Abit IC7 / 2.6C @ 3.34 benches

[oldfart]

Benches in the Thugsrook format

Abit IC7 / P4 2.6C Benches

--------------------------------------------------------------------------------

TEST SYSTEM
Abit IC7 - BIOS version 13
P4 2.6C SL6WH Pack 05/02/03
3.34 GHz | 257 FSB
Corsair XMS3200C2 (2x 512) mb
Alpha 8942 - Panaflo M1A
Radeon 9700P - 370/325 core/mem - Cat 3.4 - no AA/AF - default settings
Antec True330 - 330 watt PS
Win XP Pro SP1a
DX 9.0a

TEST SPECS
ABIT IC7 i875P = 1.6 Vcore | 1.5 VAGP | 2.8 Vdimm | PCI/AGP 66/33 | 5:4 ratio | DC DDR 412 | cas 2-2-2-5

I limited the Vcore to 1.6 max. The CPU may go higher with more Vcore, but I didn't try it.

--------------------------------------------------------------------------------

BENCHMARKS

> 3DMark2001 < default
18837

> 3DMark2003 < default
5770


> UT2003 Demo < 1024x768x32, flyby - botmatch
247.8 - 89.8

> Quake III 1024x768x32 < max quality settings
412.6

> Seroius Sam SE Crand Cathedral 1024x768x32 < Quality setting
229.3 KB used for demo profile:
Originally recorded: 11688 frames in 96.9 seconds => 120.7 FPS average.
Without excessive peaks: 11109 frames in 90.1 seconds => 123.3 FPS average.
High peak: 231.4 FPS
Low peak: 81.6 FPS
Low sustained: 92.2 FPS (489 frames in 5.3 seconds)

> AIDA32 ver 3.61 read/write
5263/1803 performance mode "disabled"
5348/2001 performance mode "fast"
There are a bunch of mem tweaks in the IC7 BIOS. Performance mode can be disabled, fast, turbo, ultra. I cant run turbo or ultra. I can "almost" run fast. Fast works fine, but will cause a Prime95 error after and hr or two. My ram is not quite good enough to run it. All benches above were set to disabled.

[jhites]

Very - Very - Nice :cool:

You said all along not to get one of those cheap granite bay boards. :D

[Duvie]

Looking good oldfart!!!!


I think I may be looking at an i865pe mobo by albatron...just the pro that is.....There isn't that much of a difference between the i865 and i875, right???


I am really into my dvd authoring and video editing...I need an extra 512mb of ram so I figured I should bit the bullet and get one of these DCDDR mobos...I will get a 2.6 or 2.8c later when the price drops a bit...I think HT can fit nicely in my system...As well as a SATA raptor drive....

Nothing like making your system seem a bit slow then editing and working on a full 4.38gb DVD full of clips, animated menus and submenus,etc....

[ThugsRook]

very nice http://www.bleedinedge.com/forum/images/icons/icon14.gif

that 5:4 ratio has got to go tho!
are you gonna be trying to chase it down to 1:1 with better ram?

i can see the new game is afoot :deviltail

what we really need is unlocked cpus so we can stop wasting all this $! :na:

:wave:

[oldfart]

Well, to get rid of the 5:4 would require ram that could do DDR 514. It doesn't seem likely right now. For now, a higher multiplier CPU is the only way to run 1:1. I wasn't into spending the $$ on one or some super ram. Not a very good price/performance IMHO.

I'm going to drop the overclock down and do a couple of quick 1:1 Vs 5:4 benches to see the diff.

[ThugsRook]

id suggest booting upto M86 and see what it has to say about your bandwidth along with Sandra and Aida.

:wave:

[Thor]

Originally posted by oldfart
Benches in the Thugsrook format

Abit IC7 / P4 2.6C Benches

--------------------------------------------------------------------------------
> AIDA32 ver 3.61 read/write
5263/1803 performance mode "disabled"
5348/2001 performance mode "fast"
There are a bunch of mem tweaks in the IC7 BIOS. Performance mode can be disabled, fast, turbo, ultra. I cant run turbo or ultra. I can "almost" run fast. Fast works fine, but will cause a Prime95 error after and hr or two. My ram is not quite good enough to run it. All benches above were set to disabled.

