M17x R2 7970m CrossfireX Ultimate Installation and Tweaking Guide

It will lock on the lowest multiplier if the cpu will draw more then 80% TDP so basicly the 4 multipliers are only used when surfing the web.
As soon as you boot a fullscreen app it will lock on.

 

At the moment I run 55W TDP/ 55TDC 26/24/22/20. No overvolt or bclk. I can't push the cores higher because I suspect my heatsink is defective. With all cores running prime95, 1 core is in 77C while the other is solid 100 and the others average out between them.

 

95w/75tdc
All 4 cores to x26
Thats it.

Don't use prime 95, it's not really suited for XM benchmark.
XM will run hot either way and prime will cook it.

Use 3dmark 11.

You might have to use +75mv... Cpu should hit between 75 and 90 degrees celcius, that's normal.

Don't use bclk overclock, its rubbish

 

Don't worry about one core getting 99 and the other 76 and the rest in between , it's exactly like that for me too. Thats why I will do stamatis trick. I will take the keyboard off , repaste and try to screw the CPU heat sink while the pc is running a benchmark.
Stamatis said that if you do it this way you will see that while screwing each screw of the heatsink either more or less you will notice the various cores changing temps. It's because the heatsink changes the way it contacts the CPU. So you balance the heatsink screws when you get the best result on all cores

Now concerning the multies I run at 27/26/25/24
AND UNPHOTO I KNOW THAT WHEN SOMETHING DEMANDING RUNS IT LOCKS ON THE LOWEST multi but even that is 24*133= 3192mhz... Super enough for gaming.


Speaking of gaming today I installed 2-3 games just for benchmarking with my 5870 cf so to compare for when I install the 7970. I benchmarked Battlefield 3, cod modern warfare 3 , and max Payne 3
Well , I was impressed and somehow troubled (that I sold my 2x 5870s) because all of these games on 1960 x1200 on ultra run at 40-50 frames...
I remind you that my 2 5870s run naturally at 800core 1200mem and score 5000+ on 3dmark 11'

What I want to say is that I hadn't realized how powerfull these old cards are if over locked .
Of course the 7970 will be even faster but I will miss those 2 old suckers... :hi2:

 

Hey guys about to purchase a r2 and wanted to ask something to the gurus/experts.
The r2 is 8gb ram, 4870, and i7-840qm.

Now i want to place 2x6990m in crossfire in the alienware cause i have 6990m already in my clevo p170hm. I then intend to place a 7990m in my clevo and finish on that.
Would the psu 240w be enough for 2x6990m in xfire with the 840qm?

Also which would score higher 3dmarks11 the clevo with the new 7970mplus 2820qm or the alienware with 2x6990m in crossfire on a 840qm?
I do know the seller i am buying from will throw in for free another 240w psu so i can hook them up for 7970m crossfire in the future.

Ideas/thoughts appreciated.

 

Why the caps? I wasn't talking to you, that was for sangenaru.
30+ posts and an attitude....

 

Yep, those are still some damn fine cards... but still, a 7970m is like 3 times the awesome

The clevo will score much higher, mostly on the basis of CPU. It's better to run a single card than 2 weaker ones anyway. But that config is not bad at all. The PSU is enough for a non-xm CPU.

Let's assume the best Maybe he just typed in a hurry.
Thanks for the advice mate I'm pretty sure my heatsink is either defective or badly seated, I'll check again but I'm getting sick of it, i'd rather buy a new one. If I try your settings it willl burn to the heavens as it is.

 

Did you monitor your clocks with ThrottleStop when the CPU is loaded? 55W TDP won't maintain a 20 multiplier with just 55W TDP on my chip, it throttles down so setting a 20 multiplier across four cores using that TDP becomes a bit pointless.

The adjustment of the heatsink while loaded is a good mehod but only really effective if you do the c-clip mod on the heatsink which is dangerous if you're not careful. That said, I did slightly better even temps across my cores with the c-clips on.

The way our CPU's work has nothing to do with whether the application is demanding or not. Those multi's only kick in depending on how many threads/cores the app is designed to use. The app could be demanding but only single threaded in which case it will use the 27x multi and will only maintain that multi consistantly without throttling if the TDP sufficient.

