M17x R2 7970m CrossfireX Ultimate Installation and Tweaking Guide

Question I have because I have a Intel 720qm overclock. Should I have the power to run the dual ati 7970. I been debating if I wanted to upgrade the whole system or buy graphics cards.

No I did not read the whole thing.

 

I'm pretty sure radeonPro helps immensely.

Chip detection is based on core and frequency. Depending on the overclock level our CPU's will be reported as i7-920, i7-940, i7-950 etc. since the 920xm are the same chips as the desktop 920's.

That CPU should hold its own pretty decently in non-CPU-focused games. Worst come to worst, you can update later, but you should still be doing pretty well with it.

 

Okay some test results for throttling. I did the same run i.e GPU cores at 950Mhz and XM at 25x on all cores 90 TDP. Score dropped but I'm not sure if its cos I'm running Cat 13.4's now or GPU-Z was logging in the background. I reverted to 13.1 since I received BSOD's again with the 13.6's when trying to change resolution in Crysis 3. The 13.1's worked fine so I upgraded to 13.4's without CCC which seems to be working well... so far. And yes sangemaru, CAPS are installed

GPU-Z graphs for the 3Dmark11 run


The GPU's don't seem to be throttling but the second card went a bit crazy during the physics test. If you look at the GPU load there's some activity but the first card isn't affected as much. I cut out some idle time and the Date/Time column to make the graph look cleaner. The original logs are below if you want to check them out, maybe there's something I missed.

Also I repasted my GPU's again and added some thermal padding to the VRM components to the best of my ability. Its still not 100% though. I need some thicker pads for the mosfets. After I was done with the first card I installed it, booted the system and ran Farcry on medium settings. FPS was around 50-60FPS and the micro stutter was still there with a single card I think the game is just messed up. I don't have any other games where I notice micro stutter atm though. I still have to test with RadeonPro, I haven't tried everything as yet.

@AlienHack AMD is releasing new drivers that virtually eliminate micro stutter at the end of July. See demo of it here.

I actually need to get a Killawatt or one of its clones to test how much power Buzz is giving me.

edit: Forgot to add the 3Dmark11 run! It's here.

 

Oh wow that prototype driver video looks awesome, but I must admit that I couldn't really tell the difference until the video speed was lowered.
Combined score seems slightly low, but not by much. I don't expect you to be able to score more than 6100 combined with this CPU anyway. It's all synthetic so it doesn't really matter in the long run.

 

thanks for this OP

 

The new drivers you mentioned is great news! I hope they DO work!! This way I will use my second card in crossfire and not let her sit there idling.



Nospheratu when you mention microshuttering with crysis with one card maybe you mean the shutter you get with vsync off. Try shuting down crossfire , run crysis with one card and turn vsync on. Tell us if you still experience microshuttering

 

There's no micro stutter with Crysis. The game has motion blur anyway so I don't notice anything at all. Farcry 3 is the problem and it isn't the screen tearing from having vsync off. I always run vsync on in my games if the FPS is 50+. If the FPS is lower I run dynamic vsync with RadeonPro capped at 59fps. I think that FC3 is just poorly optimised. The method I use to look for micro stutter is moving forward slowly (walking or crouched) and look at the ground immediately in front of me.

13.1 seems to be the driver that doesn't give me any BSOD's. 13.4, 13.5 and 13.6 are almost a guaranteed BSOD on loosing fullscreen. I'm using Crysis 3 change resolution as my test atm.

 

farcy has many issues graphicaly with many cards, either ati or nvidia.
with my 680gtx it was playing at 60 fps and then suddenly droped to 30 and 20 and was shutter all the way... without any reason and without any move from my part.

i think i solved it with driver updates. nvidia fixed it, i dont know if ati did though....

why dont you try farcry blood dragon?

 

by the way i cant make up whats going on with the TITAN results...

3DMark - test your graphics card with the gamer's benchmark GeForce GTX Titan

its a mess.. some people got 8000 score others 15000....

 

What I am wondering because my cpu draw less power then the Intel 920

Intel 720 power 45
Intel 920 power 55

Note my cpu overclock 5%.

 

Some people seem to have issues when running 2x 7970m crossfire. Nospheratu in this forum seems he has no prob. So you should wait for my card to arrive this week so we can see if in general the r2 can work with 2 7970s in crossfire

 

I honestly doubt that there's any legitimate way one can power this setup with the 240w power brick.

If the components aren't all being used to 100%, which has been the case when I tried the dual 6970M route, then yes I can understand that...

However I don't see how this is feasible WITH an overclock on the CPU too.

