AMD@ 2Ghz vs P4@3.55Ghz: 1-0?
 

I've got my new rig up and running: a P4 2.6C on a 865G motherboard (Asus P4P800-VM). If you disable the onboard grafics (which i did), it performs exactly the same as a 865PE motherboard according to several websites.


The cpu is an excellent overclocker: after playing games for a few hours, 3550mhz (273fsb) proved to be fully stable at stock voltage with aircooling.

I've reinstalled windows properly on a formatted disk, installed dx9, intel chipset drivers & the latest catalyst drivers.

Everyting seems to run ok, but 3Dmark 2001 score is ~15500 and 3dmark 2003 is 6000 (r9800 @ 425/375)... which is about exactly the same as my XP1800+@2000mhz on a budget asus nf2-400 motherboard with silent cooling.

The performance difference between the 3:4 / 4:5 memory divider is neglegible.

-edit-
P4 is single channel (1*256mb bh-5) because it fails to boot with my repaired bh-5 stick, in contrary to the nf2 rig.
-edit-

Does Intel suck (CPU costs 5 times as much as my AMD), or is something horribly wrong?

our P4 2.53, P4T533 and 256MB RIMM PC1066 performs slightly better then my KR7A, 1700+ at 2Ghz and a stick of 256MB mem at FSB 190

dazzling, I'll never buy Intel again. Intel doesn't suck, they're the fastest... but you do pay the price of it.

Okay, his gpu speed is 70mhz faster, but that DOES NOT make 6000 3Dmarks difference. Something must be wrong...

perhaps the high FSB is slowing you down?

try lowering the speed a bit, you might get lagspikes in 3dmark if you're clocking too high.

Nope, the score increases linear (14500 at 3000mhz). No glitches whatsoever.

What does dragon test stress? It's way too low, but the other scores aren't too good either.

so it's not dual channel ?

there's your problem.

dual channel does boost an intel quite substantially...
@ 3.5 (as I normally run 24/7) the difference with my 200*10 amd (office pc) is GIGANTIC in favor of the intel...

Do me a favour, dual channel does not give you 5000 extra 3Dmarks. I'd be suprised if it would give me a +500 boost.

just get you another module, plug it in and see for yourself :p

it DOES make a hughe difference.

If i gain 5000 3dmarks, i'll treat you 30L Stella. :p

The motherboard fails to work with the repaired winbond bh-5 stick, so it will take a while before i can test dual channel.

3.5 ghz :grum: :o

That IS exceptionally, no ? I'm stuck @ 3.25ghz, 250mhz fsb (IS7 - pc3200 corsair, SLK900)

3.55Ghz with air & 1.55v is very good indeed, i'm not complaining about the overclocking capability of the chip. I'm complaining that even at 3.55ghz it isn't faster than my xp1800+@ a lousy 2ghz.

I'll try dual channel before i replace the motherboard, but i don't know which one i should replace it with.

200$ for a p4c800 is way more than i want to spend on a motherboard. 150$ max:

- abit is7: 125$
- abit ic7: 150$
- asus p4p800: 125$
- MSI 865PE: 115$
- MSI 875P: 166$
- Gigabyte 875P: 145$

The gigabyte is the cheapest 875P mobo and has 6 instead of 4 memory slots. I was happy with my gigabyte 845PE motherboard, but asus / abit or even msi are probably better for overclocking.

Does anyone know whether the GIGABYTE GA-8IK1100 has any trouble running at 275FSB? I'm not planning to push my 2.6C much higher, because i will need phase change cooling or +1.55v for that.

Are any of the above motherboard a definite yes/no? I need S-ATA (no raid), 5.1 sound and 10/100 Lan. I don't need CSA GBit lan, integrated s-ata raid, firewire, ...

try other benchmarkes besides 3dmark, dont fix on 1 item
try different ones.. might just be the software

I know for fact that a P4 @ 3ghz IS faster to work with then an AMD @ 2ghz

Same results with other benchmark apps.

