Downgrading the GPU in an M17x R3

Hey folks,

First post here and I'm already asking for advice eh.

Well, I've seen quite a few threads and other posts on other forums etc about the gpu having issues - particularly the 0116 BSOD message.

I've had my m17x since April 2012. It has a 580m within.

It was supposed to be from the outlet site - but that got misdelivered and Dell, to my amazement and to their credit - delivered me a brand new one to match the spec of the outlet one I'd found...

Anyway - back on track, I was watching videos through the hdmi out to the tv the other day as I've done countless times before.

Video stopped and 0116 BSOD. Like others, I rebooted thinking "oh well, a random BSOD - first for everything". But... it happened again after a few minutes - then again.

Steps I took:

1) Updated the m17 bios from A09 to A12. Didn't update the vbios - it appeared to be up to date according to the dell support site.
2) Updated drivers.
3) took it apart and cleared out dust residue and checked contacts. Seemed gpu was ok but no signs of any weakened thermal strips. #

No such luck - still the problem occurs from time to time.

I didn't remove the heatsink as I didn't want to wreck any paste that might be there. I have since acquired some paste to perhaps clean and redo that but I'm a bit wary of trying to remove the heatsink in case I wreck the gpu.

My question... (finally eh): given that I mostly watch videos and only rarely game - do I need a costly 580m replacement or can I get away with a 460m?

My only sticking point is that I don't want to lose the integrated 3000 graphics given some of the upgrade steps for other upgrades mention disabling integrated graphics.

Would a downgrade be doable and allow me to keep the integrated active or should I just bite the bullet and get another 580m (that I'm now also concerned may fail within less than 2 years as this has). I'd have considered a 680m for more power efficiency but thoughts of disabling the integrated graphics turn me off a bit not to mention concerns about relatively quick failures (*if* that happens) for a costly gpu.

Thoughts and advice are welcome

 

Hi there, sure sounds like the 580m has gone. They are dropping like flies recently.

The 580m was the top GPU for the R3 so there should be a bunch of lower GPU's you could go for. All you need to find out is which GPU's dell offered as standard for the R3 model (anyone?). Doing this will mean no bios mods and optimus should work

Have you tried just taking it out, just to be sure it's the NV and not something else?

 

460M, 560M, 580M were all GPU's offered by Dell for that laptop when it comes to NVidia GPUs. AMD GPU's were 6970M and 6990M. If you put any of the other cards (not a 580M) in your laptop, then you may need to rearrange pads on the heatsink for the NVidia GPUs, and for the AMD GPU's you'd have to rearrange the pads & use an X-bracket specific to those AMD cards.

 

Micky - I haven't tried just taking it out. Would disabling it within Windows 7 achieve the same effect?

I'm guessing here but I couldn't see any obvious way to tie the HDMI out to the intel graphics - seemed it was only working via the Nvidia. I have no crashes in Windows with the integrated Intel stuff. I was thinking that if I could enable the HDMI out via intel (if that were even possible) then I could see if it was the Nvidia or not. Assuming here by NV you meant Nvidia - let me know if that wasn't a correct assumption.

Robbo - thanks for the info. That's what I wanted to know - what was possible without bios mods etc or disabling integrated. Reason being is I use integrated for some limited gaming - the only reason I got a laptop with a 580m originally is that it was a good deal on the outlet site. I'll bear in mind your point about re-arranging heatsink pads.

 

Disabling in windows does not seem to have the effect required. The bios still 'sees' the card. I know that when fitting a non-standard GPU disabling the iGPU in windows does not help and it needs to be disabled via the bios.

By taking it out you can see if the HDMI out works on the intel. The NV (yup nvidia ) does take control of the HDMI as far as I know so maybe it will work fine without the card installed?

 

Thanks Micky, I'll try removing the 580m probably tomorrow and watch en episode or two of something if the Intel 3000 can then control the HDMI out. It's always crashed since the first time within an hour of watching a video at some point.

 

You're very welcome .

Good luck, it could probably do with a clean anyway

 

Just as a thought, but the only time I ever got bluescreens was when I damaged my RAM after vacuum cleaning my laptop! Yep, the static killed it! Don't ever vacuum your laptop - use a can of compressed air to blow out the heatsinks. You could run Memtest86+ to see if your RAM is the culprit. Although, it sounds that your blue screen aren't random. Might not hurt to run Memtest86+ though to check your RAM. (I do think it's probably your NV card though).

 

Never vacuumed it - as per yourself, I'm a bit wary of that and always used compressed air.

I was going to check the memory after but I tried physically removing the gpu.

It's now out and this time the intel graphics did take over hdmi out duty.

Watched quite a few videos... but no crashes. Methinks the suspect is the 580m. I'll try carefully removing the heat sink and repasting it before considering a replacement gpu.

 

Good to know that it works fine w/out the 580m.

I guess it could have been getting hot over time so a repaste isn't going to hurt trying

 

If you're looking for a great GPU for that machine, the 6970M is probably the easiest to find on the market and is dropping in price very quickly. Mine still runs as fast as the day I bought it and, while mine doesn't overclock too well, you can generally expect a decent 20% overclock on the Core speeds, which should make it a bit quicker than the 580M by that point. It's only like... what? 15% slower than the 580M at stock? If that.

 

Micky - the paste that was there was pretty grey and seemed somewhat brittle. I cleaned it off as best I could and repasted it.

This time after watching some vids - no crash... I'm still a bit nervous of it so installed Msi afterburner/kombustor - should have done tat before too I guess.

Anyway, not an extensive test by any means, but after 15 mins of the 3d stress test and half an hour of the physx test, the gpu temp never got above 65c. Fan of course varied in speed to keep things level but so far so good.

Imgl- I have no problem with an amd alternative should the worst happen, if its cheap enough. My only sticking point is switchable graphics- would that be fine still with an amd gpu?

 

Seems I spoke too soon. I left the physx test running as it was stressing the gpu more and although temps never increased beyond the range I mentioned - it BSOD'ed with the 0116 error.

I think I'll be looking for a new gpu, although for video purposes, the intel one is sufficient for the time being.

 

The 69xx series has had some reliability issues as a whole though.

 

Oh well, it was worth a try. There is only one thing left to try, and I've only read about it, is to bake the card. I have seen a couple of people say it worked, it seems that softening the solder on the card can fix a poor contact? You'll need to research the correct temp and it's the last thing to try before trashing the card.

For AMD as long as you go for the cards listed earlier that were sold with the R3 switchable graphics should be fine - but you MUST get the x-bracket and screws for the card, the NV one will not fit correctly. On AMD the technology for switchable graphics is called Enduro (Optimus is NV technology).

Good Luck

 

Here's a link describing the baking process, but don't bake the heatsink, remove that before putting in the oven, just put the card in the oven:

GPU Bake again ....... - News - LaptopVideo2Go Forums

 

Just a real quick comment. The 6970m is an awesome card, sad I had to sell it. It does not use enduro but normal graphics switching the old way (better).

 

I tried baking for a VRAM error of course it didn't fix it. But yes it does reflow the solder.

 

I'll try the bake tomorrow evening - it's worth a try I guess

 

Ok, it took me a long while to actually bake the 580m as I was running HDD Sentinel on a friends external HD. One that is pretty much dead but anyway - I digress.

I baked it as per that link - 200f, pre-heated, resting on foil for 10 mins, cooldown etc etc.

However, when I put it back in after repasting the bios didn't even see the card. I took off the back, reseated it but nope - no luck.

It was a last resort anyway - nothing ventured, nothing gained.

 

Oh well, like you say, worth a try...

 

Here is the thing, what might of happened is the solder did reflow but incorrectly. Or something else was wrong and we pushed it over the edge. We may never know... unless there is a burnt place on the GPU.

 

200F isn't hot enough to melt the solder properly - I presume you did actually mean 200C though given 200f would be way to cold for anything!. The Lead free solder used on these chips starts to melt at 217C. If you are going to do the oven option go somewhere in the 215C region but I don't recommend oven baking cards. It is far better to have it done properly by a professional with correct equipment and flux.

Oh and another thing - if the chip itself (the die part) has failed no amount of baking will fix it.

 

Aargh. The instructions did indeed say 200c but I'd left it at 200f. So instead of a possible fix all I did at 200(f) was make the problem worse and consign the card to oblivion considering my mistake. Bah! I was so sure I'd followed the guide correctly too.

Well - I guess I can console myself that it was on the way out anyway.

 

Oh my yes. 200F would do almost nothing. You should of gotten it a bit hotter .. You could try baking the card again .. No harm!

Edit* thanks for the rep! Also be sure to follow all the other steps just modify the temp to around 390F to 400F, It might actually bring the card back to life.

 

You're welcome (rep).

This time I dd follow the instructions but no change. I guess I knackered it first time. Integrated gfx for the time being, discrete when I find a decent deal.

 

Found a relatively cheap 460m as I'm not too worried (at this point in time) about blazing fast frame rates and popped it in - it already came with heatsink etc so saved me a bit of time.

Straight away I went into the A12 unlocked bios...

was my reaction as I saw no discrete graphics below where it lists Sandy Bridge for the integrated on mine. I had made sure it was seated properly of course.

I set up the options back to how they were (switchable) and just exited and went into windows. "...installing new device drivers..."

To my relief, I could see it in device manager. Rebooted and I could see Nvidia Gfx below Sandy Bridge.

Back into windows, installed the latest driver and ran a couple of quick tests - the temperature seems much better on the 460m than on my 580 - the fan didn't even kick in with any force compared to the previous 580m.

It's slower (naturally) but it seems on the surface to work... no crashes through hdmi out too... phew.

 

Glad you got it figured out. And again sorry the baking this time didn't do much :\. But anyway, the 580m as stated a bit has been known not to be that reliable. But glad that the "downgrade" worked and it wasn't a motherboard related issue.

 

Once again, I'm guilty of speaking too soon...

Same issue still present - initial hour or so of hdmi watching was ok, it crashed with the same BSOD on the next episode of a show.

Then, a random different crash (in windows - this hasn't happened before) followed by a driver power state failure message on the BSOD.

On the next boot, it beeped once... but then went into windows ok.

3 or 4 more reboots - no beeps.

I did a quick google and found this regarding the beeps.

Alienware Systems: My computer shows no image, is beeping or the indicator lights are blinking (No POST / Codes Table) | Dell US

If it's accurate, then perhaps it's the laptop mobo?

At this point, I'm glad I only went for the 460m rather than anything more exotic that would have cost lots more but probably got me the same issues. At this point, I'm thinking my old 580m may have been ok given the similar errors with this 460m.

 

Sorry to hear it .

It is sounding like the MB and the 580 was fine, but I'd hate for you to replace that and still have issues.

There is a couple of things I would try:

1. Get rid of the modded bios and put the latest dell one back (default settings). Also remove the coin cell (hold power button down with no power or battery when it is out) to loose any NVram settings that might get 'stuck'. Expect some beeps on first startup as it will think the internal battery has failed (the date/time will be the default).

2. A complete stripdown as if you were replacing the MB (including removing the CPU). Sometimes just the act of reconnecting everything fixes weird random issues.

Good Luck and fingers crossed .

 

I got rid of the modded and put the latest A12 back (default settings). Removed the lithium battery, held the power button and then put it back.

As you said, there were beeps on startup and I had to reset the date time etc. It showed the 460m at the time I was in the bios setting the date time.

By the time I got to windows though, the 460m was gone again. Rebooted, back into the bios and "<not detected>" for the discrete graphics.

So no luck...

However, all is not bad.

I had previously checked the Dell website and added the m17x r3 to the list of 'my products'. I had premium phone support with "NBD" next to it and wondered if that was actually on site support or not with just over a year remaining. Given that the original machine was nearly 2 years ago and also an outlet machine that got replaced by a new one - I didn't realise that Dell had added 3 years of phone and on site support to that too.

I rang them up, dubious as to this being really true. I related my story, that I had replaced what I presumed to be a faulty gpu with the 460m, sent the engineer the BSOD crash logs and the upshot is that tomorrow an engineer is coming to replace the gpu (presumably with another 580m) and also the motherboard (in the engineer's words "just in case").

Can't be happier than that!

 

Woo Hoo, dell come through on the warranty again - more reason to stick with them - and renew warranty to the point you can't any more

The warranty (which comes with all outlet machines as well) is NBD and on-site. In some locations it can be a little work to get this level of service but in general it works well in the higher subscribed countries .

Good thing is that Dell are fully aware that the 580m is a piece of ****. They don't even force you to do a full factory restore if you have this card. Install latest dell driver, hit the card up with something like 'Heaven' benchmark. If it fails replace!

Keep us posted!

Edit: forgot to add, that bit about sending a MB as well happened to me - and is way beyond what is required of them. Strictly speaking if they find more than one piece of hardware that may be causing the problem they can ask it comes back.

Just think about what you want though. The engineer is sub-contracted and often they just replace everything they have been given before even testing it. If you're fine with that then let him loose and watch. Might happen, might not. Otherwise say you want it tested with just the 580m before replacing the MB...

 

I didn't think about what you mentioned on the 580m - but you're right, the engineer didn't ask me anything about drivers being up to date (which they are) or anything - just asked for the logs and examined them and that was it.

Warranty extension... just over a year to go - I think it's definitely worth me requesting that and seeing what they say.

I was confused by this though:

DELL PPS Premium Phone Support NBD 384
DELL Next Business Day Onsite 0-Expired
DELL Getting started support 0-Expired

So I did ask the engineer to check my warranty status if I had onsite support - he said it was fine. I hope so - don't want to be hit with an unexpected bill

I'll post here once the engineer has been though and I've had a chance to run a benchmark

 

That is confusing an looks a little wrong. They separate out the on-site so it can be an option for other machines and countries (sometimes). The AW warranty always comes with on-site and only the accidental portion cannot be renewed/extended (in US yes, everywhere else, no). They will not sell the machine in countries that do not have a local support infrastructure.

What country are you in if you don't mind me asking?

 

It does look odd eh. I'm in the U.K.

 

Thanks, at least I know the 'variances' in UK policy

You got exactly the level of service I would expect here - they do cost around 1/3rd more than the US though.

Also, you might see a lot of bad press for on-site engineers, and I have seen some atrocious messes caused by unskilled people. My guy is simply the best - and I've seen a lot in 20 years of IT. Last time he came first thing he said. 'I see you've got an R4 now, much easier to replace the panel'. Awesome, often in some locations the engineer has never even seen an AW!

 

It's reassuring that then regarding the variances and service cover. Original purchase date was March 2012.

I haven't seen a lot of bad press for them but I did see that other thread where the guy was out and the engineer paid a visit while the wife was home. Screws showing through the case seemed pretty crazy - that an engineer would leave a machine that way. Your engineer does sound good. Likewise, I've had a reasonably long career in I.T. (mainly development) so have seen ineptness and skill both in abundance

 

Yeah, I spent 10 years at the sharp end starting with desktop support and ending up running the infrastructure development team . The last ten years have been rapid management moves ending up as Director (VP) of IT and Telecoms for four countries in North and South America. The PC has always been my hobby so I've stayed on top of it - and it helps when discussing things with the 'propeller heads' and turning a technical requirement into a business opportunity.

That thread you reference was the latest one I was thinking about - didn't link it so as not to scare you!

Mantra: It's all going to be fine - everything is going according to plan, lol

Usually you can extend up to 3 years in the UK and it is cheaper to buy it up front or in a block of years. Mine started around my joining the forum but got extended by a new replacement. When it came up again I just renewed for one year. I'll do it again probably, I just wasn't sure I wanted to keep it that long - and selling one of these without warranty is worthless

 

The engineer didn't come on the day or the day after. I turned on my phone to ring Dell and noticed....

...a voicemail from the engineer wanting to confirm a time. Doh. I'd left the phone off as I wasn't expecting a call - and that it's a phone I use for non-personal/friends/family type calls so not one I often have on.

In any case, it's been rearranged for Monday or Tuesday. This second engineer I spoke to at Dell said they would replace the gpu first and then go from there rather than both at the same time. Understandable I guess although I think it's a forlorn hope that it's anything to do with the gpu. It crashed a lot yesterday - even just in windows while it was doing almost nothing.

So next week for the update...

It's a good job I have a baby m11x r2 for backup.

 

Ok, the engineer came and fitted a 580m and wanted to take away the original gpu. That's understandable but naturally, I didn't keep it when I replaced it with the 460m. He had to send something back and gave me a choice of him ringing Dell to let them know I'd replaced the gpu (which I had already let dell know on my first call) or to just send them back the 460.

I let the engineer take the 460 to send. If Dell do say anything I can point to the original conversation.

Anyway, onto testing. I decided to stress the system as much as I could.

Currently, it's driving both the laptop screen and tv through hdmi out running Heaven benchmark with extreme tessellation (don't know if the 580 supports that) on Ultra mode with 8xAA. I'm also running prime95 to stress the CPU while Heaven takes care of the gpu. It's nearly been an hour and no crashes... So far. Could I have been so unlucky that the 460m might have been faulty too? Time will tell. Leaving it running until I need to use it in a couple of hours.

 

Good news, that GPU swap could have got messy if the engineer didn't bring something back. Unlikely anyone will notice unil it hits the refurb evaluation by which time they will probably just bin it. They are a 3rd party and dell give them a credit for parts returned to them (or charge the customer!).

Sounds fine to me, you've pretty much tortured it! There are GPU programs that go even harder but not recommended for an unattended test (EG furmark). The 580 will be a little jerky in heaven at full tilt but the tesselation is the biggest hitter in that benchmark. There is a hot-key to turn it off and on - now you see just what tessellation can do.

I don't know if the R3 was sold with a 460? If it was then maybe you was just unlucky

 

Even if they charge me, hopefully not an absurd amount - I'll ask for credit based on what I paid for the 460 Hopefully that won't happen.

I left it for 3 hours and it didn't crash in that time. Once or twice Heaven slowed down more than normal - that could have been Prime95 doing something a bit extra though.

After that, I ran through a game for a couple of hours (only 20% cpu, mostly GPU) on full settings at 1920x1080. No crashes..

So far so good... I'm not expecting miracles but it seems perhaps I did have one card that went bad and another card that was a bit odd (even if not bad - perhaps driver versions or something).

You're right about tesselation making a difference. It's very noticeable in quality and a bit less so in speed (but still noticeable). Benchmarks have come a long way from the days of Final Reality. I used to like that benchmark on my first ever 3d card in a tower- the venerable old Voodoo 3Dfx (Diamond Monster version).

Someone earlier mentioned that the 460,560 and 580 were choices for the R3 so I guess I was just unlucky

I've about a year left still in case things were just unusually good today. I'll ask for a warranty extension methinks too. It's too good to ignore as many have said.

 

Ahh, flashback time . The good old Voodoo and Voodo2! My first dip into the Sli world.

Coming from many years of desktop builds with a couple of years off this 'Gaming laptop' thing is a very different animal. I was quite happy with the 90 day RTB I would get with my components, and EVGA having a (limited) lifetime warranty made me lean towards them - it did work, I used it .

Along with a fair bit of PC experience, extended warranty was something I would not normally consider. Dell have come through for me and so many others when it's hard to identify just what part has gone wrong. As long as I can do it I'll pay up. The last renewal paid for itself within three months.

Bleeding edge components come with bleeding edge reliability - and cost the most .

Benchmarking for gamers is now very different. If you're an enthuiast chasing that extra few 'points' then the mainstream are excellent for comparative purposes. I stick to the in-game benches now (A big thank you to the games authors for seeing this need!). You get to see exactly what your settings do in the real world of a specific game. One thing I like about GF Experience is being able to see all my game (27 of them!) settings on a single screen. I can see what I used in other similar games without the need to fire them up

Fingers crossed for you, maybe time for an extended gaming session

 

I never went sli there. I did think about the Voodoo 2 and 3 but never got there in the end. That was about the time when Matrox 2D quality was the pooch's privates so to speak and nothing else could compare. I miss Matrox being a presence in the 3D world. Shame about Parhelia

I never did go for an extended warranty as around that time I was more interested to build them myself. I never bought a pre-built PC aside from one time with a now old (co-incidentally Dell) machine that I used purely for video editing with a probably even older Matrox RT2500. I paid 700 quid for the RT2500! Goodness. Now it's in storage. I too would never consider an extended warranty much like yourself but the experience with this Alienware has been great.

One time too on my m11x R2, the warranty had just expired by about a month or so and it developed keyboard issues. Some keys were just not working... rang Dell, they sent me a replacement keyboard - for free.

I used to be that enthusiast who'd benchmark hard and overclock and cool etc etc etc to get those few points but as an older man - as long as the equipment does it for the stuff I need and game with, I'm good

I have had an extended session - on the Old Republic (not too taxing) and also Crysis 2 on Ultra settings. It seems to be reasonably stable - no crashes so far... fingers crossed indeed and thank goodness for that year or so that's currently left.

One question I have - I often go overseas to visit friends in Canada/USA - do you know if they'll fix things if I'm abroad should problems occur?

 

Cool man, very similar experiences .

For me it was the watercooling that lead to serious OCD and time to back out. Quad Sli, another card just for physics, and five waterblocks!

One of the reasons AW owners get a hard time in the technical forums - you bought an out-of-the-box gaming machine!!?

Looking good so far though, that crysis session should have given it a work out. I have heard of a lot of people having issues with the later drivers and SWTOR so it sounds like you dodged that bullet .

As for warranty, I've used it in quite a few countries. It does vary considerably but basically you have up to three months in any country that dell has a presence. If they sell your model then you get on-site (usually) and hardware part replacement - I had a HD changed in Brazil when I got there for a job and it failed, on a machine purchased in the US, and transferred to Canada . If dell do not sell the model then it's an overseas shipping job