M17x R2 7970m CrossfireX Ultimate Installation and Tweaking Guide

I have had two 021 original and one 017 that came with 022 undervolted and overdrive enabled im afraid. I only have Dell cards and had lagging/stuttering before installing Nospheratu's clevo VBIOS'es.

Its not mine that was shipped with the 024.
But im curious if you are going down that road with dell cards and the 024........
If newer VBIOS handles power better, my cards might not shut down at high speeds.

Odd that such a new version dont have overdrive?

 

I'm making SOOO many confusions Anyway, that overdrive-enabled undervolted 022 vbios is custom made by svl7. The 024 ones would obviously not have them.
And yea, maybe they'd help... Feel like giving it a try?

 

Hehe not sure. If i still had problems with my cards id probably go ahead, its not a big issue that i can go "only" to 975 / 1440 for benching. ��

 

My personal opinion on that is:

I have plenty of performance now, changing vbios will not improve my performance in the way i will see it, i am sure of it, so why risking bricking my card, are they not dieing frequently enough already ?

What i have my eyes on is next gen GPUs from AMD, and by gods, please prey they do not just re-brand the name and bump the clocks, cuz i am sick of watching this happening....

I want a new Architecture, don't give me another factory overclocked 7970m and call it 8990m ( like it was already with 8970m, PLEASE THE $&*!^"$* SAKE

Another thing is that 920MX and 45nm. This is an XM cpu but it really starts to get beaten by newer models ( even non XM models )

My warranty runs out in 600 days, Long live the King, but it is time for dell to rethink the whole idea and put this bloody RGB screen in newer machines ( upgrade option )

 

But when it comes to the 8970M. Didnt they also add 2 extra gigs of VRAM?
Everyone reffering to the clock increase only, but that was only the first cards or am i mistaken?

Not sure how much extra the 2 more gigs of ram gives though. Does it only help in multiscreen?

I wonder how good our AMD's are after all the new drivers.
From what i read its nvida having problems with their drivers now, just reading this fora. I havent seen any test comparing the AMD's with the newer better drivers to the 680M and our crossfire with the new drivers compared to the SLI with their current drivers.

If NVIDA drivers hasnt improved the performance of the 680 over the last 3 months the cards should be quite the same, cause the AMD has done wonders.

 

2 more gigs of ram adds almost nothing.... It is like having 16GB of ram when all you use is max 5 GB. ..

Sent from my SM-N9005 using Tapatalk

 

AMD Catalyst 13.11 Beta 9.5 is out. I thought we'd a get a WHQL before the year was out but I guess not

Feature Highlights of The AMD Catalyst 13.11 Beta9.5 Driver for Windows
- Includes all Feature Highlights of The AMD Catalyst 13.11 Beta9.4 Driver
- Resolves the issue of AMD Overdrive missing in the AMD Catalyst Control Center for the AMD Radeon™ R9 290 Series graphics cards
- Resolves intermittent flickering seen on some AMD Radeon R9 270x graphics cards
- Resolves graphics corruption seen in Starcraft®
- Improves frame pacing r​esults in AMD Quad CrossFire™ configurations for the following: Hitman: Absolution, and Total War™​: Rome 2

Download here.

 

Yeah !....oh wait....no WHQL ?....meh...

Hitman: Absolution, and Total War™​: Rome 2
hmm...who plays that crap ?

 

Lol, that's what I said Boring release, but they never really mention all the changes/fixes they implement so I'm hoping for some relevant improvements.

 

i think they will be very irrelevant...unless God himself modified the .inf and .dll files

 

Guess this means pretty much no upgrade for us, not playing those games

 

is there something big going on with this 024 bios everyone is talking about?
sorry i am too lazy to read through the pages.
if it gives a performance/cooling/powerconsumption upgrade i might ever so slightly be willing to try it.

 

Just tried out the new Beta drivers, and the 100% load at idle issue resolved itself. Maybe this release will fix other people's problems too . . . . hmmmm

 

Well I know that it doesn't resolve the HDMi audio on CLEVO cards. properhit over at T|I is running it atm and his default 3D voltage is 1.000V. He doesn't seem to be experiencing any anomalies with the RGB screen as well which is good.

I also experienced that after installation, the clocks dropped to the normal idle state after the first installation reboot on my secondary card. I was a bit sceptical and rebooted again and it was idling at 3D state again before the card went into sleep mode. I turned off ULPS anyway since I like monitoring my second card even on idle and I have a feeling it will cause issues with CrossFire.

edit: I also think performance might have taken a slight knock as well. In TR the fps is all over the place in the Shanty town level which is where I'm at atm. It could be just a badly optimised level though.

