GTX 1070s with inferior Micron VRAM

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

flexy

Diamond Member
Sep 28, 2001
8,464
155
106
No, Micron does NOT have lower OC headroom. The exact opposite, it is actually "switching faster". This issue which by the way is also inaccurately described as "inferior memory" in some mainstream media has to do with the card/hardware and how the voltage is supplied. It is not a memory issue at all.

The issue is if someone wants to apply a high (400 or 500+) overclock on the memory from a low voltage idle state. And the issue can not only be solved with a new BIOS but there is also a workaround using MSI Afterburner. I recommend Google and especially the MSI forum where they guy describes the issue and how to increase the idle voltage uging MSI afterburner so the OC can be applied without the artifacts.
 
  • Like
Reactions: nopainnogain

amenx

Diamond Member
Dec 17, 2004
4,717
3,010
136
^ OK, thats reassuring, thanks. About to get mine soon, so less of a worry now.
 

reb0rn

Senior member
Dec 31, 2009
320
120
116
I flashed Gainward bliss 1070 and tested new OC of micron, now it can do stable +800 but only if I force 1700Mhz GPU clock with afterburner, still 400+ is not stable if P12-P2 states are bouncing up and down

So it was not full fix, and today I see that bios is removed from palit/gainward site!

MSI gaming x is also with micron, have same problem, but still no bios published
+700 possible with P2 GPU state enforced by MSI afterburner

Also I will soon publish MSI gaming X 1070 problem of VRM overheting on normal usage and card blowing up, temps are way higher then top review sites reported and MSI GPU temp is also way of so I guess reading is fake!!! I had on RMA and new one also overheat (VRM)
Guru3d banned me for stating it, I will wait for my IR temperature gun to do more testing

If anyone know how to enforce idle P2 state and not use afterburner curve graph editing, I would like to know, tested Nvidia inspector and it did not do good
 

nerp

Diamond Member
Dec 31, 2005
9,867
105
106
I think this is overblown. And there are different issues being conflated. I think there was a small batch of bad cards with hot VRMs and a few blew up. The black screen/fans spinning issue is different. The Micron memory this is another. All three are converging into this false notion that the 1070 is fundamentally flawed.

If your EVGA GTX1070 or 1080 is a few months old and has been running great, don't worry. If it dies, RMA it. If you are worried it will die, it will probably die anyway so just be ready to RMA. If you order a new one, unless you're getting old stock from the bad batch, you should be fine.

My EVGA GTX1070 SC has been running problem free for a few months. Samsung memory. I am not expecting disaster at this time.
 

reb0rn

Senior member
Dec 31, 2009
320
120
116
VRM overheating is primary MSI gaming 1070 issue, I know they ALL die if used in mining even with <80 power limit, mine died underclocked after a 3 weeks

GPU temp atm is 53C, power usage 53%, fan profil is at insane 70% ad still VRM is damn hot, also no way GPU temperature reading is correct, I have Gainward phoenix 1070 which is by FAR better and cheaper card with 85% power usage and temp and VRM being over 2x less
https://bitcointalk.org/index.php?topic=826901.msg16217534#msg16217534

and I know one guy which in 3 weeks all his six 1070 gaming VRM blowed up, and sure he was not pushing them 100%
 

amenx

Diamond Member
Dec 17, 2004
4,717
3,010
136
VRM overheating is primary MSI gaming 1070 issue, I know they ALL die if used in mining even with <80 power limit, mine died underclocked after a 3 weeks

GPU temp atm is 53C, power usage 53%, fan profil is at insane 70% ad still VRM is damn hot, also no way GPU temperature reading is correct, I have Gainward phoenix 1070 which is by FAR better and cheaper card with 85% power usage and temp and VRM being over 2x less
https://bitcointalk.org/index.php?topic=826901.msg16217534#msg16217534

and I know one guy which in 3 weeks all his six 1070 gaming VRM blowed up, and sure he was not pushing them 100%
The MSI gaming x is a very popular card with thousands sold. Its one of the highest rated @ amazon and newegg. There is no statistically valid evidence to back up what you say aside from your own personal experience and that of another from a forum link you provide. Furthermore, the review of the MSI card @ guru3d with thermal imaging shows VRM temps as quite normal vs other 1070s. The Evga 1070 SC otoh has much higher VRM temps, yet no one has complained about Evga cards blowing up. Always funny when someone seeks to construe their own statistically insignificant experiences as a massive widespread problem. I can understand why guru3d may have banned you. Anyway, this does not pertain to this thread, its about Micron vram 1070s. Should you wish to carry on with this, suggest you start another thread for it, thanks.
 

reb0rn

Senior member
Dec 31, 2009
320
120
116
I agree, I will start a thread only with clear proff
Micron problem is for sure blow a bit over the head, but OC problem with low P states are clear seen with it
The "problem" should have been addressed, with proper boost voltage and frequency adjustment (bios or even maybe driver adjustments) same OC should be reached its only problem that Nvidia did all the initial tests with the Samsung and ALL the review sites got cards with samsung memory, and not more then a month after release they switched to micron which due the bad R&D had no OC and stability as the Samsung cards had

From my point its a bit misleading to have 99.9% review on internet done with samsung when you in the past few months can buy only cards with micron

Also would like to know why palit and gainward removed their micron bios fix from site?
 