Those are still pretty damn impressive for 5:4 memory ratios. Yeah, some 1:1 comparisons would be interesting. Thanks for the benchies!

[oldfart]

I just ran a few quick benches. My Corsair XMS 3200C2 is good for ~ DDR426 2.5,3,3,7. I used this number as the FSB limit.

Test setup:

2.6C @ 213 FSB | 2.77 GHz
1:1 ratio = DDR 426 2.5,3,3,7
5:4 ratio = DDR 340 2,2,2,5
Stock 325/610 9700P clock speed

3DMark03 1:1 | 5:4
5149 | 5139

3DMark01 1:1 | 5:4
16211 | 16122

Quake 3 1:1 | 5:4
346.4 | 345.3

As you can see, VERY little difference. The ability to run a more aggressive timings as opposed to maxing out my ram Freq with slow timings almost makes it even. The same would hold true with the higher overclock. IF I could get ran that could do DDR 514, I'm sure it would have to be at slow timings. I can run what I have @ 5:4 with fast timings.

Spending big $$ on super high end ram would basically get me nothing.

[Thor]

That's nice, but I'd like to see the max fsb at 5:4 vs 1:1 max fsb. There should be a significant difference there, regardless of ram timings?

[oldfart]

My max FSB @ 1:1 is 213. My ram limits me to that. To run 213 FSB, DDR 426, I have to run slow timings. My max FSB @ 5:4 is 257. This is a limit of the CPU @ 1.6 Vcore.

[Thor]

Ok, wasn't clear about the DDR340 at fsb 257. Thanks!

[oldfart]

There was no DDR 340 @ 257 FSB test. The tests that were done:

Original test
257 FSB 5:4 DDR 412

Second test
213 FSB 1:1 DDR 426
213 FSB 5:4 DDR 340

[Thor]

Clear as mud now. Thanks.

[oldfart]

If it is not clear, let me try to explain.

My Corsair ram wont go higher than DDR 426. Doesn't matter what ratio I use. To get to DDR 426, I have to use slow 2.5,3,3,7 timings.

On the first post, I ran 257 FSB, 5:4 ratio, DDR 412. At the lower DDR 412 speeds, I can run more aggressive timings of 2,2,2,5.

The was no 257 FSB 1:1 test. That would require DDR 514, which is impossible for me (or anyone) to run.

You asked for max FSB 1:1 and max FSB 5:4.

Max FSB 1:1 = 213 MHz (ram limitation)
Max FSB 5:4 = 257 MHz (CPU limitation)

Both of these tests were run and are posted.

I also wanted to show a comparison between max 1:1 FSB and same FSB @ 5:4 to show what kind of performance hit 5:4 causes. I had to drop down to 213 FSB to do a 1:1 test since that is the max my ram can handle. I cant run a 1:1 test @ any higher FSB speed.

I think this is really pretty easy to understand. If you want a different test run, LMK what it is and I'll see if I can run it.

[ThugsRook]

ive got some nice juicy Twin XMS3200LL "just sittin here" that will do 440ddr 2227 on 2.7v......

:deviltail

[oldfart]

Originally posted by ThugsRook
ive got some nice juicy Twin XMS3200LL "just sittin here" that will do 440ddr 2227 on 2.7v......

:deviltail
Here is the rub. It wouldn't even do me any good. Right now, I can run DDR 412 2,2,2,5. With your mem, I could run....DDR412 2,2,2,5. No change. Maybe I could get the "Fast" performance mode 100% stable. It still wouldn't get me to DDR514.

[Thor]

Originally posted by oldfart
[B]If it is not clear, let me try to explain.


OF, thanks, I meant that as a figure of speech. I did realize you were maxing your FSB with 5:4 memory ratios in your first benchmarks.

