M17x R2 full internal cooling mod - huge improvement!

I'll never crack 20K Vantage, but this may help DR650, Joker, and Jubbing to crack it. My GPU's were not taking any more of an overclock, and it wasn't due to temperature. After all of the kinks are worked out, and everything is proven to work with no ill side effects, I think this thread would deserve a sticky.

 

I would recommend using Tuniq TX-3 for something that is loose fitting I didn't do your mod, however, my temps are not that far off by just replacing my thermal paste, and propped up the rear end of the laptop for it to breathe. my temps never get over 65'C(which is the mem) when gaming for hours. Tuniq TX-3 is the best thermal grease

and here is a quote from their site

http://www.youtube.com/watch?v=12ce9FpKO3I

there is a video comparing AS5 to TX-3

And since it can be quite difficult to apply what I do is put it by a light bulb for 20 mins and spread it quickly using the method in this video.
http://www.youtube.com/watch?v=xTrSOywFn_4&feature=related

 

Mkelliny, you've got to run Furmark. My gaming temps never get very high either. Furmark is an extreme stress test.

 

as requested like I said a little higher than his but not drastic, but I also didn't take drastic measures(like a mod) + during gaming which is a lot less intense than furmark i never hit above 65c on my memory and the rest are in the low 60's to high 50's. (also I noticed that his was run at 640 x 480 I was running it at 1280 x 1024 I also ran it for 15 minutes not sure if that matters think it does)


Uploaded with ImageShack.us

 

OK. You're running 4870s and you're also running an old version of Furmark. The new one is in a volcano environment, and has a dragon in the center.

 

Yeah you cant compare those benches to ours.

As an aside, doing this mod will probably net you even better temp drops. Like, much better.

-Ash

 

Does that really matter it stressed it the same and the 5870's from what I understand are cooler and less power hungry. So me running 4870's at those temps would prove better for the 5870's. and this version is 1.7.0

here is the change log for 1.8.2

And why would my benches be null? My GPU should run hotter than yours, yet without ripping my laptop apart to install duct tape and what not the difference is not as drastic. I'm not slamming you or your mod I think that is great + rep for you I'm just saying that this worked for me and might work for others as well who might not want to go that route just yet.

 

You cant compare the temps of a HD5870m to a HD4870m.

Also, using the same version we have is only logical.

The main thing to keep into account is merely changing the thermal interface material for the HD5870m doesnt affect temperatures nearly as much as this mod.

Its quite a lot different.

-Ash

 

Yes, I am not trying to compare them I'm telling you that the 5870's run cooler than the 4870's which is fact, yet my temps are not that much higher than yours. and the version does not make that much of a difference there is the change log posted above no improvements in the way furmark stresses your GPU.

 

Well, do you care to humor us and download the latest version?

Thanks

-Ash

 

Sure will post back in 15 minutes

 

Dude, if your temps are that low, that's a good thing. I'm not calling your bluff.

 

Second run 1.8.2 negligible change, I don't want to seem like the bad guy who is trying to call someone out. I was just sharing my experience with the GPU temps, and other solutions for ppl like me who still have a 3 year (or whatever amount) of warranty left, and would not like to compromise that... really that was all I was contributing an experience and possible other solution that some ppl might find helpful.


Uploaded with ImageShack.us

EDIT: I forgot to mention I changed my resolution down to match the OP's run, however, i have seen that the Resolution didn't affect my results drastically. Also, the 4870 is a 55nm 65W chip the 5870 is a 40nm 50W chip so it should run cooler. (that is why I wasn't comparing)

 

4870's memory controller doesn't get hot because it's a 256 bit bus using gddr3. The 5870s use 128 bit bus + gddr5, hence the higher temperatures. They can't really be compared.

 

What are your temperatures like @ idle?

I'm running an external screen almost constantly (so 2x 1920x1200), and the temps on my 4870 while working (non-graphics intense stuff) are hovering around 60C. Therefore the fans are constantly running although not on high. When I disconnect the 2nd monitor the thing cools down a bit and the fans slow down also.

I'm wondering if it would work to put an appropriately sized piece of copper sheet between the memory and the heatsink.

Anyway, I'm considering this mod just to get rid of the fans running while im not even gaming.

 

Ram temp is a little higher.

But check the SHADER CORE TEMP!
This is using actual washers with 2-3 twist on each screw (so only barely grabbing) so as not to crack the core.

Gotta go to dinner now, but that was a fairly long run. Also, the fans never kicked into high speed.

I will never have my gpu fans running at 100% ever again during gaming.

Ill post idle temps when I get home.

Adios!

-Ash

 

My idle temps @ stock



My Idle temps @ 675/1150





Also, Here is another burn test with an extreme OC at 675/1150




and another at 625/1100


And all I did was apply Tuniq TX-3 to the Core
Uploaded with ImageShack.us

You ran it for only 4 minutes and 6 seconds run it for atleast 15 minutes

 

Wow, those are nice temps. I did a quick 5 minute run with FurMark 1.8.2 @ 550/887.5 (so stock clocks), and temps are rising towards the low 90s. This I do not even mind even though it could be better.

