M17x R2 7970m CrossfireX Ultimate Installation and Tweaking Guide

I will.
Got quite aggrevated today when nothing i did made the computer even start up past initializing the screen.

Going to try to get a new buttonpad with controls, nothing works but the power button.

Also one thing ive discovered. During startup my comuter uses quite a bit of time on first sequence of the boot when running fans etc when having two GPU's.
When using one it starts straight away.

I even pulled the CMOS battery to check it and swapped PSU's back and forth.
I started to make it work more often with the 240 than the 330, but it could be just a coincident.

 

Just did a new CPU test. This time i shall have good overhead on power since i have single GPU now.
just did Physics Score 5676 and Combined Score 5520

105 TDP and 75 TDA 25 times ratio

 

trome 71, its sad to see you experience so many problems...

as the guys asked, is your laptop working great when you have only one card seated in ? take the other out tottaly and try to see if everything is working ok

 

On the positive side, playing with BIOS OC gives me more than changing throtlestop multies.
As long as i can set TPD and TDA in TS I can get above 6050 using 25X and 5% OC in BIOS, 27 with stock BIOS gives me about 5900 tops. Dont see much change by alternating the multies actually. Something is fishy.

Working like a charm now with one card.
Will try swapping after all updates perhaps tomorrow

 

try with the other card only . if it has problem only with the second card then its a card problem , if each one works alone but when both instaled dont work its a cable -crossfire problem

 

Oh dear, here it goes again, CPU drama

.... your CPU score is so low, you scored only 5900 when i Scored 6050

We are talking here --->>> 2.5%

so about 0% performance gain in games

GG

 

Nah this was positive for me.
I only need the current and power increase of TS it seems.

My CPU gives more than 15% less than most other CPU's, but thats how it is.
Guess its part of other "issues" in my m17.

Going to work on GPU again now.

 

Well the card died during that gameplay. My computer didnt want to start with that card alone.
I was able to load new BIOS to both of the cards in DOS, trying to undervolt the cards before putting GPU 2 in GPU 1 slot.

Could of course be the cooling on other than GPU chip not good enough.
Got to think about either replacing or swapping both or staying with one for now.

(Probable cause of difficult starts could be high current on the f....ed card.)

 

I have a feeling the CPU score may be 3dmark related. It seems to be kinda glitchy. We should compare throttlestop bench times.
If the one remaining card works properly there's no point in changing it out.
Though, with the amount of troubles you seem to be having I'm curious about what voltage your PSU is delivering or if everything on your mobo works properly.
Any techie buddies with a multimeter, some electronics know-how and a bit of free time?

 

I didnt have probs with the power when both cards was running but was suspecting power issues when it just died and almost never wanted to start again giving up after windows initiate discovered the crap card.

Running the card alone gives me no problems, (on either PSU) And with one card alone i shouldnt have any probs with power?

My TS bench time of best has been 11.877 but since it was during my extreeme overclock only suitable for that test its not very good test.
Normal high clocks that could also run 3Dmark11 would give me in the high 12 or low 13 area. can try some tests tomorrow just for fun.


(Swapping both cards would then be for the 8970M cards. btw)

 

Guys, i have said that in the past. Sangemaru you remember it and said you would look into it.

I see Trome71 talking about his Physics score the past 2 pages and i didnt say anything to see if you find out the reason yourselves.

Firstly , i wanna ask. are these physics scores in 3d Mark 11? or the new 3dmark (the one including firestorm?)

 

If you're referring to HwInfo fan control, it appears to have no bearing on my scores whether on/off At least not for me I'm still trailing a few hundred points below where I'd hoped to be at my multi and TDP settings, just not as bad as trome.

Regarding 3dmark 2013, I can't even finish a damn run, the Clould Gate physics test insta-crashes on me no matter what my clock settings are.

EDIT: Just to make sure, I'll give it another hwinfo-less run tonight and see if something changes about my scores

 

just listen to the difference in physics score in 3d mark 11 , that closing the hwinfo does (closing the whole program, not only the sensors window)


......................................with hwinfo on..............................with hwinfo off
cpu multi 23x...........................6178.........................................6771
cpu multi 26x...........................6665.........................................7281


so for me closing hwinfo has a huge impact on cpu-physics scores

 

Woah... that's a pretty big gap, yea. I'll re-do the test. Come to think of it, I think the last time I did that test was back with the old PSU. Could be that the power throttling was artificially lowering my scores.

EDIT: Yea, I just checked. Physics score dropped around 300 points and the combined score dropped a heavy 500 with Hwinfo on.
This could explain some annoying CPU bottlenecks i've been having in games at high load. I think I'll be giving the Clevo vBios a try for the gpu fan control and see how that goes, since I don't use hdmi much, and I can just output audio directly through the jack anyway.
And that's using 1000ms sampling not the default 100ms one.

 

i told you.... :thumbsup: and been telling for 2 months now!!

By the way, if you choose the Clevo Vbioses with the autofan. Do you know that the gpu fans are not independently autocontroled BUT are controled pararel to the cpu main fan?

for example if you run a game or a programm that uses both cpu and gpu the gpus fans will go faster as the cpu fan goes faster. and in games you have no prob as the cpu is used alot and so the gpu fans rise rpms and cool the cards
BUT in specific cases as are the 3dmark 11 3 first graphic tests that dont use cpu but only the gpus , you will notice that their fan doesnt rise at all! although their temps rise to 80+ ! only on the 5th test that is cpu-physics you will hear the gpu fans spin

SO the clevo vbioses are actually autofans BUT controlled by the main CPU temp. Just to know.


PS.By the way, the weather has gone winterish here, temps droped from 30-34C ambient to 20-24 , and my gpus now run full gaming sessions not going over 66 degrees.. and that is with reused gpu thermal paste (its the same i have for the last 7 heatsink swaps) and with the destroyed thermal pads from my old 5870s..

 

Oh. That's the behavior I have on Dell my cards with Dell vbios when I don't use HwInfo. Crap. Ok, giving up Hwinfo is not an option then, and there's no point in using the Clevo vbios if the behavior is identical, and I'd lose overdrive in vbios and hdmi sound in the process.
I'll do a little experiment with sensor time reading and get back to you, maybe we can still use hwinfo.

EDIT: In HwInfo, under the Sensor Status panel, press configure and increase the Scan Interval to something like 10000. Once the scan interval is increased, there appears to be no difference between using Hwinfo and not using it, since the CPU time it's taking up appears to be negligible.
My score ended up being slightly (20 points) higher when hwinfo was disabled but the physics score dropped massively (since the fans spin at a lower speed by default, apparently, than the one I can set in hwinfo and my CPU ended up throttling). For some reason the Combined test doesn't heat up my cpu as much.

However, there's a deeper issue I'm still not getting the hang of. After I managed to get that 6800 physics score using 25/25/25/25 settings and 105W TDP (and my cpu even throttled during that one), I've not been able to climb above 6300 since then. The score is still much lower than you guys', even without running HwInfo.

I can't tell why.

 

Nice to know. I see big increase to. Close to 6500 now. (installing 3DMark13 and will test that too soon)

Since you do this for benching you should probably just manually set fan to full speed instead for using auto.
This should prevent your fans from starting to late under fast heavy load if you change scan interval very high.

 

I set it to 5 second scan interval (I consider that plenty OK) and my scores only appear to have dropped by around 50 points. I'll do some more testing of high-performance clocks once I get home and can tweak my cooling properly.

 

Anyone tested the 3Dmark13 benching?
I know ive got single GPU only but a bit excited about the physics scores there too to verify if it is a mark11 issue or CPU.

 

you got the free version? it takes forever to benchamrk because it does all the tests and also makes you see the intro scenes...

if only i could find the whole version...

 

I'd heard that users with low physics scores in 3dmark11 didn't have that problem in the Fire Strike physics bench. Unfortunately, the new 3dmark crashes for me always in the cloud gate physics test, so bleh.

 

And so the CPU drama continues......

Let me go for a beer...or a Sweedish Cider...

 

Well flingin, what cards did you buy? Clevos or dell? And from where?

 

2x clevo cards from Eurocom bro, second one is due to arrive on Wednesday.

If i see i will encounter problems of different sort, stability, hang ups, anything unusual, then i will go back to single card config.

I prefer to spend my time playing the accual game than to optimize forever and ever with 1000 diagnostics programs endlessly.

therefore if something will go wrong i will go back to 1 card + insane OC.

 

Hehe, I might need to buy another card myself and not sure how to get a card that i can insert that i can make possible to make it play with my other dell card. Perhaps it less hazzle selling the one 7970M and buy dual 8970M.

For me its not about getting world record benching, but it seems i have problems with my CPU since tests show very much lower speeds.

If its settings making problems then i have the explanation.
If its the bench program its fine and installing another benchprogram showing its in the area of normal speed im fine.

I can OC it quite heavily giving me quite extreeme high frequencies over time, but it doesnt end up in real CPU speed.
I can easily do settings giving me 3900MHz and do any benching with it, but i get 20 % lower actual speed on CPU than others says 3Dmark11.
Heck, any OC on my CPU gives at least 10% less than no OC for others and perhaps even lower than no TS current/power help for others.
Thats quite significant, also for gaming and I do play. I need settings that can make me able to play for hours, not for benching.

I have seen others scoring same as me. It could be due to having a QS CPU and others with same speed having the same?
Weakness in CPU?

If its the CPU i might have to buy another or perhaps if its my mainboard I might give up on upgrading this modda.

 

I don't know, if it were ES then yea, I could picture there being some differences that countered the usage of ThrottleStop and made your cpu un-overclock-able, but with a QS, if there is a problem you're probably looking at a software one, or some problem with power allocation.
That being said, i'll probably just keep 25/25/25/25 @ 100W after the heatspring tim gets here and get the CPU clocks off my mind entirely. It's plenty powerful already.

 

Its not easy getting a card not 022 to fit my other dell card

If you do TS test bench, what are the time for the small test?

 

17.171 on mine atm

 

For fun i did the heavy one with 25X and not BIOS OC just to see durability. 25X on all cores rock stable the entire test.

Didnt hit above 90 deg until 50%. Max temp 94 on some cores.
Right now i dont use my extra fantray.
Only a 2 cm elevation of the rear.

Same settings gives me 14.3 for the small one.

 

So the CPU drama continues ?

 

i beleive the 3dmark 11 cpu differences you see are software related. maybe something in wondows is laging your cpu. did you see what hapened with hwinfo? it gave 13% lower numbers when enabled, wlthough the tsk manager didnt show any cpu usage by it.

Also , give us exact cpu TS bench numbers with cpu @ 24x , try the 8 cores 32M and 1024M
if its on par with mine then your cpu is ok and you have a 3dmark software issue

 

3192 entire test 15.460 sec for 24X 32M (14.941 wo HW info (even at very low update)


For the 1024 its 486.535 without HWinfo. Ending at 87deg.

Shutting down windows saving modes etc lowered it to 465.755 but not much in it.

 

battlefield 4 beta is out!! go download so we can see how our aienwares handle this game!!

 

I did try it slightly but didnt didnt change to max performance.

Atfter that my card died.
Will try again in a few days to see how it works with singlecard and how good settings i can use.

Have you tried the TS benching in 24x?

 

not yet , i will bench later today

 

24x multy, TS bench 8 cores 32M ------> 14.108sec
8 cores 1024M ------> 447.975 seconds



AND just to mention that hwInfo takes up alot of cpu power, with hwinfo enabled the 32m score was 16.200seconds... 2 seconds up!!!

 

Definitely let us know how your rig performs. I am particularly interested in the results. I get decent performance on ultra with single 7970m with intermittent fps drops. This has occurred with many others. There appears to be an issue with PB too...after all.

 

Ki Alien.
Then it seems i see a slight dropoff on my CPU from yours. ca 3-4%. Can live with that. Better than the 10-20% I've seen in 3Dmark11.
With very high update time and fans running manual, i have apx 0.2 dropoff on HWinfo, not much.

Thx a lot
I might give it a go again and get another GPU for this laptop.

 

the 3-4% may be drop offs from some programs running on the backround. I closed everything before TS bench. For example skype alone running minimized on the backround added 0,5 seconds to the bench.
So be sure to close all the icons you can down from your taskbar (from hidden icons)

 

By the way, i benched my desktop 680gtx with 3dmark 11 (i hadnt benched it in the past)
so as to compare with my alienware. i wasnt expecting the results!
as you knoy my alienware normally scores ~9950 on 3dmark 11 (maximum is 10.074)
and my desktop i5 with the 680gtx scored just 8.404!!!!

in the graphic tests the alienware was 5-8 fps faster in all tests!! damn!! i wasnt expecting that!! i thought my desktop was faster. Now i'll have to start ocing the 680gtx!!

 

I have disabled all the powersaving stuff, and stopped all programs not needed.
Still im happy its ok. Do however get to 19 at times, dont know why and its only occationally, but after the lower times arrive its not going back up.

 

Hello all,

So finally I have come around to placing in my second 7970m.
When booting up I noticed the fan is not spinning.
The card is detected and running.
It runs around 70 degrees idle and my system shuts down after a couple of minutes.
Also I can't control the fan while using hwinfo.
Does my card has a bad vbios?
Thanks in advance

 

Have you enabled crossifre?
And then started HWinfo again?

 

Can you set fan speed for it from Catalyst Control Center? Under the Overdrive tab?
Also, I'd flash the 0975V 022 svl7 vbios to get easy access to Overdrive and keep the temps down. The cards shouldn't heat during idle, even with no fan activity.

 

if the fan is not spinning at all it would raise to 70C and more...
the heat would build up...

maybe you disconnected the fan cable or something while installing?

why cant you set the fan through hwinfo??

 

Thanks for the quick replies(I'm lost without you guys)
So I tried flashing the bios with atiflash and is get an error like 01f10
Did I brick my secondary card?
I'm also gonna recheck the fan connector and repaste the card.
I will update when I know more

 

This is what i think about this.
You will see in catalyst if you have two cards or just one once windows has started.
IF you only see one, you cant set up fans. HWINFO will also have to show 2 cards to run fans. If only one card, HWINFO cant run fans either.
If you have the possibility of running crossfire in catalyst you have two cards. You need to check it of in catalyst.

What cards do you have? (are they both dell or clevo's or what?)
Have you tried the -i message to verify both cards before flashing and have you tried now to still see that you have 2?
I'd try to get it running before flashing any VBIOS and wait with flashing until you know that its not other stuff making problems for you.
(Connectors ect)

Try reading out the VBIOS too before flashing it with new.
Try doing most of the guide on techinferno.

 

What vendor are the cards from, and what vbios did you have originally, and what vbios did you try to flash with?
Errors like that may be when trying to switch vendor vbios'es (like from Clevo to Dell) and they're totally normal, and you can force flash over it no problems.
It's important to make sure the checksums are ok and know in advance that flashing is alright though (although if you had an unflashable 022 card, you wouldn't have been able to post the system with the card in anyway).

 

The cards are both dell
And the primary card is running with the undervolted svl bios
The secondary card is running a stock dell 021 bios.
Both the card show up in hwinfo and catalyst prompted with an enable crossfire popup
I think I just messed up connecting the secondary fan.
I will try flashing the cards later.

 

Good luck