Featured Worklog

Price Search



PC Apex Sponsor


PC Apex Sponsors



PC Apex RSS Feeds

RSS Feed for PC Apex Reviews & ArticlesRSS Feed for PC Apex PC Modding WorklogsRSS Feed for the PC Apex Daily DisturbanceRSS Feed for the latest PC Apex Site NewsRSS Feed for PC Apex Affiliate and Web NewsRSS Feed for PC Apex Deals and Steals

Go Back   Apex Community Forums // PC Apex Forums // Benchmarking // Results

Results Share your benchmark results here.

Reply
 
LinkBack Thread Tools Display Modes
Old 31-March-07, 11:05 PM   #1 (permalink)
Apex Elite Tech
Hak Foo's Avatar
Default "Forgiveness" of benchmarks

Today I decided to wind up my 4600+ and see if I could earn back a few HWBoints.

Now this is without "chained" tweaks, leaving the voltage and all timings as before, just boosting the FSB.

First I tried wPrime. That would run, even the long 1024M version, at speeds as high as FSB=215 (which delivered 2596MHz and a POST screen of "2.61GHz").

At FSB=218 and 220, it also completed, but the system would bomb when I tried to restart everything so I could submit results. (for best scores, I strip down everything, including the 802.11g card driver, to bench)

At FSB=225 and 230 with reduced timings (DDR333 memory settings, LDT 4x instead of 5x), it failed to boot. 230 stopping after POST, 225 during the Windows load.

Then I tried 3DMark03. At 215, it ran two tests then bombed to desktop.

SuperPi 1M would either fail the first test, or if I tried a second test, at 210 and higher. I didn't bother trying anything between 200 and 210.

So all I got out of it were some new wPrime scores, which don't even actually move the needle that I can see except to allow me to say I've pushed my CPU over 2.5GHz.

Has anyone else noticed similar phenomena-- that some benchmarks are far more forgiving of overclocking than others?
Hak Foo is offline     Reply With Quote
Old 31-March-07, 11:27 PM   #2 (permalink)
n00b-ass reviewer
BigAkita's Avatar
Default Re: "Forgiveness" of benchmarks

Yes. More text to meet minimum required.
BigAkita is offline     Reply With Quote
Old 01-April-07, 09:26 AM   #3 (permalink)
Apex Tech God
Zaltan's Avatar
Default Re: "Forgiveness" of benchmarks

Aquamark has always been nice to me. I'm able to run higher speeds that other benches. Does your board have voltage options.
Zaltan is offline     Reply With Quote
Old 01-April-07, 12:01 PM   #4 (permalink)
Apex Elite Tech
Hak Foo's Avatar
Default Re: "Forgiveness" of benchmarks

I don't fiddle with voltages if at all possible-- I figure dialing up the FSB has no risk of damaging a nearly irreplacable S939 chip, but dialing up the voltage DOES.
Hak Foo is offline     Reply With Quote
Old 01-April-07, 01:30 PM   #5 (permalink)
A George Orwell fan...sorta
godfoot's Avatar
Default Re: "Forgiveness" of benchmarks

Depends on the particular hardware involved.
For instance, when I had the Firestix for a while (and now the same thing is happening with a current review item), the stock voltage that my mobo put the RAM at was 1.8 or 1.9, but to get DDR2 1000 (which is it's rated speed) a slight bump in voltage is needed.
godfoot is offline     Reply With Quote
Old 01-April-07, 01:50 PM   #6 (permalink)
Apex Elite Tech
Hak Foo's Avatar
Default Re: "Forgiveness" of benchmarks

Well, I looked, and it doesn't matter because my mainboard tops out in terms of CPU voltage at 1.40 (which is stock for the X2 as far as I can tell), and the RAM is good to 220 or so-- I recall being able to get that high on my old 3200+
Hak Foo is offline     Reply With Quote
Old 01-April-07, 01:55 PM   #7 (permalink)
A George Orwell fan...sorta
godfoot's Avatar
Default Re: "Forgiveness" of benchmarks

I was referring to RAM voltage, not CPU.
I think my CPU voltage tops at 1.5 or 1.6.... but never had to take it that high as yet.
godfoot is offline     Reply With Quote
Old 01-April-07, 02:16 PM   #8 (permalink)
Apex Tech God
Zaltan's Avatar
Default Re: "Forgiveness" of benchmarks