The thing I like most are your idle temps.

My M17x is going to be replaced, so once the new one is in, I will retest the temps and at least put some decent past on the core. Maybe experiment a bit with copper sheet to cool the memory as well...

Oh, and thanks for your help!

 

Can I get a show of hands (you know.. not physically) of who's undertaken this mod by Ash?

Of those people, can I request you guys (and girls.. no discrimination) to do me a favour. Specially you Ash. Do a Full Vantage run (it would be helpful if you had a pre-mod run score.. if not, its fine) and push the system. See what score's you get, see how high it helps you get. Use external cooling if you want to, but the test would be more precise if used without.

Last night, I pushed my computer to clock scores of 932/1203 without external cooling or removing any parts of the computer, temps were high, but the score maxed out around 18300 only (as opposed to my 19200 with external cooling).

So to get a true indication of the mod, try and push it as hard, and see what score's you get. (My bios is at 5% OC).

 

hey mate, typing from my n900.

i dont need to run furmark for any longer, because if you take a look at the graph you will notice temps are normalized.
no point beating down the cards more than i need to. it doesnt take long for furmark to normalize temps - usually about 4 minutes.
anything longer after you have normalized temps, and all you are doing is electron migration.

also, please try keep discussion on topic, e.g. the hd5870m mod or the cpu mod.

-Ash

 

I think it's just Ash and myself at this point Jubbing.

 

Although I have replaced my CPU in the M17xR2, I have not messed with the GPU card, and therefore, some of the items that are referred to in the first post require much imagination on my part.

It would be nice to have a post with more detailed pictures, especially of the later methods using washers, etc. And some labeling on the pictures that point to what a retention clip is and where to put the washers, etc...

I'm by no means new to hardware, but I'm just trying to imagine what I need to do before I take the whole darn thing apart! I'd be willing to do the mod and report back, but just need more detail.

Also, instead of duct tape for the ducting, how about a single wrap of electrical tape? I believe I have even seen Dell use electrical tape to close off gaps between the duct and the fins on other models of Dell notebooks.

Pictures, people! Pictures!

Good topic, btw!!! +1 for the OP!

 

I hit 100 C in game, i got a video card replacement and now i have over heating on my left hand, and the machine hits 98 C yet !

 

Well you HAVE done the testing Turbo.. have you noticed any difference in performance?

 

Little bit off topic, but how do you disable processor cores on the i7? I know it supposed to be possible, but I have never seen how to do it.

 

I've been planning on doing this if no one else has when I get my new mainboard tomorrow, I was thinking about eletrical tape as well instead of the duct tape, it's more heat resistiant and leaves less residue. the 'retention clips' are acutally just c-clips that keep the screw connected to the HSF, if you had it apart it's impossible to not see them and know exactly what the OP is talking about.

you're not going to see any difference in performance unless you've been clock throttling due to heat, and then it's probaly a defective card. I would expect vantage to go up 50-100points tops, which really isn't enough to measure between run-run differences anyway.

 

Well My apologies for ever posting thought I had something to contribute to this, however, since its been made apparent that I don't you should rename the thread to M17x R2 5870m full internal cooling mod as I thought this could apply for the 4870. I wasn't trying to steal your glory bud. Just offering another solution. I have also been trying to find how 128 bit bus GDDR5 runs hotter than a 256 bit bus GDDR3 card can't find anything about that seems to be if they had equivalent bus types that the GDDR5 runs hotter. Maybe Joker can explain it better?

 

the heat issues needs to be addressed at Dell.

IS anyone from DELL listening.

The problem will be fixed, when Dell comes out with the "Brand New Alienware M17 r3" or maybe the M18 r1

 

man, we need this kind of info from guys like you who own 4870s just as i and a lot of other people do!!! Soon these 5870s will be replaced by 480Ms and these new owners will be in the same position as us. At least the 4870s don't face any major driver problems, so while we're playing our games, there's others whining about the flickering and the lack of 5870 driver support. Let's open our own thread and discuss our 4870 overclocks and cooling mods dude.

 

Sound like a good idea(not sure if a thread exists), but I will create one later on when I get back home.

 

there is no heat issue that I'm aware of, if you think mid 80s is bad, sell your M17x and go buy a G73JH and enjoy your 100-110+C temps and GSODing cards. The only time these things see 90-100 is in furmark, which is perfectly resonable since there's no game that I know of that is as intensive as furmark

GDDR5 has many improvments that allows it to run a lot more effeciently at the same clock speed of GDDR3, the downside being that GDDR5 ram runs hotter because it works harder. It's not really the MCH's that run hot, rather the ram, they halved the number of MCH's to save some die space and some core heat, but it doesn't effect performance since GDDR5's bandwidth is effectively double that of GDDR3 at the same clock speed.
so in short, it's not the MCH's that create the heat, it's the ram modules, Dell uses Pads that are too thick which cause poor contact on the modules and even effects core pressure, so by leveling the HSF to make better contact on the modules and core, it helps temps immensly.

