haven't played BF3 on my machine so don't know what the power draw with that game is like. I know Crysis 3 draws up to +-275W and Bioshock Infinite is almost always around +-220W.

From your description it seems that the PSU is tripping as it can't supply the power requested by the laptop but it isn't pulling your entire machine down with it, which to be honest is better than a total blackout. Can you confirm that your cards are running at 0.975v in 3D mode with GPU-Z? A quick and easy way to do this is run the built in GPU load test in GPU-Z in fullscreen mode (the little icon next to the PCI-E lane speed field) and then check the vddc values on the sensor tab.

 

guys could u inform me on this ULPS issues-bug?

what problem does it create?

welcome to the heavy weight m17 club mate!! -) with 10.000 3dmark11 points that your setup can reach there Is no game to be afraid of

 

Use ASUS GPU Tweak to underclock your cards until you do the PSU mod... because you will need to do that mod, apparently. Use throttlestop to lower TDC to ~47, and limit the max consumption of your CPU (I think it's at 62 by default).

My PSU can't max out without shutting down either. I use 800/1100 clocks at the moment and it seems to be able to carry that.

Problems with stuck idle or load clocks on either card, easily triggered by sleep/hibernate, instability and bsod's, you name it.

Downside is that with ULPS disabled, your second card will no longer power off when unused... which as far as i'm concerned is better, since shutting it down interferes with HwInfo fan control and monitoring, and causes other issues such as crossfire not running even enabled.

 

Its almost surelly your psu. try some else, take out your battery use your laptop only with your psu. run bf3 and check your temps of you cpu not going over 100 and of yoir cards not over 85. If they dont and you still get shutdowns with the psu led turning off and then needing to take the psu off the wall output and back In for It to work then you have a low output psu.

also try playing bf3 with the cpu not overclocked. see If thats ok with your psu.

also try using throttlestop. Its a great app. use It to raise your tdp to 85-90 . also click on the throttling options and use It to manage your cpu muktipliers.
with throttle running you dont need any other program to setup your cpu

hmmmmm . that solves the mystery why my second card stays stuck at full 3d speed even after I exit a game. I was trying to find out why and thought It to be a vbios problem. maybe I should try enabling ulps and see If It fixes this. I had no problem with It on. no bsod no stuck speeds...

 

Double check if you disabled all EnableULPS entries. I've only experienced this issue with the 13.8's. Check the known issues change list here, they mention the secondary GPU running at 99% load although I still experience it with OverDrive disabled.

 

@Nospheratu: GPUz doesnt have VDDC sensor in the tab for either card. Weird? Hwinfo reports VDDC of .975V for both cards. Is not accurate? Also, I am using EVGA E LEET to tweak the CPU.

@AlienHack: Yeah, my temps look good. I did a log with Hwinfo and reviewed it after shutdown and power supply giving out. I am afraid it looks like the PSU is giving out. I dont have a meter to measure what my PSU is putting out. I will have to ask a friend this weekend. I will try using Throttle stop.

@Sangemaru: I will give ASUS GPU tweak a shot as well.

It looks like the PSU is not holding up. I still need to try the other PSU I have. I am concerned my electrical engineering skills aren't up to par to successfully complete the 330w mod. I have read over the post and update by Imsolidstate. I just have not cracked open a brick myself and am not too familiar with it. But it appears to be the direction I am heading. I am happy to see that my temps are otherwise holding up.

 

I've noticed the same issue with GPU-Z depending on the vBIOS I have loaded. AlienHacks vBIOS doesn't give me a vddc reading in GPU-Z but HWinfo does. Since you're running the Clevo vBIOS on the Dell cards that could be causing it.

I wouldn't worry too much about it since from what I understand HWinfo reads the values provided by the AMD's ADL library which is accurate. GPU-Z just simplifies the process since HWinfo logs a lot of different sensor info and could cause stutter ingame. You could disable most of the sensor's that you don't need.

I would throw EVGA ELEET in the bin though lol As AlienHack mentioned ThrottleStop is your CPU's best bet. You could use it for only monitoring temps but the functionality built specifically for our CPU's make it the best tool available for our machines.

 

LOL, will do. Will switch over to TS now. I am currently using the .975V vbios from SVL7...thus dell.

As an update. I used my other PSU. played 15 mins with no shutdowns. I also noticed CPU clocks stayed constant as well. I had seen them flucatuate quite a bit with the PSU that gave me the shut downs. I just wanted to see what would happen before I started throwing other variables in. I try the recommendations you guys suggested.

 

Awesome news!

Strange about the Dell vBIOS not reading the vddc sensor, I thought you were experiencing problems with the Dell vBIOS on your new card. I also experienced the same no vddc reading when I first tried svl7's vBIOS but after a few reboots/driver installs it showed up. I'm not quite sure how or if it was the particular driver version I was using.