Have you changed the HTT multiplier Hak Foo? That could also be the problem. Though, I would really recommend increasing your Vcore to 1.5v at least.
Zaltan is offline     Reply With Quote
Old 01-April-07, 03:17 PM   #9 (permalink)
Apex Elite Tech
Hak Foo's Avatar
Default Re: "Forgiveness" of benchmarks

Well, the only options are 1.4 and below. I tried knocking the HTT multiplier down to 4x (well, they call it '800MHz', but AFAICT, it's the same thing)
Hak Foo is offline     Reply With Quote
Old 01-April-07, 03:31 PM   #10 (permalink)
Apex Tech God
Zaltan's Avatar
Default Re: "Forgiveness" of benchmarks

Quote:
Originally Posted by Hak Foo
Well, the only options are 1.4 and below. I tried knocking the HTT multiplier down to 4x (well, they call it '800MHz', but AFAICT, it's the same thing)

800MHz will help overclocks. What board do you have?
Zaltan is offline     Reply With Quote
Old 01-April-07, 03:39 PM   #11 (permalink)
Apex Elite Tech
Hak Foo's Avatar
Default Re: "Forgiveness" of benchmarks

ASRock 939SLI32-eSATA2.

ULI M1695 chipset (yes, it theoretically will do dual-X16 SLI, if you use very specific hacked driver versions, but what I liked about it was that it was fanless unlike almost every nF4 board)

Edit:

Knocking the HTT to 4x and the memory timings down to DDR333 settings seems to allow SuperPi 1M at FSB=220. Of course, 177MHz memory clock is not hugely great for overall performance.

Last edited by Hak Foo; 01-April-07 at 03:51 PM..
Hak Foo is offline     Reply With Quote
Old 02-April-07, 11:34 PM   #12 (permalink)
Apex Elite Tech
Hak Foo's Avatar
Default Re: "Forgiveness" of benchmarks

Quote:
Originally Posted by Hak Foo
ASRock 939SLI32-eSATA2.

ULI M1695 chipset (yes, it theoretically will do dual-X16 SLI, if you use very specific hacked driver versions, but what I liked about it was that it was fanless unlike almost every nF4 board)

Edit:

Knocking the HTT to 4x and the memory timings down to DDR333 settings seems to allow SuperPi 1M at FSB=220. Of course, 177MHz memory clock is not hugely great for overall performance.



Damn it. It's still not stable enough for long runs. I thought it was heat (dialed the rear fan to full blast, no benefit) but I fear it's voltage which I cannot fix.

One of the more spectacular failures:

wPrime 1024M. Core 1 stops trickling results at about 64%. Core 2 keeps ticking to 65%-- like 3 updates of the screen unmatched. Enough time to go o.O. Eventually the system finally gives out with a resounding front-panel-speaker beep.
Hak Foo is offline     Reply With Quote
Old 02-April-07, 11:39 PM   #13 (permalink)
A George Orwell fan...sorta
godfoot's Avatar
Default Re: "Forgiveness" of benchmarks

Methinks you need a better mobo, but only your wallet can decide that.
godfoot is offline     Reply With Quote
Reply

Bookmarks

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is On
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Slashdot // "Market Share" "Installed Base" and Consumer Electronics Gizmo Slashdot RSS 0 18-March-07 01:30 PM
Ars Technica // Conservapedia hopes to "fix" Wikipedia's "liberal bias" Gizmo Ars Technica RSS 0 05-March-07 01:11 AM
Ars Technica // Blake Ross has a "tip" for Google: stop using "tips" Gizmo Ars Technica RSS 0 29-December-06 06:30 PM
Ars Technica // Cable and phone companies call Net neutrality "silly," "mumbo jumbo" Gizmo Ars Technica RSS 0 07-September-06 11:01 PM
Ars Technica // Women's "don't date him" gossip site sued by "victim" Gizmo Ars Technica RSS 0 30-June-06 03:01 PM


All times are GMT -5. The time now is 10:07 PM.


Powered by vBulletin® Version 3.8.3
Copyright ©2000 - 2009, Jelsoft Enterprises Ltd.
Search Engine Optimization by vBSEO 3.3.0
Copyright PCApex.com, GameApex.com, ForumApex.com 2001 - 2008
Advertisements

Page generated in 0.22071 seconds with 9 queries