 

4870's shouldn't even be hitting 80C @ stock clocks. Mine don't even hit 70C in furmark at stock speeds.

 

If you guys wanna use a great TIM I highly recommend Tuniq TX-3 for this mod(considering it can handle low pressure HS)or not.
there is a link with specs
Tuniq TX-3 is the best thermal grease

and here it is in action vs AS5

YouTube - ‪Thermal Paste Showdown: Tuniq TX-3 vs. Arctic Silver 5‬‎

and here is how to apply it
YouTube - ‪Tuniq-How to apply TX-3 thermal grease‬‎

Only cons is that it is more expensive than other TIM's and difficult to apply, but like I said before put it by a light bulb for 20 mins and it will help alot with application.

 

GREAT thanks!!!!

 

Thanks for the input mate.

Just bear in mind TX-3 wont be suitable for the GPU die, only the ram. You need to have 2 different thicknesses of heatsink material for the best result.

So TX-3 is a great suggestion for the ram, but stick with a thin one for the GPU like AS5.

-Ash

First post updated with a boatload of info and step by step guides.

-Ash

 

Lozz is right. I really don't see a difference in performance. I've got a hell of a laptop cooler, and my fans don't really come on in most of the games I play now. I was just playing Fallout 3 on maximum settings and my fans didn't come on. As someone mentioned earlier, if there are no ill side effects to this mod that we figure out down the road, it will lengthen the lifespan of the cards.

 

There wont be any difference in performance from doing this mod, but it will allow you to run higher voltages 24/7 and higher overclocks with it.

-Ash

 

i agree with this, therefore it will lead to higher overclocks in the end.

 

These are the temps on my m17x while running Furmark stability test after replacing thermal paste on the GPU's with Noctua NH-2 paste. I left the pads alone and did not do any ducting mods. Ambient temp is 80F and I am using a Coolermaster NotePal U2 cooler.

 

are you using furmark multi-gpu? Just wondering if Hwinfo32's report of 40% GPU utilization on the second GPU is accurate.

 

I was running the Multi-GPU test. I'm not sure whats up with HWinfo32. But the temperatures are at least the same ones I get with GPUZ. Also bear in mind that this is the "stability test" which the OP ran. I only checked off what the OP did to produce similar results. Temperatures are higher when I run the extreme burn in tests.

Edit: Here's the another test. Also, I realize that my original run was at a higher rez. Here's a new run:
I am noticing that my clocks are reading only 300/400. I need to find out what's going on...

Found out what was wrong. I disabled Powerplay. I will get new temps up and take down the earlier runs. 2nd picture is with the maximum clocks running the same test the OP did. You can see the modifications he made have made a difference to his machine.

 

rez difference between 640x480 and 1200p is very small so we can overlook that. Still decent results. I'll have to start using HWinfo32 looks like so I can get readings from both cards at the same time.

 

Lozz, I just open two different instances of GPU-Z, each showing a different card from the dropdown menu at the bottom.

 

I figured you could so something like that with it, since I'm getting my mainboard installed right now though, I couldn't test it Thanks.

I've just contemplated something, obviously the heatsink is capable of keeping the 5870 very cool, which means they cooling system in the M17x Is better than we imagined, good enough for SLI 480M's maybe? With a simple thermal Pad change, maybe a thin sheet of copper on the backside of the card and slightly more tension on the bracket, it certainly seems that way.

 

Will a pair of 480Ms work in the R2?

 

That is a question that no one knows the answer to but it seems that the answer is NO due to cooling issues.

 

with stock cooling pads, no. However as I was saying, with the improvements the OP made from essentially doing nothing except removing the defects in the design, I can't see how it wouldn't run two of them. The improvements are minor and composed of nothing except common sense, a M17X R2 GPU HSF would have no problem running a 480M with those improvements I would easily imagine. All Dell would have to do is adjust the backplates a bit and use thinner Gap Pads, easy stuff even on the Scale they make the things at.

 

Not to mention adding a PSU capable of supporting such a set up.

 

yup, Can't comment on whether the mainboard is designed for the extra capacity, I know it's only a 3 moffset power design, and if it were loaded with a fully unlocked 920 or 940XM that can up clock all 4 of it's cores to 3.2/3.4ghz *and* running 200W worth of GPUs not sure how it would handle it. However, most of people's concerns have been with cooling, The benchmark thread is full of people that prove that the 3moffsett design has some room left over to allow extra power to the overclocked components. Really you would want it to be able to handle 70-80W, so you're not maxing them out all the time and you can still overclock decently.

 

With Ash's cooling mod, plus whatever Joker is cooking up over in the other thread, PLUS two 240w adapters in parallel, we're golden right!