Last edited:

flexy

Diamond Member
Sep 28, 2001
8,464
155
106
This is what ManualG (Nvidia) yesterday posted:

As far as I am aware, we did not complete the full qual back then so the Palit one from September does not have the changes.

And this is likely why it was removed.
 

bystander36

Diamond Member
Apr 1, 2013
5,154
132
106
Until the proper timings are changed for the Micron VRAM, the Samsung has faster VRAM. After they change the timings for Micron, it is unknown which will be better (at least this is what we are told).
 

Dufus

Senior member
Sep 20, 2010
675
119
101
If anyone know how to enforce idle P2 state and not use afterburner curve graph editing, I would like to know, tested Nvidia inspector and it did not do good

What do you mean by idle state? P2 on GTX 1080 for instance is a compute state where clocks are the same as P0 except memory gets down clocked. For GTX 1080 1251MHz to 1128MHZ or something like that but personally I keep both P2 and P0 memory clocks the same speed.
 

reb0rn

Senior member
Dec 31, 2009
320
120
116
What do you mean by idle state? P2 on GTX 1080 for instance is a compute state where clocks are the same as P0 except memory gets down clocked. For GTX 1080 1251MHz to 1128MHZ or something like that but personally I keep both P2 and P0 memory clocks the same speed.

I mean P satate for GPU clock, problem is when card is idle and you kick in/start compute software, micron is crashing when P states change if memory OC is more then +250-400
also while mining there might be gap in work and GPU downlock to some Px state and it again crash
 

Dave2150

Senior member
Jan 20, 2015
639
178
116
It seems the Founders Edition cards really do use premium quality components ;)

All FE cards are Samsung, even the most recent batch. That says alot IMO, I think NVIDIA is doing very well selling their own cards. Very happy with my 1070 FE :)
 

ultima_trev

Member
Nov 4, 2015
148
66
66
I thought it was strange that my friend could only increase the base memory clock on his EVGA GTX 1070 SC by about 75 MHz before artifacting. Wonder if this had something to do with it.
 

bystander36

Diamond Member
Apr 1, 2013
5,154
132
106
I thought it was strange that my friend could only increase the base memory clock on his EVGA GTX 1070 SC by about 75 MHz before artifacting. Wonder if this had something to do with it.
The cool thing about my GTX 1070, is that it's already self clocking about 200Mhz over advertised. Nvidia's new Boost 3.0 will OC beyond the stock clocks already, without your need to do anything yourself.
 

Dufus

Senior member
Sep 20, 2010
675
119
101
I mean P satate for GPU clock, problem is when card is idle and you kick in/start compute software, micron is crashing when P states change if memory OC is more then +250-400
also while mining there might be gap in work and GPU downlock to some Px state and it again crash

Well unfortunately the forcing p-state function is also bugged with Pascal otherwise you could just force P0 or P2.

What voltage did you want to lock to?
 

MrBeeline

Junior Member
Sep 12, 2010
2
0
61
Strange thing is... I have an EVGA 1070 FTW with Micron memory (or so says CPU-z) and the memory has been OC from day one at 8,800MHz and it's just fine (also +100 on the core and hitting 2050/2100 under load). It was stress tested for several hours when I initially bought it and I've since played both Witcher 3 and Deus EX MD for many, many hours since and no sign of a problem.

So it looks like there isn't even an issue with "all" Micron memory on these cards.

I see EVGA has now produced an updated BIOS for my card. But I can't see the point in installing it, as I don't appear to have an issue.
IE. If it aint broke, don't try and fix it.

PS. My card was shipped from EVGA on 31/08/2016 (according to EVGA WEB site when I put my serial number in). So not like it's even one of the latest batch of cards.
 

reb0rn

Senior member
Dec 31, 2009
320
120
116
well 8,800 is not in Mhz its quad rate, so its only +400 in afterburner
MSI afterburner is OC-ing dubble rate so +500 would be 8000 -> 9000
 

MrBeeline

Junior Member
Sep 12, 2010
2
0
61
My bad :( ... Fair enough. To be correct ... In Afterburner I'm setting +400.

A lot of folks with the Micron memory appear to be struggling to OC the memory at all.
So the main thrust of my comment still applies:
"So it looks like there isn't even an issue with "all" Micron memory on these cards."
Which is really what I was trying to get at :)

GPU-z_zpsgf3mpleu.jpg
 
Last edited: