Underclocking memory on 5830 causes crashes

[Infinity1]

I have the regular Sapphire 5830 and overclocked the core from 800mHz to 1000mHz while maintaining extremely stable temperatures during Bitcoin mining (67C).

My memory clock is currently at 620 mHz because anything below 600mHz seems to causes a crash. My conputer doesn't crash immediately but when I launch a program such as FurMark or GPU-Z.

The OS this occurs on is Windows XP 32bit and I'm planning to switch to LinuxCoin or Ubuntu, whichever is better on a 4GB USB.

I'm using MSI AfterBurner to overclock.

[1714179200]

1714179200

[1714179200]

1714179200

[pekv2]

I have the regular Sapphire 5830 and overclocked the core from 800mHz to 1000mHz while maintaining extremely stable temperatures during Bitcoin mining (67C).

My memory clock is currently at 620 mHz because anything below 600mHz seems to causes a crash. My conputer doesn't crash immediately but when I launch a program such as FurMark or GPU-Z.

The OS this occurs on is Windows XP 32bit and I'm planning to switch to LinuxCoin or Ubuntu, whichever is better on a 4GB USB.

I'm using MSI AfterBurner to overclock.

Try this. Lower the core back to normal "800" then lower the memory down to 600 and see it it crashes again while what ever you were doing to make it crash, this is purely a test. Then post back with results.

[ssateneth]

Are you inching the memory down? Or are you bringing it down suddenly? Are you doing it while the card is idle or under load? Some cards have very unstable speeds that you should avoid. My asus 5870 absolutely HATES anything from 212MHz and higher to an unknown amount, but its safe at 300MHz. It's the same at around 380+ and 600+ as well. Also some cards or motherboards absolutely HATE changing memory under load (hard freeze). Set your target memory while idle. You may have run into a deadzone. Just keep trying different "regions" of memory frequency. Doesnt like 150? Try 175. Doesnt like it? 200. And keep on going. You may get BSOD's or hard locks. Won't hurt your card, just reset and try again.

[crazyates]

Those 58xx cards hate any memory speeds from the 350-600 range. It's a little uncomfortable at first, but drop those suckers right down to 300, and you shouldn't have any issues.

[Coinoisseur]

It's worth trying 1000/300, my Sapphire 5830s are rock solid with 300Mhz RAM settings. But the cores are at 950, and overclocking can have some seemingly odd quirks. Might need to drop the core clock a bit.

[Infinity1]

Okay, I managed to overclock to 990mHz and downclock the memory all the way down to 300mHz.
After I came back, some ATi processes crashed after 20 minutes of mining halting my mining at 70 shares.

So I'm at 990/300 although I don't really see a problem with those speeds as they seem stable to me unless the card is pulling more than 1.25V

I wonder how people handle over 1000mHz unless they overclock with the Sapphire 5830 Xtreme.

Anyone know how to play with the voltages?

[Coinoisseur]

Going beyond 950ish on a 5830 is really a roll of the dice. There is variation in core chip quality and overall board quality, just as not every CPU overclocks to the same maximum.

[Infinity1]

Going beyond 950ish on a 5830 is really a roll of the dice. There is variation in core chip quality and overall board quality, just as not every CPU overclocks to the same maximum.

Here was the error I got:

[crazyates]

Ya not every chip can go to 1000MHz. Try bumping it down by 10MHz and see if the errors continue to appear. I think I ran my Sapphire 5830 at 940,300.

[ssateneth]

Bad memory clocks shouldnt allow you to do a VPU reset. They will often cause hashing errors (hash returned <1 difficulty) or hard freezes. A VPU/driver reset is usually because your core speed was too aggressive for the voltage you're at. Extremely high core speeds can cause a hard freeze too (1050MHz @ 1v as an example). Either decrease your core speed or increase your core volts (I suggest the former, unless you dont pay for electricity)

[zvs]

sapphire xtremes @ 1000/195 @ 1.163v (at night) or 920/195 @ 1.063v (during the day) were stable for me over a period of about two months on 3 cards.  something to start at when changing voltage, anyway

(as i recall, they were able to do 1040/195, 1035/195, and 1010/195 as max stable)


https://bitcointalk.org/index.php?topic=16406.0

use that for Windows, it's so much easier than afterburner or w/e else =p

[crazyates]

sapphire xtremes @ 1000/195 @ 1.163v (at night) or 920/195 @ 1.063v (during the day) were stable for me over a period of about two months on 3 cards.  something to start at when changing voltage, anyway

(as i recall, they were able to do 1040/195, 1035/195, and 1010/195 as max stable)


https://bitcointalk.org/index.php?topic=16406.0

use that for Windows, it's so much easier than afterburner or w/e else =p

Or CGMiner...

[zvs]

sapphire xtremes @ 1000/195 @ 1.163v (at night) or 920/195 @ 1.063v (during the day) were stable for me over a period of about two months on 3 cards.  something to start at when changing voltage, anyway

(as i recall, they were able to do 1040/195, 1035/195, and 1010/195 as max stable)


https://bitcointalk.org/index.php?topic=16406.0

use that for Windows, it's so much easier than afterburner or w/e else =p

Or CGMiner...

but then you couldn't use phoenix

[crazyates]

sapphire xtremes @ 1000/195 @ 1.163v (at night) or 920/195 @ 1.063v (during the day) were stable for me over a period of about two months on 3 cards.  something to start at when changing voltage, anyway

(as i recall, they were able to do 1040/195, 1035/195, and 1010/195 as max stable)


https://bitcointalk.org/index.php?topic=16406.0

use that for Windows, it's so much easier than afterburner or w/e else =p

Or CGMiner...

but then you couldn't use phoenix

That is true, but whatever kernel Phoenix uses, CGMiner also has.