-edit-
I don't want to bash intel, i just want to know what's wrong. We've already established it's too low.

fsb: 273mhz
mp: 13 (=> 3550mhz)
mem div: 4/5
mem: 218mhz
mem timings: 2.5-3-3-6
AGP: 8x
Catalyst: 3.5 i think
R9800: 425/375
Dx: 9
Latest intel chipset drivers

is it a clean install, or did you change mobo/cpu and kept the OS install?

clean install on formatted disk. As clean as it can be. :)

the fact is that sth is terribly wrong, cuz I have the exact same comparison with a 2ghz (10*200) xp and my gamerig (3.6), same amount of ram, and the difference is catastrophic for the xp.

or it could be your special program "clockgen" that isn't working verry good :p

cool, didn't know about clockgen

http://www.cpuid.com/clockgen.php

true, synthetic benchmarks dedect default speed, but programs like 3dmark do increase in speed.

"some versions of windows"...? i'll install SP1, see if that solves it.

I don't think that will help

It would be pretty odd, if a program released on 18th august 2003 doesn't work well in Win XP...

It doesn't hurt to update, i'll report back in half an hour.

SP1 didn't do it. Installing other updates too.


I'm sure now the incorrect mhz recognition is the cause of appearantly low performance!

When I boot at 2600mhz, and change the speed to 3500mhz, i'm 100% sure my machine is actually running at 3500mhz, but windows doesn't adapt to the changed speed.

When i launch a game of vice city, it's like the game is on crack: everything goes a bit too fast. Your car feels like it does 300kph, but your speedometer still says 200kph.

An excellent example is an old DOS game without a frame limitter on a new machine: they don't adapt to the high speed.

I hope there's a fix for this error.

my guess is, you put that CPU in another motherboard and it will not perform like this.

like said on shrimp, I can clock much higher with clockgen then when I use the Bios.
I don't believe it.

You don't believe the cpu is running at 3550mhz stable, or you don't believe my system would not crash with a cold boot at 3550 mhz?

What happens if you compare your 4ghz bios oc with a 4.3ghz clockgen oc?


-edit-

If i set clock speed to 3500mhz, and i reset, the post screen says "3.55Ghz". Is that cheating too?

-edit2-

Latest version of cpu-z shows correct speed.

I'm going to compare these two situations:
a) Boot at 2600mhz, oc in windows to 3300 & bench.
b) reboot at 3300mhz & bench

very very interesting what we see here!

I wanna know how this turns out :)

CPU scaling seems to be ok, and there is no difference between a) & b) in performance.

(In situation a, not all apps dedect the correct speed, in contrary to situation b).

difference 2.8 <> 3.0 is almost nihil, but 3.0 <> 3.2 is alot
how come?

You tell me? I'm just showing the test results. All options were left the same.

try sisoft sandra cpu benchmark (it still is freeware) and compare it to other CPU's

other mem settings ?

Not CPU limited, not graka limited. Must be mobo/memory then.
Dual channel should give a huge boost, and if memory is all that's keeping you back, it might actually gain a couple of thousand points. That and the 875's PAT might be something.

Unless somthing's wrong. Try memtest 86, and prime to see if any of them fail.

Prime fails above 3330mhz, but 3d stable up to 3550mhz. Memtest reports no errors.

I will be able to test dual channel tonight, i'm curious about the boost.

@piotke: max vdimm: 2.65v. bh-5 sticks do about 220mhz on relax timings (2.5, 6-3-3) on default voltage and memtest reports no errors.

Dual channel barely makes a difference:

So what do you guys think? Any other option left but to buy a true overclocking board? If so, which one for <150? Abit IC7?

Well if all things ARE about equal, there's somthing that's gotta be keeping you back.

I only know that I want the P4C800-E Deluxe. P4C800 vanilla is known to do good too.

it's a beauty, especially if you have a good cpu like richy's cpu

I'm loving my P4C800 Deluxe :ws:

I think i'm going to keep the P4P800-VM for a while. After playing GTA: Vice City and C&C generals i can say it's defenitly faster than my AMD setup, even if most benchmark apps tell otherwise. It's probably the clockgen app.


I don't feel like spending 220 euro on a motherboard that isn't much faster, but doesn't screw up benchmark apps... I'll wait till next month. ;)

oh my god, the man does have brains ;)