Bitcoin Forum
November 19, 2024, 02:09:24 AM *
News: Check out the artwork 1Dq created to commemorate this forum's 15th anniversary
 
   Home   Help Search Login Register More  
Pages: [1]
  Print  
Author Topic: Harm in Underclocking Mem?  (Read 1582 times)
aceman1011 (OP)
Full Member
***
Offline Offline

Activity: 142
Merit: 100


View Profile
June 09, 2011, 07:08:52 PM
 #1

By underclocking the mem clock on my 5830s, will I be doing any harm? they are at like 800/500. What's the recommended level? This doesn't seem to be bad since the temp only goes lower when I do this.

Rent my 800 + mh/s rig. PM me!
Goldenmaw
Member
**
Offline Offline

Activity: 84
Merit: 10


View Profile
June 09, 2011, 07:11:27 PM
 #2

Good question.  Every card has a lower limit to underclocking - too far and it'll function poorly or not at all.  And every card has a sweet spot, where it runs the coolest and hashes the best.  But is there any danger to this?  I can't imagine one, but I'm no expert.  I'd like to know if anyone can think of a problem here.
Braden
Newbie
*
Offline Offline

Activity: 42
Merit: 0


View Profile
June 09, 2011, 07:12:42 PM
 #3

No. At idle the 5800's underlock the memory to 300 anyway. Anything above that is 100% fine.
bcpokey
Hero Member
*****
Offline Offline

Activity: 602
Merit: 500



View Profile
June 09, 2011, 07:12:57 PM
 #4

There is already a thread about this. But anyway, no, underclocking is a function of powersaving features. The only problem can arise is if you turn the clocks ridiculously low and there is not enough internal bandwidth, which would usually cause a driver malfunction, but no physical harm.
urtur
Full Member
***
Offline Offline

Activity: 211
Merit: 100



View Profile
June 09, 2011, 07:16:24 PM
 #5

My 5970 works well at 175MHz memory (and the performance is the same within the range 175-450, so I use the lowest value to save energy).
Freakin
Full Member
***
Offline Offline

Activity: 154
Merit: 100


View Profile
June 09, 2011, 07:16:56 PM
 #6

most people get the best hash rates on teh 5830s with mem speed about 1/3 of clock speed.  i'm using clock at 990, mem 330 and getting 307MH/s at 61C.
Jack of Diamonds
Sr. Member
****
Offline Offline

Activity: 252
Merit: 251



View Profile
June 09, 2011, 07:17:37 PM
 #7

It's also individual to every card.

Some of my 5850's go down to 300mhz. Some I can't even put to 900mhz or they will freeze when I start mining so I have to run them at full mem speed at the cost of bigger energy cost and heat.

1f3gHNoBodYw1LLs3ndY0UanYB1tC0lnsBec4USeYoU9AREaCH34PBeGgAR67fx
Uzza
Newbie
*
Offline Offline

Activity: 35
Merit: 0


View Profile
June 09, 2011, 07:18:59 PM
 #8

What damages a card is too much heat, which wears the components out. Cards have automatic temperature limits at where the cards underclock to prevent damage.
The other thing that can damage cards is too much voltage. While more voltage leads to more heat, voltage itself can be too much for the components to handle, literally frying them.

So underclocking puts less stress on components, and actually increases their lifespan. And as have been said, the thing that can happen is that the drivers malfunction, and have to be reset.
Goldenmaw
Member
**
Offline Offline

Activity: 84
Merit: 10


View Profile
June 09, 2011, 07:19:34 PM
 #9

Makes sense to me.  Thanks for the input, guys!  I'm running my 5830's memory at 200 mhz.  Seems to be my sweet spot.  Strangely, I've noticed instability at settings like 225 and 250, though.  I wonder why?
LegitBit
Full Member
***
Offline Offline

Activity: 140
Merit: 100



View Profile
June 09, 2011, 07:24:51 PM
 #10

Just because a card can underclock to 300 does not mean it can do all clocks in-between 300 and the standard speed.


Clocks are based on specific ratio timings, screw any of that up and the card wont work well.

Different RAM and other mitigating factors can cause some to have greater or lesser range.

300 just so happens to be a 'normal' speed, so most RAM will run at that... but that doesn't guarantee the core clock will like it.

Also be sure to know the difference between regular and "effective" clocks.

Effective clock takes the combined speed of each individual memory module and totals it. Don't try to down-clock based on that number.

Donate : 1EiAKUmTVtqXsaGLKQQVvLT9DDnHsT7jTZ (Block Explorer)
aceman1011 (OP)
Full Member
***
Offline Offline

Activity: 142
Merit: 100


View Profile
June 09, 2011, 07:27:42 PM
 #11

So what is a good speed for OCing then? Is 875/602 a decent speed? MSI Afterburner is crashing on me, so I have to use the limits set by CCC. The top card is running at 77 C, while the bottom is running at 70 C at those speeds. Is this safe for long term use?
P.S.: I have another fan that will bring them both down to about 67 C.

Rent my 800 + mh/s rig. PM me!
Goldenmaw
Member
**
Offline Offline

Activity: 84
Merit: 10


View Profile
June 09, 2011, 07:28:36 PM
 #12

Just because a card can underclock to 300 does not mean it can do all clocks in-between 300 and the standard speed.


Clocks are based on specific ratio timings, screw any of that up and the card wont work well.

Different RAM and other mitigating factors can cause some to have greater or lesser range.

300 just so happens to be a 'normal' speed, so most RAM will run at that... but that doesn't guarantee the core clock will like it.

Also be sure to know the difference between regular and "effective" clocks.

Effective clock takes the combined speed of each individual memory module and totals it. Don't try to down-clock based on that number.
Good info.  Thanks for explaining that!
LegitBit
Full Member
***
Offline Offline

Activity: 140
Merit: 100



View Profile
June 09, 2011, 07:49:25 PM
 #13

So what is a good speed for OCing then? Is 875/602 a decent speed? MSI Afterburner is crashing on me, so I have to use the limits set by CCC. The top card is running at 77 C, while the bottom is running at 70 C at those speeds. Is this safe for long term use?
P.S.: I have another fan that will bring them both down to about 67 C.

Don't use the temps generated by afterburner, it only shows core temp, not shader temp. Use HWINFO32

http://www.hwinfo.com/download32.html   Run this in sensor mode, no need to use the exe, just download the zip file so you don't need to install it.

Most AMD/ATI cards have another sensor for detecting shader temps, which during mining is up to 10C higher than core.

I personally don't ever like to go above 65C though most cards are rated somewhere between 90-100C but that isn't for continual use. Heat slowly destroys the card, higher the heat, the faster it dies.
If you set the airflow properly you can cool a 5830 to 55C 24/7 on air in a cool room, even when overclocked to CCC's limit on core speeds.

I don't know whether that speed is good for your card, though in my testing 5830's seem stable at "(Core speed) X (0.65) = Memory clock"

Again, at least with my findings.

Good info.  Thanks for explaining that!

Thanks for the courtesy.

Donate : 1EiAKUmTVtqXsaGLKQQVvLT9DDnHsT7jTZ (Block Explorer)
urtur
Full Member
***
Offline Offline

Activity: 211
Merit: 100



View Profile
June 09, 2011, 07:55:14 PM
 #14

Strangely, I've noticed instability at settings like 225 and 250, though.  I wonder why?

I have instabilities for 205<MEM<240. I don't know why either.

My sweet spot is: GPU=900, MEM=175, which gives ~395MH/s at 84C,65%fan.
Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!