MSI 1070 Longevity? Aetina 1070 Viability?

[Naberius]

Hi, all. Two questions in one post...

 

First: Did anyone else have their MSI 1070 flame out after a few years? I ask because...

1. The first one (that I purchased in 2019) died in January 2022 (graphics output artifacting, laptop acting like it was having a heart attack)
2. When I looked around for a replacement, they were seemingly impossible to find; Even Eurocom said that their on-hand stock didn't pass tests and had effectively gone bad on the shelf.
3. The second one I purchased (by... cannibalizing an MSI laptop I picked up on Kijiji) died in the same way as the first... last month.
4. In both cases, I don't think I missed anything in butting-up the components against the heatsink; The first one I used silicone pads... The second one I used K5 Pro... Both cards lived relatively cool lives never going above 65 degrees under load...

 

Second: Did anyone try out the... standard-MXM-sized-1070 in their M17xR4 or 17r1? I believe they were an Aetina design (could be wrong on that). They're seemingly going down in price on eBay... And I reckon one of those might be a reasonable (albeit somewhat slower) fit? 

[Tenoroon]

May I ask what revision MSI 1070's you had? If you had the revisions 1.0 cards, those are of worse build quality in comparison to the revision 1.2 cards and from what I can remember, have some problems operating at elevated voltages.

If your cards were revision 1.2 cards, then you might just have some bad luck. However, with how old these cards are getting and how they were built, it might just be the beginning of the end for them sadly. 

I've tried two of those standard MXM 3.0 1070s in my 17 R1, but they never worked well. If you can find one for cheap enough and you have a refund window, go for it, but the two I tried never worked right and didn't get along with the BIOS all too well.

[Naberius]

First was a ver 1.0

The second was... a ver 1.2... Interesting.

 

We'll see how the standard-sized card works - a vendor on eBay is blowing them out for $200usd with a 30day return window.

They've apparently sold... like 40.

https://www.ebay.ca/itm/164357449828

 

[Csupati]

I used gecube  1060 , p3000, p3200, ebay china 1060 , every one of the still working .

Probably you cut out the base for the Msi card to fit,  there is no chance its make contact with the base,  you fully dissssembled the laptop and no shaving left in it?

[Naberius]

Correct - I removed a full section from the laptop chassis to make room after a full disassembly. Two 1070s died in the same way (over the course of years). Whenever a GPU died, I was able to roll to a 970m I had lying around without issue.

[Csupati]

3 hours ago, Naberius said:

Correct - I removed a full section from the laptop chassis to make room after a full disassembly. Two 1070s died in the same way (over the course of years). Whenever a GPU died, I was able to roll to a 970m I had lying around without issue.

Whats the full build ,like cpu rams display? What psu are u using?

[Naberius]

i7-3940XM | MSI GTX1070 | 32GB DDR3 | 120hz screen | 330w PSU

[Csupati]

Im thinking about you are doing something wrong with the termal pads,  van you do a pic how you done the thermal pads?

[Naberius]

It's entirely possible I fouled something up in an unobvious way. No pictures of the pad placement now though -- The first card was installed in 2019 and died in 2022. It was a conventional thermal pad placement covering the parts in the VRM section. Because of the conspicuous living death it suffered (wherein the card was still semi-operable but would flake-out as soon as any sort of load was put on it), I assumed I'd missed something in the pad placement so... with the second card I just sludged the area with k5 pro and called it a day; Not much chance of missing anything in that case but... also no guaranty that the act of sludging the whole area didn't conduct heat somewhere harmful (unlikely afaik). Given that these cards seem to be going the way of the dodo in general... I'm finding it less-so odd that both cards died and moreso odd that they both died in the same way.

 

Anyway... new standard-sized 1070 arrived today. I should be able to install it on Tuesday...