:beer:

[Gigawhat?]

so it looks like the best performance is at 5:4 in your situation (at least in 3Dmark 2001/2003?)

Does it feel faster at 213 sync or 257 asynch?

I have just ordered the IC7 but I am planning to use OCZ EL3500 so hoping to get a little more than 213 synch...

[ThugsRook]

it seems as tho not everyone is getting the same results from these setups

:scratch:

[oldfart]

Originally posted by Gigawhat?
so it looks like the best performance is at 5:4 in your situation (at least in 3Dmark 2001/2003?)

Does it feel faster at 213 sync or 257 asynch?

I have just ordered the IC7 but I am planning to use OCZ EL3500 so hoping to get a little more than 213 synch...

213 FSB (1:1) = 2.77 GHz
257 FSB (5:4) = 3.34 GHz

In my case, the higher CPU speed is by far the way to go. You can see the results. If you have better ram and/or higher multiplier CPU, yours may well be different.

The other thing my testing showed is mem speed Vs mem latency can make a big difference. Running my ram @ max stable frequency which required slow timings was ~ the same performance as running it 5:4 with fast timings.

To get absloute max performance, you need BOTH a high mem freq @ 1:1 AND still be able to run at fast timings. I think you would need to use a 2.8C and some high end ram to be able to do that.

Lets look at the numbers for a 2.8C. Lets say you can get to 3.4 GHz as an example.

14 x 243 FSB = 3.4 GHz
You will need ram that can run DDR486 and still have tight timings. Even 3.3 GHz will still require DDR 472.

People seem to think its so easy for everyone to run 1:1. Not so.

I was willing to take a (very small) performance hit on my setup to save a bunch of $$ by using my existing ram and a 2.6C. Yeah, I could have had a little bit of a faster setup, but it would have cost a LOT more $$ to do it. Not worth it for a small gain.

[oldfart]

Originally posted by ThugsRook
it seems as tho not everyone is getting the same results from these setups

:scratch:
True. Every system is different. I just read BT's OCZ PC3700 Gold review. Lets compare 3DMark2K1 scores

BT's setup:
2.6C P4P800 250FSB 1:1 DDR500 3.25 GHz.
High zoot $$ OCZ PC3700 ram, 2.5,6,3,3 (I think). I'm not clear if he was running 2 or 2.5 with that bench.
9700P @ 410 core and 350 memory
Score: 19379

My setup:
2.6C IC7 257FSB 5:4 DDR412 3.34 GHz.
Corsair PC3200, 2,5,2,2
9700P @ 370 core and 325 memory
Score: 18837

The difference? My score is 2.7% lower.

I am at a large disadvantage with the slower core/mem speed of my 9700P. If I could run 410/350, I'm sure I could easily match or exceed that score. I do have a 9 MHz CPU speed advantage.

So why am I supposed to worry about not running 1:1 ratio? It doesn't make much of a difference. The little it does is mostly made up with the ability do run tighter ram timings.

[ThugsRook]

oh yea ~ you say that now :yummy: (mr. i845 3:4 ratio)
ill give you a month tops.....

congrats on an awesome system :wave:

[oldfart]

LOL....I was thinking the same thing. I was THE guy pushing the 3:4 on the 845, now the opposite. :na:

I do always back up my recommendations with numbers though.

Times change. Right now, I think 5:4 with tight timings performs about the same as 1:1 with slow timings. You can save a pile of $$ by going the latter route.

[ThugsRook]

times change

you dont have to tell me! ;)
ive been 4:5 to 1:1 to 3:4 back to 4:5 and now happily on 1:1 again :wack:

ive been checking out the ddrII specs ~ looks like 266 and 333 to start, 400 to come later. intel must be thinking the same thing about current ram quality.

when the prices come down on the higher mplier chips the game will shift a little. for now ~ you gotta pay to play.

we really need a mplier hack ~ think of all the $ we'd save :smash:

:beer:

[oldfart]

Yeah a 3.0 would be nice with some high end ram. Part of the overclocking game is to get the best bang for the least $$. The high multi CPU and the high $$ ram are not it right now.