 

Theoretically I couldn't agree with you more. I don't understand how my setup is working. I even ran Crysis 3 on High with a 25x across all cores oc on my XM yesterday and logged GPU clocks with GPU-Z... no clock drops at all I'm not complaining though

I am running one GPU at 0.975v and the other at 1.000v... still, its crazy.

 

Am I the only one not surprised at all? Why do you think I started this thread, because there was no reason for this setup to not work :thumbsup: And nospheratu proved it. Still, even I didn't expect he'd be able to OC the CPU to that extent while still keeping stock clocks, but one thing should be obvious from this:

1. the most important thing is the PSU quality and
2. any m17x-r2 user giving up his machine is a sucker

Consider the following: At the time when techreport (i believe) measured power draw on the 7970m (AT THE WALL - estimates), they were running 1.05V cards, as that was the default value at the time. Newer revision 1.025V cards are more efficient. Dropping the voltage even lower will bring power consumption WAY down.

I'm sure Nospheratu's PSU is struggling at these settings though It's entirely possible that after a few months at this usage level, the PSU would not be able to keep up and would start supplying less and less, but there's a few factors here:

1. you can replace the PSU.
2. you can replace the PSU with a more powerful one (clevo 300W recommended).
3. you can undervolt/slightly underclock the cards and limit the CPU's TDC.

Either way, it remains for alienhack and me to confirm nospheratu's observations. But basically? I'll be keeping this machine for a long long time.

 

i will use this machine to play star citizen smoothly in the end of 2014.

by the way as you know my card is stable at 0.975V and oces to 940mhz with that volts. looks like it is a good card for crossfire...

 

Wow. Yes, that sounds real good. I suppose you'll keep it then. When will the second card arrive?

EDIT: Omg, Star Citizen is being developed by Roberts' CIG? I can't wait I loved all their games. Haven't had a good spacesim in a while, and a mmo too. Wonder if they'll offer Oculus Rift support.

EDIT2: Oh yeaaa http://oculusrift-blog.com/star-citizen-oculus-ready/270/

 

Ofcourse they will. I have already spent 700usd on that game. Its gonna change the gaming world forever

 

CryEngine 3 space-sim VR Dogfighting while on the go with your machine nearby and the rift on your head I can't picture a more pleasant train trip.

 

so if the 240w psu can power xfire 7970m and an oc cpu. can we use a 330w psu?

 

We've yet to play around with the 330W mods, but the evidence points toward the Dell 330W suffering some kind of incompatibility with anything except m18x's (since non-dell users have also had power-related shutdowns on that PSU). Meanwhile 240W psu's can handle more.

If you can locate a more reliable 300W PSU (such as the one used with Clevo X7200 machines) and perform the Chip ID mod at the motherboard level (read through the thread for more information), then you should be able to really push those cards and CPU.

 

sangemaru!!!!! You jinxed my no shutdown run!!!

Loaded up Crysis 3 and as soon as I entered a level instant blackout... everything off including PSU indicator light The only good news is I booted up again and played for another 30mins I don't think its related as I didn't have my battery in and the PSU did do this previously with my 5870M's.... hope its not related, cross fingers

On an unrelated and purely superficial note... windows performance index rates my CPU and RAM the lowest subscore of 7.5

 

So did you change any settings/clocks/volts (radeonpro stuff)? If it's a one-off thing it may be that the PSU hadn't had time to warm up properly and you shocked it, or your PSU is having trouble keeping up.
My CPU and RAM have 7.4 each, and I'm running OC, so you're golden there (maybe too golden, *hint*).
I still think you should aim for lowering consumption as much as possible on that setup, since you don't want to ruin a good thing by burning the PSU

 

Yeah I cheated a bit for the WEI run.. 90 TDP and Dell's default 24/24/24/24. For games I'm running the machines default TDP which is 62. Not exactly in line with Intel's specs...

Its a bit disappointing that I'm stuck on these 13.1 cats though. The latest CAPS do help but I feel as if I'm missing on on important optimizations. The only thing I haven't tried as yet for the BSOD's is swapping the cards around. I noticed in between AMD driver installs when the standard vga drivers are loaded, my primary card has the "This device cannot start. (Code 10)" error... strange. Although as soon as I load the AMD drivers everything works as normal.

Yeah I definitely don't want to ruin this PSU. Who knows if the next one is as good.

edit: I've also managed to flash the vBIOS to the Dell 0.975v and both are reflecting correctly now. I'm not sure why it didn't reflect last time. So both cards are on the same voltage and the BSOD's continue...

 

the BSOD continue only when changinf resolution in crysis 3?

