Aw m18x R2 Dual 980m SLI upgrade!!

Just the default performance mode. Full test.

 

Wait wait wait... DSR... WORKS?
@octiceps @Ethrem @n=1 @Mr. Fox

 

I dont think it's related to Dual PSU. On single PSU you can push 980ms upto 1300Mhz (CPU @ 4.4Ghz). Now tht stupid throttle gone, You can get stable performance. Try different driver.

 

I wonder if that was an oversight. I'm in bed right now but maybe I'll get to test it tomorrow. Watch it not work on Maxwell lol

 

Yep, DSR works, just tried it! I ran F1 2012 benchmark at 3K downsampled to my 900p resolution, I turned off AA due to downsampling - it looked rubbish (bit blurry & some small jaggies) in comparison to 900p + 16CSAA and average fps dropped from 86 fps down to 54 fps. Also, you can't use Custom Resolutions when using the DSR factors, so I'm limited to 60Hz, my 78Hz overclock is not available. I'm going to stick with 78Hz and 900p, no DSR factors for me! You have to turn DSR Factors on in the Global Settings part of the NVidia menu.

 

Works, but at what cost? I don't give a rat's butt about DSR if my GPUs throttle when overclocked. The new 353.00 driver is utterly and completely worthless on my three 780M systems if I have to run stock clocks. I'll test it later this weekend on 980M SLI, but I'm really pissed that this is the third garbage driver in a row. Went back to my 345.20 DT CUDA driver mod. Having the ability to run an in-game framerate killing DSR junk isn't worth sacrificing benching. There's not any any feature worth having if the trade-off means throttled overclocking.

NVIDIA GeForce Driver 353.00 Discussion Thread

 

I dont know what happens, now this driver is throttling but yesterday was ok with no throttle att all.

 

I am getting some throttling too with the hotfix driver. Its defiantly better but it still there, it doesn't happen while gaming, it only throttles during 3dmark 11. I guess when you push ur gpus to its limits it throttles anyone else experiencing this?

Nikkos, to get this score did you need to over volt your cards?

okay so I cant get a stable overclock. Stock clocks are rock solid with ZERO throttling. when you overclock it keeps throttling to stock clocks so Im just gonna give up overclocking my cards till prema releases the magic vbios. check this out, tell me your thoughts guys

CPU @ 4.3 but temps are getting a bit on the high side.




not so bad for a noob like myself

I went back to my stock vbios. gotta say, it seems to be working better with the newdrivers (hotfix 353.00) for me.
im gonna wait till prema releases the official vbios for now im gonna stick with the stock vbios of my cards.
I think im capped by my power brick by now. I got 16800 a couple of times then I unplugged my mouse and everything I had connected on the usb ports then turned off my keyboard lights and dimmed my monitor and I got 17364... FYI this Is probably the max performance you can get with a single brick for people interested in the upgrade of the 980ms but (not interested on dual brick mod) like myself.

outstanding video cards, you can feel the power on every single game.

Also guys, my GPU 2 gets hotter than gpu one during gaming.... I would guess is normal due to the second gpu is closer to the CPU and that would create more heat. Is this a normal behavior?? the one is about 70c the other about 74-76c

 

Upgrade to dual-brick power™

 

No such issues on my severely overclocked 670MX. It's a Kepler card though (and a different system!), and overclocked it only pulls about 100W (a bit less) from the measurements/inferences I've taken with a KillaWatt, so perhaps it doesn't pull enough Watts to trigger any throttling in the driver. I checked the vBIOS, the modified vBIOS I'm using is good for up to 140 something Watts from memory.

 

I think that may actually have something to do with it... distinct possibility.

Good news... that minor difference is expected. No need to worry unless the temps are way off for GPU2. As far as a slight difference this has always been the case for me with the M18xR1/R2... 6970M CF, 6990M CF, 580M SLI, 7970M CF, 680M SLI, 780M SLI and 980M SLI... GPU2 always runs warmer than GPU1. It is because the heat sinks are slightly different and the fan intake is oriented differently underneath each of the GPUs.