The 5870's are very good but where they fall short is the poor tessellation performance which most games these days use at high settings. I am able to run Crysis 3 at native res at +- 35-40fps on the medium setting which is still impressive for these cards.

The 2x 6990M will work with the 240W but barely. You wont be able to oc and maybe have to drop voltage to get them to run stable. The Clevo with the 7970M will be almost as fast as the 6990M's in xfire.

 

So after a bit of nosing around I found the following information:

http://forum.notebookreview.com/alienware-m17x/700280-situation-330w-psus-5.html

In this thread, it was revealed by imsolidstate that the power trip point for the R1 motherboard is set at 11.69Amps, which would come out to around 227W power draw at 19.5V which ties in nicely with power draw reports from Kill'a'watt users. This means I have 220W of power to play with (more than enough for decent speeds imo) and I'll see about sourcing a 330W PSU, doing the chip mod and having a friend check out the r2 mobo and upping the resistor limit. But first I want to confirm that the R2 is usable with 940XM and 7970m crossfire providing a little bit of undervolt and keeping decent power draw limits on the cpu. I'm thinking something around 3GHz across 4-cores, 3.6GHz single-core, power-limited to 50 Amps in throttlestop would do the trick for about 55-60W CPU consumption, leaving almost 180W for the GPU's to play with.

 

Yes, I did monitor throttlestop. I was very surprised running Dishonored last night that with 55W TDP/TDC at 27/23/21/19 multis the cores were running @3.6GHz really evenly. Unless you're really pushing the utilization on these chips, the cores will clock very high even at such low power restrictions.

 

That's great news But that would mean Dishonored is only making use of one core I have to admit this thread vs core gets a bit confusing especially when you read about how windows handles threads.

That's the perfect test though... games Everything else is secondary especially when we want to stick two 7970M's in this beauty. Hopefully I receive mines before the end of the month. Glad you almost have them now, I'm really excited for you!!!

 

Thank you thank you Thing is, dishonored is not just making use of one core. Utilization across most cores would fluctuate between 25% up to a full 100% depeding on moment and load. And the temps were very good. Sure, there are more stressful games available, like Crysys 3, Dishonored is just so very well optimized.

 

Yeah I meant single thread, my bad. I noticed that when I was figuring out what TDP was required for each multi even on a single thread test, all cores, even the virtual cores were active.

This makes me think that even though an app is single threaded the way our processor/windows handles it is by spreading the computation of that thread across all cores... which doesn't make sense to me, then again, this might be utter nonsense I'm typing for all I know

I'm just glad you started this thread so we can possibly compare results/solutions to get these cards working without any throttling or shutdowns :thumbsup:

 

I hope the postal factor drops off the cards today, I don't want to wait until next week
Yes, especially with HT activated the CPU's will do their best to balance even single-threaded load on different cores to keep them cool as well as be as efficient as possible (in terms of performance, but NOT power consumption )

 

Are you planning on using 0.5mm or 1mm pads on the vRAM? I'm currently using 0.5mm Fujipoly Premium pads on my 5870M's and contact is perfect. I have another sheet I intend to use on the 7970M's but not sure how that will pan out.

You could also use thick thermal paste but that would be a lot of paste and a bit messy imho.

 

It wouldn't be the first time I have to use paste but I plan to use 0.5mm phobya 7mW pads and if necessary stretch them out.
Sure, some stickier pads would have been nice, these phobyas are very rigid, but meh.

 

Maybe i dunno, but he did have a good point about removing the keyboard and tightening the screw on the CPU HS while benching.

Even 1/8 of a rotation will adjust your temps of one core or 2,it's the best way to get all 4 cores around the same temp.

 

Tried that, no change. I have a feeling I have to remove the gums around the screws to let them tighten more. I don't have C-clips on my heatsink screws and was kind of bewildered about how to mod them.

 

So I'm currently conversing with a schematic shop online and looking to find and provide schematics for the R2 board. I'm curious about previous reports that there is a 220W resistor-based trip on the board and why this doesn't engage with dual-240W PSU's but does with the 330W ones.

I'll probably only have time to check out my board sometime tomorrow, I'm busy with work, but if anyone feels like opening up his R2 and taking a look, I would need the board Part-number/revision number or similar information to identify it and send it on to the schematic vendor. It should look something like this:



The vendor usually sells schematics for something between 10$ to 15$. If he locates one for me for a decent price, I plan to probably buy it out and have an electronics specialist look over it for suggestions on whether or not it's possible to update the resistors on the board. If anyone wants to do a pool, pm me

 

No change from all angles? That's not right, it is obvious that on corner has WAY more pressure the other 3.

What do you see when you repaste the gpu and do a mock up?
So paste it, close it, tighten it, let it set for a fee minuten and then dissasemble it again.
You should be able to see if the HS touches the complete surface.

I did this in my thermal pad tutorial and saw that my HS was warped

 

Obvious yes, but this heatsink is a total conundrum. I got the machine second-hand, and I suspect the previous user was inattentive when repasting/reseating the heatsink and may either have touched the heatsink with bare fingers or damaged a heatpipe. The heatsinks has contact with the core... somewhat. The pattern is more obvious in the middle of the cpu but pretty much absent on the sides.

I had to replace the thermal pads on the heatsink with much thinner phobya versions to even get contact. Before that, unless i used HUGE amounts of thermal paste the core would remain almost untouched. It was very frustrating. It's pity that the design would not be bad... if it worked. But it feels as though someone made a good gooling design, passed it on to dell, where they made random retarded decisions about what thermal pads to use, with what fans, what pressure etc. and it all got messed up.

 

Damn, sorry to hear that man. Do you have the same problem with both left and right heatsinks?

 

No, CPU heatsink only. The current 285M GPU's are nice and cool, running maximum 77C under furmark or gaming. Just the CPU is having trouble cooling.

 

I didnt mean to offend you , but at your last posts you mentioned repeatedly that when on load (by a multithread program) the cpu uses the lowest multi. I also though that it was about my posts where i mentioned my multies thats why i used capitals. Sorry

PS. although i am new in these forums I am a professional IT (working currently for 2 International companies) for 16 years. So i realise deeply the ways pcs work. nevertheless i firmly beleive an ancient Greek says " as long as i grow old i always learn new things" "γειρασκω αει διδασκομενος" . so i always have open ears for new things.

Thanks

 

if i understand correctly what you say about the cpu-heatsink making contact only in the middle of the core and not on the edges, you can solve this easily. repaste using the alternative old school method of using a credit card to evenly make a "sheet" of thermal paste over the core. DOnt use the usual method of just leting a drop of thermal paste as big as rice in the middle of the core and then put the heatsink on.

Also what you said about the previous owner puting his fingers on the heatsink, doesnt matter, because although body oil from his skin is a bad thermal compound, if you clean the core (i am sure you have ) with alcohol the body oil is easily removed from the equation so its not a problem now.

if you manage to make the 330w power brick work with the R2 then i give you my respects. it will be a first. I really hope you succeed, cause noone has managed to make the 330w REALLY work , i mean they do change the chips inside so it is recognized by the R2 BUT it doesn't actually give the needed power and shuts down after some intense gaming use (or as some people say maybe the r2 with a 920xm and 2x 7970s oced needs more than 330w) .

 

if u watch the video of the dual 240w psu mod, only one of the psu works when u start it up. from there u would have to unplug the working one and transfer the load to the second psu. replug the psu, then suddenly the load is almost balanced.

Alienware M17x R2 - Dual PSU Modification - YouTube

what i think is happening is, u start off with one psu to get pass the psu check. from there u can turn it up to whatever u need.
honestly, i dont understand why a diode bridge would do that. a diode bridge is usually used to convert ac to dc. this is some voodoo magic

 

Hehe, yes, you know what they say. Any sufficiently advanced technology (as perceived by the uninitiated) is indistinguishable from magic

Regarding the heatsink, yes, I will first clean it (i need to locate 100% isopropyl alcohol, since all I have atm is rubbing alcohol) and then reattempt to use it. If it's still not working properly, I will replace it.

Regarding the 330W mod, it is my suspicion that while the PSU is properly detected, the motherboard will issue a warning signal when approaching "full load" of around ~220W telling the PSU to switch off. It's the only thing that would make sense to me. In that case I would have to locate where the signal is received on the PSU and simply remove the mechanism. I am guessing this, since I haven't yet had a chance to see exactly what happens when power limit is tripped.

 

This is all speculation and has been discussed many times before but the simple fact that the dual PSU mod works absolves the motherboard from any power restriction theories. It may be the way our boards draw the power that doesn't work well with the 330W... I'm not sure. I was so excited when they finally got the 330W to work as the only thing that was seemingly holding us back was the ID chip.

I don't think its our boards sending a shutdown command as there's no signal pathway anyway to send the command. The ID signal is a "closed circuit" and doesn't interact with the PSU circuitry at all. From what I understand its the PSU thats initiating the shutdown due to some built in safety overload mechanism. It could be a spike in current, I honestly don't know... its all magic to me

 

Well, I just got the cards on the table in front of me, so tomorrow will be updates day (providing I don't run into any issues).

 

what ? tomorrow!!! oh no my friend!!! you have us hanging all week ! now you must cancel all you friday night jobs and apointments and sit at home installing these puppies!! WE WANT ANSWERS!! :thumbsup:

 

what about the possibility that a 330w is just not enough??
stamatis has mentioned that on full load ( 920xm 2x7970 oced) the wats coming from the wall were more than 330w!!
and well actually it makes sense!!! just add 100w the oced cpu , 100 x2 =200 wat the cards =300w allready without screen, mobo, leds etc.., and not to mention that this nuber might go up on heavy oces..
so only the 480w mod is sure to provide enough

 

Then I'll finish my DnD before midnight and start assembling/disassembling during the night

StamatisX used them on full load with heavy OC My plan is to use either stock or slightly OC'd 940XM with heavily undervolted stock GPU's, or (if we can fix the 330W mod), heavily OC'd CPU with slightly OC'd GPU's. You should take StamatisX's usage case as an extremely heavy load. I plan to go a more moderate route.

I'll have to see about getting a 330W PSU for experiments if staying on the 240W PSU undervolted doesn't go through. That will probably take extra time, especially since I'm under extreme budget constraints.

Ideally, if the cards I got from Riri-Fifi are good undervolters, I may be able to keep their consumption low enough that I would not over-tax the 240W PSU.

I'm estimating 5-8W for Hard-drives, 5-10W for the screen, who-knows-how-many for AlienFX and another 10W for other components. That'd be say 40-45W.
The CPU with a moderate (single and dual-core-focused) OC should stay under 75W under load. That won't leave me a lot of space for the cards, but I believe I'm being pessimistic about these numbers so far. There is also the tolerance of the PSU to take into account, that can make a world of difference.

imsolidstate had stated at one point that he could push up to 280W consumption on his 240W PSU. I have had similar experiences on a different machine pushing 230W consumption on a simple 150W Delta PSU with massively overclocked 920XM (26x across all cores hyperthreaded running prime95) and massively overclocked 7970m (1000/1500 1.1V 7970m).

The quality of the PSU is very important.

 

Congrats!!! Best of luck with the install!

Yeah it definitely isn't with his setup as he's got his XM and GPU's overclocked. But for standard clocks on the GPU and CPU I would think 300W would be sufficient. The problem we are facing with the limited amount of info we have is that those who did successfully do the 330W mod and measured the power draw from the wall, it never exceeded 220w - 240W before shutdown.
edit: seems sangemaru beat me to it

 

doesnt the m18x use a 330w psu with 2 gpus?

 

Yes, but that 330W was built specifically for the M18x so it works as intended. Its also been subsequently used for the X51 desktop without issues afaik.

 

i beleive the same. the 330w works as intented on the m18 , but on the m17 maybe the power requirements are not as optimised and power spikes make the 330w brick shut down

 

tick... tock... tick... tock

 

So the cards arrived and were installed in my machine. I won't go through describing the horrors of getting image on the lcd
I finally managed to get image this morning, however... one of the cards (the main) refuses to install drivers. It gives me error 10 (device cannot start). If i try to install drivers for it windows will hang at boot.
I took the card out, cleaned the slot, checked for any issues on it re-inserted it and tried again, it still refuses to start. I'll swap it for the seconday card and see if I can run the other one at least... off to a bad start
Oh, and the m18x (or whatever laptop this cable came from) crossfire cable is MUCH too short for the m17x-R2. Make sure you use a proper m17x-r2 crossfire cable.