[Csupati]

5 hours ago, Naberius said:

It's entirely possible I fouled something up in an unobvious way. No pictures of the pad placement now though -- The first card was installed in 2019 and died in 2022. It was a conventional thermal pad placement covering the parts in the VRM section. Because of the conspicuous living death it suffered (wherein the card was still semi-operable but would flake-out as soon as any sort of load was put on it), I assumed I'd missed something in the pad placement so... with the second card I just sludged the area with k5 pro and called it a day; Not much chance of missing anything in that case but... also no guaranty that the act of sludging the whole area didn't conduct heat somewhere harmful (unlikely afaik). Given that these cards seem to be going the way of the dodo in general... I'm finding it less-so odd that both cards died and moreso odd that they both died in the same way.

 

Anyway... new standard-sized 1070 arrived today. I should be able to install it on Tuesday...

If you gonna get a regular "quadro layout" card then those are critical to be cooled, if you want to cool the capacitors you can do it ,but not required.

[Naberius]

All right - here's the install report--

 

Bone-stock install: I used the original 3-pipe heatsink because the 5-pipe I have on hand is tailored for MSI cards (and as a result, misses the right-side mosfets).

Regardless, performance and temperatures appear to be on par with the MSI 1070(s) I had previously (17XXX gpu on fire-strike, about 65 degrees under load).

 

That said... quirks... there are some new quirks:

1. Display-port to Ultra-wide monitor at 60hz: As expected
2. Display-port to Ultra-wide monitor at 120hz, 144hz: The display locks down and refuses to show a picture _even after_ it automatically rolls back the refresh rate change. I had to clear my registry entries to get the monitor back.
3. Display-port to VR Headsets: No picture. Steam VR throws a display disconnected error.
4. HDMI to UItra-wide monitor at 60hz: Not detected
5. HDMI to 1080p tv: As expected
6. Internal/attached display at 120hz: As expected

 

Not sure if the above could be sorted out with a different vbios but yeah... display connectivity is not great. For 200usd it's not bad but the laptop's utility is definitely down some (compared to the MSI card).

 

 

[Csupati]

13 hours ago, Naberius said:

All right - here's the install report--

 

Bone-stock install: I used the original 3-pipe heatsink because the 5-pipe I have on hand is tailored for MSI cards (and as a result, misses the right-side mosfets).

Regardless, performance and temperatures appear to be on par with the MSI 1070(s) I had previously (17XXX gpu on fire-strike, about 65 degrees under load).

 

That said... quirks... there are some new quirks:

1. Display-port to Ultra-wide monitor at 60hz: As expected
2. Display-port to Ultra-wide monitor at 120hz, 144hz: The display locks down and refuses to show a picture _even after_ it automatically rolls back the refresh rate change. I had to clear my registry entries to get the monitor back.
3. Display-port to VR Headsets: No picture. Steam VR throws a display disconnected error.
4. HDMI to UItra-wide monitor at 60hz: Not detected
5. HDMI to 1080p tv: As expected
6. Internal/attached display at 120hz: As expected

 

Not sure if the above could be sorted out with a different vbios but yeah... display connectivity is not great. For 200usd it's not bad but the laptop's utility is definitely down some (compared to the MSI card).

 

 

You have the same thermal with the 3heatpipes cuz the chinese 1070(this card) is only 90W , compared to the msi 115w.

You could play in the  inf mod with sections, I dont prefer the nv software to automaticly add the card id.

[Naberius]

Re: power draw: Interesting... I'm not sure that's the case for this card. GPUZ is reporting it at 110-114w while running Furmark (which was what the MSI card reported in GPUz also). 

 

This is especially interesting because I had one of the 90w capped 1070s at one point (just before switching to an MSI card) and its 3D mark Firestrike GPU score was closer to 14000 (rather than 17000, where this card is scoring).

 

When I asked the (hk-based) seller which vBios it was running, he said it was an MSI vBios... I didn't press him for more details on that though.

 

(driver is 537.42 with the card manually added to the INF file)

 

[Csupati]

24 minutes ago, Naberius said:

Re: power draw: Interesting... I'm not sure that's the case for this card. GPUZ is reporting it at 110-114w while running Furmark (which was what the MSI card reported in GPUz also). 

 

This is especially interesting because I had one of the 90w capped 1070s at one point (just before switching to an MSI card) and its 3D mark Firestrike GPU score was closer to 14000 (rather than 17000, where this card is scoring).

 

When I asked the (hk-based) seller which vBios it was running, he said it was an MSI vBios... I didn't press him for more details on that though.

 

(driver is 537.42 with the card manually added to the INF file)

 

The more interesting thing is the gpu chip power draw is 60W ..... thats not good compared to the total card power draw 111W . holy sh@t. 

Thats a clevo 1060 mxm that has the 1080 VRM on it . We know that VRM is power hungry cuz its the powerful one just less then was it on the 1080 .

the 3.rd pic is a p3200 in a 17 ranger with the 2 heatpipe heatsink 

[Naberius]

For sure - that said, it makes me wish I had a GPUz screen grab from the MSI card running Furmark;

Without the apples to apples comparison, it's hard to say what is or isn't normal, I reckon...

(This is me hoping someone still has an m17xr4 or 17r1 with the MSI card and can do a quick capture).

[Reciever]

The MSI 1070 has to deal with a lot of heat around the VRM's since they are packed neatly in a single line. I used to get shutdowns on the 17 R1 until I added some heatsinks to the passive end of the heatsink assembly so it wouldnt surprise me if running them hot for prolonged periods of time would cause what would otherwise be premature failure.

[Naberius]

1 hour ago, Reciever said:

The MSI 1070 has to deal with a lot of heat around the VRM's since they are packed neatly in a single line. I used to get shutdowns on the 17 R1 until I added some heatsinks to the passive end of the heatsink assembly so it wouldnt surprise me if running them hot for prolonged periods of time would cause what would otherwise be premature failure.

 

Super interesting - in the M17xR4 with the 5-pipe, it ran normally until it didn't. Given your note about overheating in the R1, the way the card would still post in the xR4 but crumple as soon as load/power was applied, and... this disquieting photo I just snapped wherein the heatsink doesn't actually line up very well with the mosfets... I'd say we have a pretty good hypothesis as to why 2 cards, over an extended period of time, failed in the same way.

 

(this might also make sense of why the first card, which I bricked up using silicone pads, lasted much longer than the second card, which I just sludged with K5)

 

 

[Reciever]

21 minutes ago, Naberius said:

 

Super interesting - in the M17xR4 with the 5-pipe, it ran normally until it didn't. Given your note about overheating in the R1, the way the card would still post in the xR4 but crumple as soon as load/power was applied, and... this disquieting photo I just snapped wherein the heatsink doesn't actually line up very well with the mosfets... I'd say we have a pretty good hypothesis as to why 2 cards, over an extended period of time, failed in the same way.

 

(this might also make sense of why the first card, which I bricked up using silicone pads, lasted much longer than the second card, which I just sludged with K5)

 

 

 

I asked cicichen some time ago if he can make a heatsink that is similar to that of the clevo models of the same era but he was firmly adamant about it not being useful. It was a shame since the heatsink did very well otherwise, I ran my 1070 @ 120w @ 54c with maximum fan in the 17 R1 and later traded the GPU+cash for a GTX 1080 that is in my P750FM now. 

 

Until then however, I would experience just straight shutdowns until I added copper heatsinks pretty much all over the passive end, trimmed them to be flush with the clam shell and opened up the GPU slot to allow it to breath with a custom laptop cooling pad that most "enthusiasts" tend to go. Never had a problem again in terms of GPU health.

[Csupati]

19 hours ago, Naberius said:

 

Super interesting - in the M17xR4 with the 5-pipe, it ran normally until it didn't. Given your note about overheating in the R1, the way the card would still post in the xR4 but crumple as soon as load/power was applied, and... this disquieting photo I just snapped wherein the heatsink doesn't actually line up very well with the mosfets... I'd say we have a pretty good hypothesis as to why 2 cards, over an extended period of time, failed in the same way.

 

(this might also make sense of why the first card, which I bricked up using silicone pads, lasted much longer than the second card, which I just sludged with K5)

 

 

The crumple at load means that its not vrm related did You tried lowering the clocks on the card?. The GPU chip it self has died, im thinking about mounting pressure issue, or the card was bent when it was inside the laptop. 

[Naberius]

On 10/7/2023 at 1:21 PM, Csupati said:

The crumple at load means that its not vrm related did You tried lowering the clocks on the card?. The GPU chip it self has died, im thinking about mounting pressure issue, or the card was bent when it was inside the laptop. 

Ah I should have used clearer language - When I say crumple under load here, I mean that when the card exceeded base clock (which I think is 135mhz - don't quote me on that), the display would start to artifact (primarily additional green pixels) and the system would respond to input much more slowly (though it's difficult to tell whether that's a failure to repaint the display or to respond in a timely fashion).

 

But yeah, you're right / to that, without a smoking gun, it's hard to say what exactly happened without tools we don't have (and... after a point... undue effort).

[Csupati]

1 hour ago, Naberius said:

Ah I should have used clearer language - When I say crumple under load here, I mean that when the card exceeded base clock (which I think is 135mhz - don't quote me on that), the display would start to artifact (primarily additional green pixels) and the system would respond to input much more slowly (though it's difficult to tell whether that's a failure to repaint the display or to respond in a timely fashion).

 

But yeah, you're right / to that, without a smoking gun, it's hard to say what exactly happened without tools we don't have (and... after a point... undue effort).

Artifact usually means Your gpu chip is died, or you have a broken soldering ball under the chip. Maybe a reflow would help, its worth a shot. Or a memory problem, u could go with integrated on and do a gpu memory test under dos, to see thats the problem, but for repairs im not sure woth it.

[Naberius]

2 hours ago, Csupati said:

Artifact usually means Your gpu chip is died, or you have a broken soldering ball under the chip. Maybe a reflow would help, its worth a shot. Or a memory problem, u could go with integrated on and do a gpu memory test under dos, to see thats the problem, but for repairs im not sure woth it.

 

Yeah sounds like tooooo much effort at this stage heh heh (especially with P5000s coming down in price finally). I'll likely sell this dead 1070 on eBay for parts or repair and call it a day. I'll post back if the inexpensive 1070 starts doing anything interesting; Given that my m17xR4 doesn't spend as much time _on_ anymore, ideally it will be the last GPU it gets.

[Ares]

Hello, could you take a screenshot with Gpu-z to see the version of vbios that that card has 1070 of ebay? Thanks

[ÊtaPegasus]

Any news about the "standard" Gtx 1070?

Still working great ? Or do you already have issues with it 4 month after installation ?

[Naberius]

22 hours ago, ÊtaPegasus said:

Any news about the "standard" Gtx 1070?

Still working great ? Or do you already have issues with it 4 month after installation ?

 

TBH I ended up selling it; The issues with driving my high refresh rate monitor and VR headsets really limited the laptop's usefulness. Later, I was able to get another MSI v1.0 1070 for a reasonable price. I added some material to the heatink to make better contact with the VRM row and called it a day.

 

On 12/9/2023 at 2:43 PM, Ares said:

Hello, could you take a screenshot with Gpu-z to see the version of vbios that that card has 1070 of ebay? Thanks

And sorry about that, @Ares - I missed your reply and... have since swapped it out. That said, the seller indicated the card was running an MSI vbios. Hard to say whether he was mistaken or they actually did swap in something that might have given the card a shorter-than-you'd-expect lifespan.