 

How many watts can Slv standard modified vBIOS deliver with a max voltage on gtx780m ?

 

I'm not really sure. I haven't used that vBIOS for benching since it was first released for 780M. I tested it and confirmed that it worked fine, but I never checked power draw. I have another person's M18xR2 that I am working on. It has 780M SLI with that vBIOS, so if I have time after I am done working on it I will test and see how it compares. I will also test it with the 353.00 driver to see if it throttles using that vBIOS compared to the one I am using.

 

FYI - in case you are still having some issues, this might be something to investigate.

3DMark 11 pulled 675W on my UPS digital meter in the run below.

http://www.3dmark.com/fs/4903861

http://www.3dmark.com/3dm11/9849353

 

Wow, 675W, that's more like desktop levels of power (most my system's ever drawn in gaming or gaming benchmark scenario is 180W - SkyDiver benchmark)! Were you not getting system freezes with 353.00 though? I was having Chrome hangs, and I noticed 353.00 was causing red DPC latency spikes every 15s - reverted to 352.86 & now all in the green for DPC.

 

I was quite sure about that. Thanks for the confirmation.
I am waiting to replace the card which refuse to play and after that I will try to find the problem. 675W is too much, it was a peak or it was sustained?

 

After your screen goes black, does the laptop reboot by itself?
Those numbers are insane. I can't even get close to them. Every time I try to do such a thing, the system reboots.
Do you know if this behavior is the same for all the 980M owners? Or by the other hand it depends on which one you get?
This is making me crazy I can not keep testing vbios as my system is not stable. I get rid of good part of the issue swapping the cards of side but it is still there.
I am asking Eurcom to replace both of them but... I do not know man, I am not quite happy with the cards. I know it is not Eurocom fault but if I can play with the cards even a little bit is not worth to me anymore.
PS: Do you mind to record a video with the reboot issue on it? I will do the same and maybe we can find any info from each other.

 

Yes, I was getting system hangs and Chrome freezes also, but only with 780M. Has not happened to me yet with 980M.

Peak, but sustained about 3 seconds in 3DMark 11 Test #1 only. Most of the rest of Test #1 is in the high 500W to low 600W range. The rest of the benchmark is typically in the high 500's at those clocks. But, WAY OVER 330W most of the time. Same for Fire Strike.

No reboots really... just black screen freezes. I think I had a reboot once, but because I was pushing the 4930K too hard trying to force it to run 4.7GHz on all cores/threads. I cannot get anywhere even remotely close to those numbers except with AC cooling. 980M is not a cool running GPU. The core is cool running, but that's not the only thing to be concerned about.

 

@Mr. Fox So to really tap into 980M overclocking you want really low ambient temps so that the whole cards has a chance to cool off?

 

Well, I don't know if what I am experiencing is affecting all 980M users that want to do extreme overclocking or not. @johnksss and @Santander both doing their extreme overclocked benching on AC, and the only way I can do it is using AC cooling. Something (not sure what) HAS TO BE overheating without AC when using an extreme overclock/overvolt. I have no issues with lower overclocks and voltage, and the issues go away with AC benching.

A week or so back Santander did an experiment for me as a favor and tried to overclock the 980M SLI setup in his Clevo without AC and his crashed immediately using clock speeds that he always used for benching without issue on AC. The temps you can see (core is the only sensor) are always fine, but something else obviously is not fine. The AC cooling is only a workaround for extreme overclocking. I game with stock clocks, so maybe I should not burn any more calories worrying about it.

Just connecting the dots here, but I'm not sure you can get ambient temps that are low enough unless you live in a meat locker or an arctic climate during the winter.

 

Makes sense, I would hope that the crashes are from a sensor shutting the card down not some little chip giving up/failing when the temps get too high....
Haha there is no snow in Western Australia but it is winter so 0-10C + fan is about as good as I'm going to get, still it makes a big difference with CPU overclocking so hopefully it will have a good effect on the gpu's once I get get all the bits for my dual PSU and the new vbios is released.

