I run a memclock of 180 on 20x 5970s and have tested as low as 120 (efficiency started to decline) without any crashes.
Now IF you are using that GPU to drive a monitor and your display is 3D accelerated (like windows Aero desktop) then low memclock might cause a problem but I haven't seen any on dedicated rigs.
Topic: Underclocking memory causing crash? (Read 1583 times)
no crash in a normal delta of mclock.
On one machine I have win7 x64 and a 6950, no lower than 409 MHz with MSI AB if I want it stable.
I have a 5830 and a 6870 and underclock the mem on both to 290 and it is very stable. I use MSI Afterburner.
When you want to game you can hit reset on Afterburner ans it sets the clocks back to stock. You can also just close Afterburner and that resets the clocks as well.
When you want to game you can hit reset on Afterburner ans it sets the clocks back to stock. You can also just close Afterburner and that resets the clocks as well.
I just tried underclocking my Asus HD5870 (reference model) on Windows, beyond the 600MHz point offered by MSI Afterburner. I had done this on Linux somewhat succesfully. On Windows, it worked at first, but it's practically unusable. If I set the memory clock anywhere between 150-400MHz, and then try to change it back to 1200MHz for gaming, the computer always crashes. At some odd numbers (eg. 240MHz) it seems to crash right away (tested repeatedly). Hashrate also drops if the memory clock is below 300MHz, even though my 5850 cards can go all the way to 150MHz without problems. I'm using SDK 2.1 with the newest drivers (cgminer -n confirms this). My 5870's behaviour wouldn't be a problem for dedicated mining, but since I'm also gaming, I settled on the MSI Afterburner limits for it. Seems they are there for a reason, after all :/
I've had problems with this in the past. If I didnt leave the core clock at default before I made my way to 300mhz on the memory clock ( first stop was at 500mhz ) my computer would lock up and crash. From what I understand, these are "holes" where the core/memory clocks just cant function properly. Batches of cards tend to be slightly different, just got to play around and figure it out.
Hmmm... GPUZ reports the same clocks, could it be showing incorrectly as well?
I tried 650 and its working fine at this time... so you might be onto something.
I tried 650 and its working fine at this time... so you might be onto something.
It can, yes. Usually, the screen will show a bunch of garbage, then crash - if it's set too low. Generally, once you go below 325MHz on modern cards, you're in a territory where the card may crash. It's possible MSI AB was displaying the DDR memory speed, which is probably 2x or more greater than the speed most commonly used to report GPU mem speed. (so, it's possible that Trixx reported it in the common way, which could be 700MHz if displaying mem speed in "DDR" speed, which it's possible MSI AB is doing)
Try 650MHz. Fwiw, I don't believe you can cause any permanent damage by underclocking or overclocking mem speeds -- only problem comes from overvolting & high temps.
Try 650MHz. Fwiw, I don't believe you can cause any permanent damage by underclocking or overclocking mem speeds -- only problem comes from overvolting & high temps.
Obviously overclocking memory will cause a crash, but has anyone had an issue with underclocking?
I have a MSI Twin FROZR III 6950 and I bought a Sapphire 5830 for pure mining in the same rig. I have a dedicated rig with 3 x 5830 running underclocked mem to 350 Mhz just fine, but that is with Trixx.
I am using Afterburner on this one and it only lets me drop to 500 Mhz... but anytime I do, its an almost immediate crash...
I have a MSI Twin FROZR III 6950 and I bought a Sapphire 5830 for pure mining in the same rig. I have a dedicated rig with 3 x 5830 running underclocked mem to 350 Mhz just fine, but that is with Trixx.
I am using Afterburner on this one and it only lets me drop to 500 Mhz... but anytime I do, its an almost immediate crash...
Jump to: