2.4C on P4P800 Deluxe (ongoing drama thread, al a THUGSROOK)

[ErikaeanLogic]

I switched from the P4P800 non-Deluxe to the Deluxe version of the P4P800 for the Firewire feature (I'm an iPod man;)), but initially experienced a bunch of issues in the transition:

- Heat! My system and cpu temps were 8-10C higher. Killed that by applying some of the controversial AS3:).

- Instability! My 2.6C, which was playing ball on the non-Deluxe to the tune of 262 fsb (3410MHz) was no longer stable (ie: Prime-able) at 255 fsb, same bios/settings (v1007). Tried 2.4C I have here and same issue, applied bios v1008 and 2.4C is now happily Prime-ing away for over 5 hours at 279 fsb, more than 12fsb higher than was possible on the 1007 bios!

- Over/Under-volting. Okay, Asus has improved the overvolting situation of the P4P800 (when compared with the last few boards of theirs I've had) but it's still 0.03-0.05V over at idle, which I can live with. What's bizarre, however, is that with some Vcore settings, the voltage drops in a big way, sometimes 0.12-0.13V! I know that the Intel P4 Northwood whitepapers suggest that this is about within spec, but the Abit users don't seem to be experiencing this wild variation, at least not that I've read. 1.6Vcore seems to be the sweet spot: it idles at about 1.648V and drops to 1.55V under the heaviest load (default Vcore).

- Confusion! Despite the common variation of results we witness from member to member with similar setups, I cannot determine why some score higher with one bios than others, given approximately equal settings on same bioses. Weird. All I can say is that v1008 beta is doing right by me (for now). Also, kudos go to Asus for frequently updating the bios for the P4P800! :rock: I felt that they left us rather high and dry with the P4G8X bios but they've been updating this one at the rate of about one bios per week. Not too shabby, eh:)?

To be honest, I was going to dump my Springdale setup until the next revision of these boards comes out, but bios v1008 has changed my mind (again, for now). Have I gotten a speed increase over my P4G8X with this upgrade? Marginally, in games and apps, but the the real benefit for me will be implementing a HT-enabled system in DVD/MP3 rips.

Without further ado, here are my current system specs:

P4 2.4C @3350, 279fsb, HT not yet enabled, 1.6Vcore (1.648/1.55V idle/load, actual)
various, 5:4 *
Radeon 9700 Pro, default (325/310)
WD 80 JB
Pioneer 115
Pioneer 105
SB Live! 5.1 OEM
Antec TruePower 550
H2O cooling
Windows 2003 Server, Enterprise edition

* memory config will change: I have XMS3200C2 and TwinMOS PC3200 CH5 to play with, too

I'm going to bench with each memory timing first, then update the thread when I switch hardware (memory, cpu).

CONFIG#/TIMINGS

BUFFALO PC3700
1: 2.5-4-4-8/Turbo/MAM auto
2: 2.5-3-3-7/Turbo/MAM auto
3: 2.5-3-3-7/Standard/MAM auto
4: 2.5-3-3-7/Standard/MAM on
5: 2.5-3-3-7/Turbo/MAM on
6: 2-3-3-7/Standard/MAM auto

SiSoft Sandra Memory Bandwidth (buffered)

1: 5803/5758
2: 5937/5830
3: 5922/5810
4: 5966/5992
5: 6003/5888
6: 5856/5972

SiSoft Sandra Memory Bandwidth (unbuffered)

1: 2756/2810
2: 3031/3113
3: 2943/2999
4: 3089/3099
5: 2967/3025
6: 3002/3106

SuperPI (seconds)

1: 42
2: 41
3: 41
4: 41
5: 41
6: 41

3DMark2001SE, build 330

1: 16,822
2: 16,988
3: 17,010
4: 17,055
5: 17,093
6: 17,045

Unreal Tournament 2003 Benchmark (1024x768)

1: 211.5/90.9
2: 213.5/92.7
3: 213.4/92.8
4: 213.6/93.6
5: 213.5/93.7
6: 213.7/93.0

Q3 Bench (1280x1024, max settings)

1: 287.5
2: 291.6
3: 291.9
4: 294.4
5: 294.3
6: 293.6

CodeCreatures

1: 38.7
2: 38.8
3: 38.7
4: 38.8
5: 38.8
6: 38.8

[ErikaeanLogic]

Well, the TwinMOS CH5 wouldn't even post at 279fsb/5:4 = 223, not even at SPD timings, not even one stick at a time. I guess I'll let one of those AMD guys buy it;).

The XMS3200C2 works just fine at this speed, but not at any better timings than I've already test with the Buffalo so I'll not bother to bench it.

I've only drawn these conclusions:

1) THUGSROOK definitely found something special, that the "Standard" setting yields equal results (in some cases slightly better) than the "Turbo" setting. MAM on or off (with the 5:4 ratio), this is the case.

2) 2-3-3-7 and 2.5-3-3-7 are just about equal in results, although the 2.5 CAS setting yielded higher real-life (as opposed to synthetic) marks.

3) I've spent entirely too much time and money on this project!:smash: :eek2: :cry: :rock:

Off to see if the v1008 bios helps out my 2.6C. . .:wave:

[ThugsRook]

this is very interesting.

you are running 446ddr right?

[ErikaeanLogic]

Originally posted by ThugsRook
this is very interesting.

you are running 446ddr right?

Yep. Well, I was. . .back on a 2.6C for now;)

[jhites]

Less than 2% gain/loss in real world performance regardless of memory settings. Interesting!

Does it feel or respond any different with that small of a range?

[ErikaeanLogic]

Originally posted by jhites
Less than 2% gain/loss in real world performance regardless of memory settings. Interesting!

Does it feel or respond any different with that small of a range?

I can't really tell, John, all I can say is that there is a difference in system responsiveness when HT is enabled. . .the only reason to upgrade, IMHO**. Or, if cost is no obstacle, a 1:1 setup with a 3.0C and some dynamite memory (al a Bigtoe) would be the best of all possible worlds. And a Vapochill/Prommie. And a Ferrari Enzo. :na:














** (aside from profound mania:deviltail !)

[ThugsRook]

do you plan on doing any HT comparisons?

:deviltail

[ErikaeanLogic]

I may consider doing a HT comparison, THUGS. How is it that you time a DiVX rip? Is there a utility for that, which displays the amount of time it took to convert to DiVX? How about MP3's? These are real-world uses for HT but I don't remember how it is that you and Duvie bench 'em. Suggestions? Thanks:).

[ThugsRook]

i use a program called DVD2Avi for encoding.
it tells me total encoding time when its finished.

i just walk away and come back when its done.

:)

[Mikki]

Excellent info, Erik...;):)

[jhites]

Originally posted by ErikaeanLogic
I can't really tell, John, all I can say is that there is a difference in system responsiveness when HT is enabled. . .the only reason to upgrade, IMHO**. Or, if cost is no obstacle, a 1:1 setup with a 3.0C and some dynamite memory (al a Bigtoe) would be the best of all possible worlds. And a Vapochill/Prommie. And a Ferrari Enzo. :na:
** (aside from profound mania:deviltail !) Actually, looking to build another rig. I see quite a bit of advantage to the HT chips when running Seti but not really much with games or other programs. Don't run DivX so can't really say one way or another.

I think I am going to get a P4P800-VE I865G and a 2.8C and run it 1:1 with TwinMos PC3700 just for kicks and :-) . I don't think for $100 bucks more on a 3.0C there will be much performance difference if I can run at 240-245fsb with a 2.8C. I know that the TwinMos-3700 will do DDR492 2.5-3-3-7 with 2.85vdimm.

[ErikaeanLogic]

Originally posted by jhites
Actually, looking to build another rig. I see quite a bit of advantage to the HT chips when running Seti but not really much with games or other programs. Don't run DivX so can't really say one way or another.

I think I am going to get a P4P800-VE I865G and a 2.8C and run it 1:1 with TwinMos PC3700 just for kicks and :-) . I don't think for $100 bucks more on a 3.0C there will be much performance difference if I can run at 240-245fsb with a 2.8C. I know that the TwinMos-3700 will do DDR492 2.5-3-3-7 with 2.85vdimm.

not to burst your bubble, but I had alot of trouble getting the TwinMOS PC3200 CH5 to work on this board at all but stock speeds/timings (from what I've read, has to do with the SPD programming on the mem); good luck with the PC3700, tho:)

[oldfart]

Originally posted by ErikaeanLogic
I may consider doing a HT comparison, THUGS. How is it that you time a DiVX rip? Is there a utility for that, which displays the amount of time it took to convert to DiVX? How about MP3's? These are real-world uses for HT but I don't remember how it is that you and Duvie bench 'em. Suggestions? Thanks:).
I use LAME for MP3 encoding. Its probably the best MP3 coder out there. I use a GUI front end for it called Razor Lame. It creates a log with encode time. I don't do any DivX stuff, but do quite a bit of MPEG1/2 encoding with TMPGEnc. This program is awesome and displays encoding time. They are both consistent. You can encode the same file several times, and get consistent results.

[Thor]

Originally posted by ErikaeanLogic
[B]I switched from the P4P800 non-Deluxe to the Deluxe version of the P4P800

Yeah, keep reading the deluxe board OC's better. Damn.



Tried 2.4C I have here and same issue, applied bios v1008 and 2.4C is now happily Prime-ing away for over 5 hours at 279 fsb, more than 12fsb higher than was possible on the 1007 bios!


Ahh, the 1008 beta bios didn't help me. Damn again.

Nice work on the benchies and write up there Erik!