 

Well, it doesn't seem anything is hurt, so that's good. The additional benefit of cold temps is lower voltage is necessary for CPU and GPU. 0°C (32°F) will be plenty cool for benching on the patio during the winter... just run the fans full blast and prop up the back end so it can breathe well and that will work really great. I used to do that when I lived in a place with nice cold winter months.

This is what I use: De'Longhi Pinguino Portable Air Conditioner - 12,000 BTU
It has a dehumidifier for those that are in humid climates and it makes my office nice and chilly for work, too. I put a clothes dryer duct with a damper (to keep bugs and lizards out) in the outside wall of my closet to make it permanent. When I don't need to use it, I just slide the closet door closed. But, the temporary window outlet works fine, too.

 

What do you think about making a peltier cooling pad for the laptop? Do you think it will do the trick?

 

You just have to be careful with the lag it introduces. Also you will need a brand new heatsink to cope with the extra heat it produces.

 

Did your mod to the card (MOSFET/VRM) fix the black screen issue? May you share the info?

 

Yes it did fix it, a shared it with Mr. Fox and I'll leave that up to him.

 

I am planning to buy the chips, a hot air soldering station and other supplies very soon. I'm a bit cash strapped right now, but should be able to do it fairly soon. Now that I see my cards still overclock fine when I keep them chilly on AC, I am more motivated than I was before. I was getting pretty discouraged by the black screen freeze thing and was concerned the GPUs were bad. I'm very glad there is nothing wrong with them other than being really poorly engineered by NVIDIA.

I have some old AMD MXM video cards that are dead and I am probably going to practice on them to teach myself how to do this. I have an M18xR2 motherboard that needs to have a chip replaced on it, so if I can master the process and find success, then I will also fix that motherboard (already have the new chip) and then look into replacing the BIOS chip on the Alienware 18 with a removable socketed chip programmed with an unlocked BIOS.

It looks like the MOSFETS on the 18 motherboard are lower capacity than the M18xR2, so I may try replacing those with higher capacity MOSFETS to match those on the M18xR2 as well. They look exactly the same, just seem to have a lower capacity. I need to research that more to confirm it. But, it may be why the 18 does not benefit from the dual AC adapter like the M18xR1 and M18xR2.

 

Are you talking about the CPU power phases or the AC input for the MOSFETs? The CPU ones would not impact it too much.

You do have to be careful about how exactly they are rated in all areas, do you have links?

 

I have not looked them up online yet. I need to look at a schematic also to be sure what I am looking at are the AC input MOSFETS rather than CPU phases. That's why I mentioned I need to research it more. I'm not that far along in the process. The 4930MX I have in this 18 is a really fantastic overclocker, but being only able to overclock the snot out of one GPU with a measly 330W power limit is a real deal killer that keeps the 18 from being as amazing as it should have been.

 

It is this you want to beat now when you finish modify the AW18? http://www.3dmark.com/3dm11/7593998 Keep 3Dmark11 record of all Haswell MX / mq mobile processor's and gtx780m sli.

 

Wow, I thought I got a good OC with 1058/6000 and 3.8GHz on firestrike... here's 1111/6000 with 4.5GHz on a single brick too XD.

 

this maybe incorrect place but can't find the right thread pretty sure its lying somewhere in this forum but here goes.

a friend of mine interested getting a single 980m, not dual (it'll be nice to know if dual 980m would work or not in R1 too) installed into his m18x R1, not R2. afaik I read up it was possible but can't find out how and where the thread is located. i told him 780m is definitely doable but unsure about 980m and i dont have a whole lot of time to check. any helping pointing me to the thread? would process be similar to m18x r2?

 

As far as I know, you can't use it at all, because due to the SLI nature of that board, you can only choose either dGPUs or iGPU with its MUX switch. I'd say flat out it can't be done to a M18x R1. 780M SLI is the best upgrade for that machine.

 

Dear # Mr. Fox, regarding one of the posts where you show the Thermal pads distribution (Page 218) I had a question regarding the 3mm pad.
The GPU I bought had all pads pre-attached on the heatsink, the 3mm one was a little bit uneven with the large area next to it to the left above the main GPU.
Is this normal or it should be perfectly even so that the 3mm pad is even with the 1mm pad. I really hope I explained it well

 

I'm going to take the cards to a lab to do the mod. Do you have any further information to point the lab's guy?

It is very cheap to do the mod. 20-30€ both cards. It is called SMD welding if I am not mistaken.

 

980M (single or dual) will not boot in the M18xR1 motherboard. Same applies to M17xR3. It must have full UEFI support and the Sandy Bridge motherboards don't have it. I tried single and dual, and single with Optimus mode.

Don't leave the pads on the heat sink. It is common for them to not be applied precisely where they need to be when then are left stuck to the heat sink. Gently remove them and apply them directly to the components on the GPU and then everything will be aligned correctly. Avoid handling them with your fingers. If you have an Xacto knife or tweezers to lift them off the heat sink and nudge them around, try using those, or a small flat-blade screwdriver to move them around, but be careful because you will tear them up unless you go slow and gentle.

 

thanks, please excuse my noobness. Once I install the GPU will I be able to switch back to the integrated Intel GPU? will Fn+F7 still work normally?

 

Maybe. Single or SLI? I am assuming you have an M18xR2 since FN+F7 does not apply to AW18. FN+F7 switching usually does not work with one GPU installed. It may or may not work. If it does not, you can still go into the BIOS and manually change from PEG to IGFX. That's what using FN+F7 in Windows does.

I just ordered the chips to do the 980M hardware mod. Now, I just have to figure out how to do it myself, buy the tools needed to do the job and work up the nerve, or find someone that has the tools and experience to do it for me. I hope I get the same good results that @Meaker did.

 

What are the benefits from the hardware mod? Higher overclocking?

 

Yes. No other benefit that I can identify. If you are not going to do any number-chasing benchmarks, then this mod is probably not something that needs to be considered.

My GPUs work flawlessly up to around 1325 on core. Above that I have to use AC cooling to keep the screen from going black because something (not the core) is getting too hot when heavily overclocked. I am hoping this mod fixes that issue. As you can see from the benches I posted the other day, with AC cooling, 1450/1450 is stable.

 

oh I see... thats pretty high overclock. I manage to pull Over 17300 on 3dMark11 with my current setup cards oced core+120 memory +400 CPU 4.3. I am pretty happy with the results. i Noticed there still some throttling specially during benching, But it doesn't throttle below stock clocks(3dClocks I think 1036Core). During gaming there is hardly any throttling and when there is it doesn't really affect it much because it throttles to stock clocks. These cards are absolute beasts for sure. I am running stock Vbios and waiting for premas release of the vbios, hopefully that will stop the throttling for good.

 

Be aware memory clocks may be impacted. It certainly made my memory volt mod pointless.

 

For exceedingly hot environments I'd say they could extend the lifespan of the cards, even at stock. I've seen such temperature variances that I can be sitting at 65c on my CPU at night and 80c on my CPU in the day; just environmental temperature alone.

Of course, it's a drastic mod just for a "longevity" bonus, and most of the world that uses machines like these in much cooler and less harsh environments... but you know.

 

thats pretty damn high for a laptop man i bet it cooks the keyboard lol

 

Cooks the keyboard? What O_O.

Sir you must be new to Clevo, there is no "cooking the keyboard" XD

 

What was that old saying? " If you can't take the heat, stay out of the kitchen."

 

Or buy a portable AC

 

@Mr. Fox, Any idea about that AW18 BIOS hardmod will allow the Maxwell to be pnp with legacy or still latched to the full UEFI ??

 

Nope, almost positive it won't work in Legacy mode. Doesn't work on the M18xR2 except with UEFI and no Legacy/CSM and the 18 behaves exactly the same as the M18xR2.

 

tbh im waiting for next gen mobile GPU maxwell with more cores, it just seem to me the current one not powerful enough . or wait for something coming from amd.