 

Exactly... my idea of how life and vacations should really be... in a beach under the sun, drinking a corona and playing star citizen with oculus rift...

 

No anything that forces the cards to exit fullscreen to desktop and/or back. Alt-tab also triggers it. I'm assuming it has something to do with the cards switching from 3D-2D clocks. 3Dmark isnt as clear cut though since it falls back to the desktop after each test and I only had a BSOD once with that. Its not guaranteed, sometimes it works randomly but I've never experienced it using the 13.1's.

 

@Nospheratu: Can you try disabling vSync in all drivers/game options and try alt-tabbing from fullscreen?

@Alienhack: I can't wait.

 

I tried the no vsync and alt-tabbing with the newer drivers and it worked perfectly though multiple alt-tabs... then I enabled vsync and it still worked. Crysis 3 still BSOD on res change though. I also managed to get a BSOD on FC3 as well while playing with the video settings... on 13.1 Although the BSOD on 13.1 was different.

So it seems there's no hard and fast rule. I might have to resort to extreme measures like disabling TDR and setting 2D clocks/voltage to 3D like in Mr. Fox's post here.

 

If it's a random thing which looks likely then psu is my hypothesis as to the cause of BSOD.

 

I am eliminating PSU from my personal hypothesis. If vSync disabled yielded positive results, I would attempt to disable vSync in all options other than RadeonPro and forcing it from there, and using the Catalyst 12.6 beta.
There are known issues using vSync in CrossfireX, on desktop as well as mobile cards. One other thing you might want to try is to increase voltage to 1V for both cards and attempt alt-tabs, resolution changes, etc. You may be able to do this in software, without having to reflash bios-es.

 

I wonder what these monkeys at AMD did with the drivers after 13.1... the change logs never reflects all changes. I can game fine with the newer drivers it seems. I just shouldn't change anything that causes the cards/screen to switch modes on the fly.

My suspicion is the catalysts buggy implementation of power management that gives us crossfire users a headache.

 

Have you tried disabling ULPS?

Personally I only want to game in Borderless Fullscreen mode. It saves a lot of headaches and it allows the use of the RGBLED screen while calibrated.
I have to say, the 13.1 drivers I didn't like at all, but most betas run great for me. Still, it's never sunshine in driver-land.

I love seeing new nvidia drivers posted, with thousands of people complaining about crashes/bsods/shutdowns/throttle etc. It brings the overall fanboilvl down.

 

Lol yeah but if considering the ridiculous amount of BSOD searches I've done for AMD drivers I've realised its fairly even across both camps. That said I haven't read anything recently about any AMD drivers apparently killing users cards like the debacle with a recent set of nVIDIA drivers

I've had two system freezes when starting up HWinfo which I'm assuming cos it tried to query the 2nd card which was in power save mode so I've disabled ULPS. I'm also running both cards on 0.975v atm.

I wish we had the option of SLi with new cards though. This SLi certification is complete bollocks honestly.

Just FYI, I've noticed dropping shadow settings on Crysis 3 and FC3 to low nets me around a +-10fps gain. Also I get the least stutter experience in FC3 running vsync (any flavour) from RadeonPro and disabling it in game.

 

Lolyea, I swore not to touch nVidia cards after I had some that died on me... I have friends working as repairmen for IT equipment not touching nVidia chips except for personal favors.
Anyway, enough of my own brand of fanboyism

Crysys 3 shadow settings yield so much gaing because shadows are CPU-rendered, and Crysis already puts a massive amount of load on our CPU's. When I was playing on Very High settings with FXAA activated (2x), I noticed that while the GPU was pushing 77C, it didn't seem to be fully utilized, and performance increased the more I pushed the CPU.

So the best way to optimize Crysys for our machines is to lower CPU-scaled effects while pushing GPU-heavy operations to the maximum.

FC3 vsync issues are known, and I believe not just with AMD. It's a good thing RadeonPro does so much for us. I'm curious how the prototype driver will perform.

 

well i am no fanboy, neither nvidias nor atis.

over the years i bought cards from each company , it depended on the price and how good a card was.

for example in my desktop i have a gtx 680 cause its one of the fastest and i hapened to find it in a great price.

BUT for my alienware i wouldnt touch nvidia cards as they are EXTREMELY INSAINLY overpriced.


By the way from years of using catalyst and nvidia drivers, the feeling i have is that usually nvidia ones are more robust and easier.


PS. concerning problems with games and cards from either nvidia or ati, well its really random. there were times a game had problems with only nvidia cards and other times another game had problems with specific ati ones. its really random.

 

Very true, well at least from my experiences. Only ridiculous problem I had was with the nForce3 chipset and Geforce 6800 incompatibility back in the day.