Unlocked P4s....dream on! :yummy:

[ThugsRook]

true true true ~ but then again, if you woulda got ahold of a 3.4ghz 533fsb chip you wouldnt have done the 800fsb upgrade at all ;)

thats why im in no hurry to get off my current chip.
id go for a better moboard ~ but so far it doesnt exist for 179fsb! :wack:

hopefully the p4p800 im getting will impress me.
if not maybe ill send it your way for you to play with :wave:

[oldfart]

Your 2.53 @ 3.4 on a G-Bay is an awesome setup, no doubt. I wanted to give the Hyperthreading thing a try just for the heck of it.

You gotta remember my luck with CPU's! I would have got the 2.53 that was a cr@ppy overclocker. I didn't see anyone getting a dud 2.4 or 2.6C so I was pretty safe. As it is, I got an "average" 2.6C, not one if the ones that go sky high on default Vcore.

I just got rid of a 2.4B SL6EF C1 that was not stable @ 140 FSB!!

LMK how you like the P4P800. From what I've read so far, it looks like great one.

[ThugsRook]

pulled this thread into the edge room

:wave:

[Thor]

Originally posted by ThugsRook
ive got some nice juicy Twin XMS3200LL "just sittin here" that will do 440ddr 2227 on 2.7v......

:deviltail

That ram is just asking to be run at those speeds. I'm currently running 426 at 2-3-2-6 @ 2.75 Vdimm. :-) :D :-)

[Gigawhat?]

Oldfart,

I have had my IC7 for about 5 days now and I would like to thank you for your benchmarks which have helped me invaluably to know a baseline for what my system should do. I also have the 2.6c and IC7 and my cpu seems to top out at 257 just as yours did!

I am running OCZ EL3500 that would run synchronous at 228 but I feel I am getting more performance overall (according to 3DMark2001 SE) with the 5:4 ratio at 257. The OCZ would not run tight timings at all (just as Prometheus said) but liked 2-6-3-3 at any fsb. Although the Sandra scores suffered a little with 5:4, I was never able to run tight timings (2-2-2-5) synchronous so my Sandra unbuffered scores never approached what others were getting anyway.

I am very happy with my system and now realize how much work people like you Prometheus and Bigtoe really put in to getting all these comparison numbers that us hardware junkies like to read.

:beer:

[oldfart]

Glad to see this thread helped you get yours running. Its odd we both wound up @ 257 FSB......a little TOO odd. :scratch: I'm gonna have to do some more testing to see if some setting is limiting it. You are at 1.6 Vcore? Mine will error in Prime95 after an hr or so @ 258 FSB.

The memory thing is interesting. It shows that different memory behaves differently. My 3200C2 is not the greatest stuff. It :beat: @ DDR426 2.5-3-3-7, which is pretty low these days. BUT, it is perfect for my rig since it can do DDR412 2-2-2-5! I wont be replacing it anytime soon.
:rock: :rotflmao:

[Gigawhat?]

i am at 1.575 Vcore in the bios but winbond says it is 1.54. I am getting a new power supply this week that may make a difference. 3.3V is also showing way low.

What temps are you getting under load? I got up around 63C when playing a game. Is that getting it too hot? Do you know what the recommended max temp is?

This is also with the 1.3 bios. The beta 25 bios made my ide1 disappear. I am booting from SATA with serillel2 adapter.

Thanks again for your help.


:-r

[oldfart]

The undervolt is normal. New PS wont change that. The temp reading is ~ 15C too high (known "bug"). Your temps look fine.

[Skully]

Heres a link to my IC7 scores and settings! http://www.bleedinedge.com/forum/showthread.php3?s=&threadid=407&perpage=15&pagenumber=2

[SupermanCK]

Originally posted by oldfart
The undervolt is normal. New PS wont change that. The temp reading is ~ 15C too high (known "bug"). Your temps look fine.

either it's a known fact or a known bug...