 

Are you installing it on a clean install of Windows? If not, try to remove every bit of nVIDIA drivers using Driver Fusion/Driver Sweeper or any driver cleaning app. Use a registry cleaner as well to get rid of any trace of the old nVIDIA drivers. I don't think its a problem with the hardware. Also there shouldn't be a difference between the two cards, they can operate as primary or secondary.

 

i am horrified nobody did a proper 240W to 330W PSU mod yet, there are some workarounds, with 2 PSUs but who want to have that ?

i will speak with my friend who is working deeply in the electronics, maybe something can be done to finally get m18x psu to work in R2.

So much drama .....grrrr

 

@Nospheratu: One card boots just fine solo. The other card displayed corrupted alienware logo at boot (horribly so). I quickly disabled it, took it out, cleaned, repasted, reseated. Black screen boot with backlight on. Windows did not load in the background. The card appears pretty much defective. I've noticed there were traces of dried thermal paste on the IC's, VRM's and even on the mxm connector traces. I've tried cleaning them as best I could with isopropyl and leaving them to dry, but it seems to make no difference. I will try a couple more times (if anyone has some suggestions I'd appreciate them) following which I'll return the GPU to the seller for refund and try again with some other card... but that will take a while
@Flingin: That would be great mate I also wanted to contact a friend who deals in the same, but I'd need to have the 330W PSU to show him first.

 

Maybe they are running two different vBIOS versions. You can check if both have the same vBIOS versions. Boot with the working card in the primary slot and the seemingly faulty card in the secondary. Use ATI flash to check what the versions are flashed on each card. I think the command is "atiflash -i 0" for primary and "atiflash -i 1" for secondary.

I thought Riri-Fifi was running these two cards in his M17x R2 though?

edit: You could also dump the vBIOS from the working card and then flash it to the card in the secondary slot. Crossfire cable does not need to be installed to do all this.

 

I'd already flashed identical bios versions when the screen didn't work. I'm currently running identical dell stock vbioses. The problem is not in the vbios, they share it and i can flash it just fine. The card displayed corruption and then refused to boot.
I'm sure the cards worked just fine when Riri-Fifi was using them, but they must have gotten damaged during removal/transit (and i don't like thermal paste on MXM contacts). Right now, the card is as good as dead. I hope to ressurect it, though. The other card is a very nice performer, it would be a shame to have it go alone.

 

. I am really sorry and sad...
IT was supposed to be plug and play because rrififi was already using them in his m17

Anyways. Until you can get your hands on another one for SLI at least give us your opinion on the single one. Run any benches? Any games?

 

damn man... I'm really sorry to hear this. I don't have any other ideas to try apart from baking it but I wouldn't do that unless it was a last resort. Rather return it. Did the shipping box show any signs of damage that could have attributed to it?

 

are the sli cable and the crossfire cable the same on the laptop? or is there a special one for nvidia and one for amd?

 

They are different and cannot be used interchangeably. You would have to get the SLi cable for nVIDIA and the crossfire cable for AMD cards.

 

Sure thing mate I didn't get to run any games yet, reinstalling my system, but here's a quickie: AMD Radeon HD 7970M video card benchmark result - Intel Core i7-940 Processor,Alienware score: P6835 3DMarks
I'll be back with others as time passes and I finish installing things.

I have to admit, while the packaging was underwhelming (the two cards sandwiched together with bubblewrap between and around them in a cardboard box padded with foam) I don't believe any damage occurred during shipping unless the customs office opened the box and did something stupid (i have no way of knowing if they did). No baking then. I'll have to return it.

Don't forget to make sure you also get the correct length. The m18x (or other model) cable i got from Riri-Fifi is too short for the m17x-r2. Sure, you can mod it, but...

 

Yeah packaging sounds pretty good. I've been reading up on baking again and it seems like that card would be a prime candidate considering the symptoms you experienced. But I can understand from Riri-Fifi's side he wouldn't want a card back that's been baked by someone else.

I think the cable you received might be a Clevo version as they are normally shorter than Alienware/Dells.

Some crazy clocks on your CPU and GPU man Glad you having some fun

 

Thanks, thanks . I'd normally be more enthusiastic about clocks like that (they seem totally stable, and I have lots of headroom to OC more), but eh... I had my hopes up for some experiments...