 

Not sure if we ever experience the power issues when using the 330 W one,but a thought comes to mind.
I think this really should work. OR will it?

The new external mod from Nospheratu with the chip in the wire, should make the setup for dual PSU a lot easier and smaller than before.

Since both id chips are grounded in the mod, AND there is one chip going for the laptop, there really is no longer a need for the voltage and amperemeters.
The supplies should run when needed and no need to think about interrogate ID for each supply like they do on the 2*240W mod.

What you really need to do from what i see:

Mount two inlet plugs.
Ground the ID wire in the inlet for both plugs as in the plug mod.
Get a diode bridge and wire inlet like on the 2*240 mod, and do the ID mod on the outlet again as shown in the plug mod.
Problem beeing, 20 amps might not be enough when using the 330W, so there is a need for a heavier bridge. This mod can of course be done when using 2*240's to.

 

@Trome71, There's not much space saving honestly. Mr.Fox posted up his dual 330W mod on T|I which if I understand you correctly is what you had in mind. No voltage or amp meters used.

@NiveQ, yes it is risky especially since its a hybrid Enduro/Non-Enduro vBIOS built for CLEVO cards. I don't think Dell implemented a hybrid vBIOS for their cards.

 

May I ask which vbios you are using and where did you get it?
And does this enable automatic fans?


Coming back on psu modding.
I found a medical psu with 650w capability.
And can be set to almost the right requirements.
Would the m17 take 15v instead of the specified 20v?

 

It would take the 15V, but you might have to pump it in through the battery connector. It all depends on if the two power systems use the same regulators or not and if they are rated to produce 12V, 5V, 3.3V, 1.8v, etc from the an input as low as 15V. Routing it through the battery port should work in theory because the battery is only rated at 11.4v or 14.1v or whatever it is.

However, I experimented running my M17x with dual 6990m's at 18V a long time ago and it ran, but wasn't happy. The cards did not perform well under load at all. This is why using low VF schottky diodes is really important in the dual or triple PSU mods. A 0.7 to 1v drop is really all you can afford and not affect functionality.

You also have to consider the fact that with the lower voltage, the computer might potentially draw up to 25% more current, which while that translates into the same power, the increased current might damage some components. Lower voltage, high current devices usually run a lot hotter than high voltage, lower current devices.

 

*I use Nospheratu signature VBIOS. Clevo, slighlty undervolted.


It is almost as i thought of it but he doesnt use the plug mod, making you able to run this with one psu easy.

Then again, using such big power diode bridges, it developes a lot of heat and will be bigger due to cooling

Not sure 15 will work.
The current needed for the high effect will have to be a LOT bigger if the MOBO approved.

 

Stop thinking about this.
It is inefficient, and dangerous for mobo.

This laptop was not designed to draw so much power, so if you intend to make 650W psu for it, and even if you will succeed, and even if you will be able to overclock the hell out of R2 ( IT ALREADY IS ), then you will probably face your components life to shorten dramatically.

It is NOT worth it.

The best thing we can do to make R2 last another 2-3 years is to constantly put newer GFX into it and undervolt so components are Stress-Free, keep power draw within 280-300W MAX and Keep CPU at @25 All cores which will not bottleneck the GFX so much.

Overvolting GPUs and OCing is just not worth it, those 7970 cards tend to Die very frequent even without OC.

I would say, ondervolt to 1.0V and slight OC to 900/1300 is ideal to have a Little more power and keep everything cool, even though this level of overclock is pretty useless as it will give you ..what ?....from 30 to 32 Fps ?

Last thing i want to see is a bloody black screen and to find out one of 7970 is dead.....AGAIN !

That is why i am re-designing My R2 GPU heatsinks to keep the cards even cooler, i know it takes long, but believe me, it is not easy when you have almost no room for adding a 3rd heatpipe and no space on the heatsink to add anything.

Anyways i already added 3rd Heatpipe, and i need to only do some more modifications with the top heatsink part.

 

If we were worried about inefficiency, we'd all be using ultrabooks. And compared to a similarly specced desktop, the R2 and gaming laptops in general are actually quite efficient. . . . Also something being inefficient doesn't mean it's going to damage itself.

But like you said, the 15V idea imo is a bad idea. It will be bulkier, more dangerous for component lifespan, and more difficult to implement than a modded 330 watt PSU or a dual PSU setup.

For the past year and a half, I've been running my R2 with close to a 400 watt draw from the wall without a hitch. The 240 watt R2 PSU can deliver ~280 watts. The internals are rated to at least this. It's common practice in the PCB design industry to spec your critical power components to 20%-60% above what your intended power draw is. I've also done a bit of research on the actual components used on the R2 motherboard. A fairly educated guess would put the limit of the motherboard to be around 340-400 watts. If you account for a 15-20% PSU inefficiency, my 400 watt draw, puts the actual power my motherboard sees at ~320 watts.

In my opinion, you will have to do some serious cooling mods to the R2 (think watercooling) before you will be able to reduce your thermal limiting factors far enough for you to be even capable of pushing 400 watts through your motherboard.

Unless you can somehow manage to overvolt your cpu and gpu's far beyond our current capabilities and have them be thermally stable at those voltages and full load, "too much power" should not be a problem unless you have a faulty or damaged motherboard to begin with.

Just because you could hook up a 650 watt PSU to the M17x, doesn't mean it will draw 650 watts. And any draw above ~340 watts (after PSU inefficiencies) isn't feasible with our current thermal limits. At this point in time, our motherboards are safe.

Another thing people don't realize is running your cpu or gpu literally on the thermal limit will not harm it. The actual thermal cutoff for our cpu's where heat induced damage starts occurring to the die is right around 115C. Our chips will automatically throttle voltage and clock speed at 99.9C to reduce die temperature. If the temps can't be decreased in a few cycles, the cpu will perform a hard shut down. Due to this, it's physically impossible for your cpu to get anywhere near a temperature that can damage it.

Same with the gpu's. Mobile chips usually have a thermal cutoff of about 120C before die degradation starts occuring. AMD employs the same method as intel, but their throttling que varies by chip and is usually around the 106C mark.

It's voltage that kills chips (and voltage combined with high frequency), not heat. There's no reason why everyone shouldn't be aiming for stable 99C temperatures because that will result in the best possible performance you can achieve and still be safe all day. NVIDIA is basically blatantly advocating this fact with the way their GPU Boost 2.0 oeprates. What really puzzles me is where this 80C magic number came from for die temperature. It isn't based in anything logical or concrete.

My 920xm has been overvolted 75mv and forced to 3.5-3.66ghz with a 110 watt TDP. It's been running like this for almost a year now. Under load. Temps are usually 95C+. Never had an issue.

Same goes for my gpu's. I overvolted my 6990m's and ran them at 102C limits for over a year. No issues.

Same goes for my 7970m, it's overvolted and running at 1ghz all day. When I finally get another one, I'll see if that can be pushed any higher.

You should only bother with undervolting and underclocking your chips if either you don't have the PSU to back it up or stock settings are already pushing the 100C soft thermal limit. Otherwise you're needlessly cheating yourself out of performance. And the performance gains are not negligible, in the case of the 7970m, a 20+% performance increase above stock is not unreasonable. That's almost on par with a generational leap in architecture.

(NOTE 100C heat won't hurt our dies, but combined with other factors, it can warp our PCB's which can lead to loss of solder integrity. It won't affect our motherboards, but it can affect the MXM modules. There are ways to minimize this risk though - and some series are more at risk than others. This is what resulted in so many 69XX series and 7970m failures. PCB warpage. High pressure heatsink mods from day one ( before the card has failed) can go along way in preventing this.

This is why I wish gpu dies were socketed like cpu's instead of bga mounted. It would go a long way in preventing PCB warpage.

 

You have somewhat aaaaaalmost silenced me with that post

However, your opinion is based on YOUR personal experience without taking others into account bRO

For some it works, and for some it will result in a component failure, i have once added some voltage (0.75mV) to my 920XM, and i damaged it memory controller so it could only see 1 bank of memory. Other Bros had that problem too, not to mention Posts with dead 7970m cards.

 

But you see, that's my point exactly. Damage occurs with increased voltage. Not temperature.

EDIT: And you're right, I only have my personal experience to throw at it, but I've been doing this on a multitude of laptops over the years. Gateway FX P173X, Asus G71gx, and now the R2. I've fried tons of components, but they've never, ever been heat related die failures. All of my failures have been either voltage related or personal stupidity =P

Also, GPU Boost 2.0 is the best backing for this ideology.

EDIT #2: On a side note, I'm really intrigued in your heatsink mod. Have you posted pictures before or do you think you can post some here when you get a chance?

 

In these days of Tjmax near and over 100C, heat issues are, like aarpcard said, not as much of a biggie (although they're still noisy, unpleasant, may bend or warp or deteriorate nearby components and may throttle).
But yea, you never know what that extra .75v will do. I don't know for what reason my cards started disliking any voltage other than 0.975v, but I'm grateful they run this voltage properly and stay very cool.
For some dubious reasons my GPU temps these days are usually under 70C. Playing AC4 Black flag maxed out has my cards at around 66C.

And I know for sure that pushing my old 920XM on a different machine (not even overvolting it) lowered its max stable OC while temps stayed the same.
I didn't overvolt it, just overclocked it as far as I could (3.6GHz). The more I used it at that speed, the more I had to start dropping clocks as the weeks passed on. After a month of gaming OC'd, the only stable clock remained 3.32GHz on that chip, whereas when I bought it initially I'd managed to rise it all the way up to 3.8GHz once.

 

For me,now playing BF4 for about 2 hours my GPU tells me they are at 62deg still.

Running 900 / 1300 atm.

Stock Voltage.

 

Liquid metalpads arrived. A single pad was enough to cut into 3 layers to put on the cpu (used only 2 though).

@ 140bclk, +75mv, 25/25/25/25 (3.5GHz), TDP 105 / TDC 90A, in throttlestop @ all cores max temp is 88C. 3 degrees difference between hottest and coldest core.
Using it @ stock voltage, 139bclk, 24/24/24/24 (3.33GHz), TDP 105 / TDC 90A, max temp on hottest core is 80c, coldest is 75.

Ironically enough, my 3dmark physics and combined scores are still crap Probably need windows reinstall or something.

 

And where is one for me ?

Bro ? !

Are those temps result of 3dmrak only ?
For more detailed comparisons, if you could, you could run TS bench 1024M, 25@all cores TDP105/TDC90 ? PLease ?

 

Coollaboratory Liquid Metal Pad Kit Processor Cooler: Amazon.co.uk: Computers & Accessories

Check the alternative prices, there's some at 13GBP.
That kit is huge enough to use for like 3-4 machines.

 

OH !
I thought you have Those HeatSprings !

I still have metalpad...IN MY DRAWER !

did you bought this particular KIT ?

Why did you used 2 Layers ?

 

Hahah bro, u have it and not use it? Yea, I bought that kit.

I used one of the big CPU pieces, cut it in layers and used those. I did that because I'm unsure my pressure is any good since I had much worse temps than you guys. So I stacked 2 and it works great.

 

I just found out that my girlfriend threw it away during last Drawer Clean-Up, Just LOL.

Yes that is a pretty big pack, it is 13.39GBP for me because i have no VAT in Jersey

Bro, can you run TS at the settings i mentioned please ?, i could quickly compare it with my IC Diamond

 

One of those "Sweetie, don't you effin touch my shtuff" moments.

What settings bro? I can't find them.
EDIT: Found them. I will a little later, have to piece my machine back together again, I was running it disassembled.

Results: Completing the test duration made the temp slowly build up. I admit I hadn't let the test run all the way through before.

@ 25/25/25/25 TDP 105 / TDC 90A Tsbench 1024:

504.418 score;
Hottest core temp reached: 87C
Coldest core reached: 84C

I'm downloading 3dmark (finally bought the key) and browsing on alt-tab.

Is 3902 Firestrike Extreme score any good?

 

Something is really...Throttling in your system or what ?

Look at my score, i repasted my 920XM like 2 months ago... temps usualy peak at 81-82 at approx 75% into the test and do not go higher



Did not run Extreame Firestrike just yet.

 

Yea, told you, I was downloading off steam in the backround at like 10MB/s, also browsing, etc.
Yea, your temps are quite better with IC diamond, but for me they're near the best I ever had, and the only ones that allowed me to overvolt and run good temps at 3.5GHz.
Now the real test is, will these temps stick?

 

Yep, they will stick, and should improve a bit over time after long-time burn -in, in this case time is in your favor, not like normal paste that degrades with time..this thing..improves with time..which you know already

Try frying your CPU a bit more, even up to 95'C, for this, change voltage to +150mV, but be carefull with multis during this, start low at @20 and see what happens, then increase by 1 if temps are ok until you hit approx 95.

It should help those two pieces to stick together, and reach optimal performance.

I think anything up to 5"C improvement is possible, but not guaranteed though....

 

Cool, although to be honest I don't really feel like running overvolted. Unless I fail to obtain 50-60fps in some game because of CPU bottleneck, I don't plan to overvolt.
I just gave it a try at +75mV 27/26/25/24 @ 140bclk (3.78GHz single-core) @ 90TDP/85TDA and only got some 7500 physics, but it seemed pretty ok to me. Temps around 87C.

 

why did you not tried 1 layer ?

 

Better more than sorry? I'd read a few reviews online that sometimes one layer didn't do the burn-in properly because of bad mounts. Wanted to avoid that issue and just decided to stick another layer. Can't really complain. It's a great product, where none of the other TIM's I used managed to provide this kind of performance or allow me actual OC headroom. You realize I reached 3.78Ghz and still haven't touched over 90C?

 

That's great news sangemaru! Glad it worked out so well, you must be really pleased with the result after all that you've been through

In other news... it's a sad day in South Africa. The country is in mourning as Nelson Rolihlahla Mandela passed away last night

 

I'd heard. From what I've read, he was quite loved by the people.
Any word on the street about who's expected to take over?

 

Yeah, he was the face of the fight against Apartheid and if it wasn't for his guidance and decisions under his presidency we would have most likely seen a civil war. He's been retired for 14 years now, there's been two successors after him but neither could fill his shoes. If you're interested, you could watch the movie, "Invictus" that came out in the year 2009 for a rough idea of the type of man he was. There's also a new movie that came out this year called "Mandela: Long Walk to Freedom".

Anyway, I'm thinking I should just order the ICD and get my CPU some relief. I'm hitting 99C with 70TDP as its summer now and I can notice the difference with high TDP as there's no stutter when loading new areas and physics related areas in TR.

Anybody here running ICD for around a year? I'm just trying to figure out the time period before I start seeing a drop in cooling performance.

 

I'll be doing re-tests of the Liquid MetalPad at the same settings as the one previous every 7 days to see how performance holds.

EDIT: Currently running 27/26/25/24 @ 140bclk, +75mV, 100W TDP / 85A TDC. 3.8GHz single-core - 3.33GHz quad-core. Temps are very comfortable on average, low 80's-high 70's. Only very demanding apps push them into the high 80's. Lovin this.

By the way. I bumped it up to +150mV and ran 4GHz no problem. The chip only starts to get smoking if I run quad or triple-core at high multis (above 3.5GHz). I have to get psychologically used to this and see if it's stable long-term etc, but at this temps I might end up running it all the time at 4GHz or more single-core with something like 3.4-3.5 quad.

I'll have to do some serious experimenting to find the highest stable single-core clock, and see how I can tweak the multipliers from there to get the best temps and best clocks possible.

EDIT2: Highest stable OC appears to be 3.94GHz @ +150mV. I'll stick to using my previous 3.8GHz settings for daily use. Playing starcraft 2 on extreme settings with a few thousand units now only drops fps to 30 instead of 10-15. Max temps after a session of sc2 are 83C. Pretty good aye?

 

Getting almost exactly your 1024 score flingin. (434,662) But my CPU runs hotter, i get to about 87 max core. .
This in about 23 deg inside.

But our temps are not easy to compare if we all run on coolingboards, and they are not the same.

How high do you run directly on the desktop?
And are you running 100% internal fan control or HWinfo?

Running the internal fan control for CPU, no cooling pad, and setting as you guys describe, and 1024 gets me to about 97 at apx 50% so i stopped it.

 

Hi bro !

This are my temps on the desktop (idle)


I never ran Fan on 100%, it is always on Auto...

 

Hehe, i was more curious if you actually are able to run the 1024 directly on the desktop and how high u go if you are?

Still i do see that i might have to do a repaste, or a new internal clean to reduce temps.

What settings did u use?

GPU and CPU?
Here is a pure 25X CPU with 105 / 90.
and 900 /1300 gpu

End temps: GPU's 56 and CPU 73.
With cooling pad.

 

I was always wondering about those low temps. It will show about 60 *C on core max but when you will touch the gpu heatsink ( with keyboard flipped over ) it will almost burn your finger...strange isn't it ?.... Isn't it ?!

Sent from my SM-N9005 using Tapatalk

 

So, no more heatsprings?

 

I know that even 60 would feel hot, but 60 in the GPU should be colder on the heatsink.
It is odd, truedat

Havent tried on mine