I got some interesting results with testing shadows on Crysis 3. I ran everything on high settings except for shadows on low with my CPU at 20/22/23/24 with a TDP of 65 and loaded up a save that loads at quite an intense scene in the game (Looking out the window into the rain where Psycho gives you the bow). FPS was 47. Then I set everything on High including shadows with my CPU at 24/24/24/24 with a TDP of 85. Loaded the same save and fps was also 47. I never move my mouse at all to keep the test as accurate as possible.

So it seems that we do need more power from these chips to make full use of the power of these cards... at least for Crysis 3. Honestly though, accuracy of shadows isn't something I find I'm missing out on so I'm happy with the compromise.

Another thing worth mentioning is I booted up after coming from work to what looked like a black screen with brightness turned on full... it looked like extreme light bleed from the edges. I thought it might be the white screen that other members posted about that killed their RGB and immediately pressed the power button. The machine went into sleep mode. Pressed it again and it woke up to my Windows desktop and everything was normal. So I decided enough with the modded Dell vBIOS I'm going back to stock. Rebooted and got that same screen again. Pulled power, battery was already out. Plugged back in and rebooted and everything was fine. Immediately booted from USB and flashed my cards back to stock. Hopefully I don't get this again.

 

well i cant understand from where you get the idea that the cpu is underferforming. You changed your CPU multi (thus speed) but you saw no change in FPS . that means that the game is not bottlenecked by the cpu. The cards do their best and manage 47fps.

just be carefull ... we dont want you to damage your rgb. you give us so great info. So remind me please what vbios did your card have and what do you use now that the white screens started appearing?

 

Maybe my explanation was a bit convoluted. Basically I'm saying stock(ish) CPU with shadows on Low gives me the same framerate as OC'd CPU with shadows on High.

If I run the stock(ish) CPU with shadows on High, I drops about 7fps.

The undervolted 0.975v Dell vBIOS (017) used from SVL7's mod pack gave me the screen. I am running Clevo cards so that may be where the problem lies. Now I've flashed back to the stock P270WM vBIOS that came with my card.

 

I sometimes get the black screen with backlight bleed (and not just on this machine either) when there's driver issues with sleep/hibernate mode. Were you resuming or cold-booting?

 

Cold boot. I never use sleep or hibernate mode on the laptop... ever.

 

Yea, sleep and hibernate don't have a place on gaming mod machines They almost never work, although hibernate is not bad, since the computer actually DOES turn off. I find it pretty useful, and I don't have to force close every app because Chrome is a piece of trash that crashes every time.
Hmmm, damn. Quite a few software-side CFX issues. I guess we'll see once I get my second card. Do m18x users have these kind of hiccups? I don't think your issues (except that one turn-off) are power related.

 

Hmmm interesting. As I too run the clevo 7970m have you used the eurcom vbios from rsgeiger? It has worked great for me since I got the card. Maybe these other bios's have issues? Just speculation really.
Curious.

 

I think they do, well at least on the M18X R1. See Mr. Fox's post here and his workarounds.

Yes that's the vBIOS that came standard on my cards. Its actually the same vBIOS that everyone calls the RJTECH vBIOS. It was made for the 7970M that was to be used in the Clevo P270WM.

I think you may be right and that's the same reason why we're apparently able to get HDMi working with the P270WM vBIOS. The HDMi and our screen connections may be swapped around on the Dell vBIOS so it works well on Dell cards but not Clevos. I'm not sure though, either way I'm not going back to the Dell vBIOS.

 

what about my bios? the strange one? anyone wants me to upload it so he can test it?

 

Uploaded, definitely Should be worth some experiments.

 

buy the way, as i already beleived, DHL sucks big time.
They brought my card in the morning to my house but they say that they didnt find me although i was here drinking my morning coffee.

AND ON TOP OF that when i asked them how i could go myself and get it they said "sorry there is no way" you will have to wait till tomorrow and we will again bring it to you.
?? i cant go myself and pick a parcel up? i mean i am saving their time!!

 

Damn sorry to hear that man. I would have expected more from DHL, they're behaving like a local Post Office service. Your parcel is most likely lying in the local depo. You should be able to go fetch it, I think you just happened to speak to a clueless support center agent. Try calling again and hope you get someone competent.

 

Heh, that even happened to me with UPS and FedEx, in my country they suck, I have to walk them through procedures on the phone, so many stuck parcels.... I swear, if one good shipping business popped up they'd control the world.

 

lets hope that in the morning (in 10 hours) i will get it! and try it!

 

new puppy!!!
blue pcb