Drop in 3dMark Score after new processor

[Lost_in_VA]

I recently upgraded my AMD Athlon XP 2500+ Barton (ABit NF7-S) to a Athlon 64 3000+ (ASRock 939Dual Sata II). I am using the same memory (1GB, 2x256 + 1x512) and same video card (EVGA 6600GT), but I added a SB Audigy SE PCI card (used on-board sound for old build).

My old 3dMark score was 3415 (higher if I overclocked) and my score on the new rig is 2851. I know my memory situation is a little odd (running in single channel), but I can't imagine that is causing such a large difference in scores. I am using the same memory in the new build that I used on the old build.

Any insights would be appreciated.

Athlon 64 3000+ Venice
ASRock 939DualSataII, BIOS 1.9, APG FIFO enabled
EVGA 6600GT
1GB memory (2x256 OCZ Premier 2.5-3-3-7, 1x512 CORSAIR ValueSelect)
SB Audigy SE PCI

 

[Jebazor]

That's weird, did you by chance use a different version of 3dmark?
That is about the only insight I would have on this topic.

 

[Lost_in_VA]

Used 3dMark05 free version on both.

 

[Jebazor]

I don't know the exact problem, but is your cpu underclocked for some reason. The RAM may or may not be your problem, but I don't trust synthetic benchmarks in the first place.

 

[sailer]

I can think of two things. First, the addition of the sound card has added overhead to the system, therefore slowing everything down. Second, you ram sticks were better than required for the 2500 chip, but worse then required for the 3000 chip, therefore putting a bind in the system. The new motherboard also may slow things a bit.

 

[MasterLee]

Your 3000 is only 1.8 GHz, even stock 3000 in 754 pin is a tad faster at 2.0 GHz.
Love that 2500 Barton, overclocking monster it is, try matching the speed you had on the 2500 to your 3000 and compare. By the way, what are your settings ?

 

[Lost_in_VA]

Sailor, I am not sure when you say that the memory is worse than required on the new board. All the sticks are DDR 400 (PC3200) which is as fast as the mothrboard supports. I also thought that the soundcard was going to move cycles off the CPU as compared to the on-board sound.

Everything on the new boad is running at factory speeds, video card included.

 

[Datman]

Try removing the 512mb stick and run the 2x256mb in dual channel mode.
See if this makes any difference.

 

[rodney_ws]

I can think of two things. First, the addition of the sound card has added overhead to the system, therefore slowing everything down. Second, you ram sticks were better than required for the 2500 chip, but worse then required for the 3000 chip, therefore putting a bind in the system. The new motherboard also may slow things a bit.

Last time I ran 3dMark, it didn't have sound... barring some unusual resouce conflict I don't see how it's possible for a sound card to reduce performance... especially when stepping up from onboard sound.

 

[jaguarskx]

silly you rodney, you forgot that not only does onboard sound sound better than a sound card, it makes a system run better too

Huh? You're saying an add on sound card like a Creative Audigy 2 or X-Fi sounds worse than onboard sound and eats up more CPU cycles?

:?: :?: :?:

 

[three0duster]

Try removing the 512mb stick and run the 2x256mb in dual channel mode.
See if this makes any difference.

Yea try that, memory bandwidth from dual to single is a big change. for socket 939 single 2500 MB/s to around 5000 MB/s for dual channel, so imagine that along with the possibility of the sound card messing with things, which i dont think that has that great of an effect, but try the memory swapping or pick up another 512 stick ~ 40 bucks for a single value stick.
So try all that and report back

 

[pat]

You did make a fresh Windows install and make sure the AGP driver is loaded, don't you?

 

[Lost_in_VA]

Yes, fresh install of XP, loaded ASRock APG driver and then loaded latest driver from EVGA website.

APG is first device in BIOS, have APG P2P Deep FIFO enabled, Apeture memory set to 128mb (will back this down to 64mb tonight), fast write enabled.

BIOS 1.9

 

[Heyyou27]

I was under the impression that the AGP+ PCIe boards usually perform very poorly on the AGP side thanks to it using the bandwidth of a PCI card.

 

[pat]

Yes, fresh install of XP, loaded ASRock APG driver and then loaded latest driver from EVGA website.

APG is first device in BIOS, have APG P2P Deep FIFO enabled, Apeture memory set to 128mb (will back this down to 64mb tonight), fast write enabled.

BIOS 1.9

There is an option for nvidia card, in BIOS to improve performance, something like AGP P2P Deep Fifo .... anyway, search for it in the BIOS and try this option.

 

[sailer]

I was just speculating at possibilities. At the same time, the individual timings of the ram sticks can cause a conflict. with a default going to the slowest piece. In my hypothetical example, the older board might only have supported the slower timings, so there appeared no confict, but the new board might allow faster timings and then a conflict appears. This is hypothetical, of course. Also, since two that the sticks were 256 mg, while one is 512 mg, that conflict may appear with even greater significance than if both were 512 mg. The different sizes of the memory mean that they aren't balanced with each other. It might be best to just run the system on the 512 stick until a matching stick can be installed. It may make no difference, but then again, it could make a large difference. Only running the 3dmark in both configurations will tell for sure.

As to the effects of the sound card on 3dmark, since 3dmark doesn't have sound, there isn't a direct cause and effect. At the same time though, the card does require additional resources compared to relying on the on- board sound. Its an easy enough thig to test; run 3dmark with the sound card installed, then remove the card and run it again. If there is no difference, then the sound card is not cauasing a problem, but if there is, then that is known too.

 

[pat]

I was under the impression that the AGP+ PCIe boards usually perform very poorly on the AGP side thanks to it using the bandwidth of a PCI card.

This Asrock board is not one of them. it has a real AGP port thanks to an AGP southbridge. You can have an AGP and a PCIe video card running at the same time, both at full speed. I know, I have one and tried it.

The board that you are talking about are those who don't use an extra chip and rather put AGP compatibility to a PCI slot. those perform bad.

 

[uber_g]

shakes head in dismay

freakin ROFL !


heh

 

[Heyyou27]

Ah, maybe I'll pick up one of these boards for my AGP 6800GT.

 

[Lost_in_VA]

Try removing the 512mb stick and run the 2x256mb in dual channel mode.
See if this makes any difference.

Yea try that, memory bandwidth from dual to single is a big change. for socket 939 single 2500 MB/s to around 5000 MB/s for dual channel, so imagine that along with the possibility of the sound card messing with things, which i dont think that has that great of an effect, but try the memory swapping or pick up another 512 stick ~ 40 bucks for a single value stick.
So try all that and report back

I pulled the 512mb module. On reboot showed Dual Channel on the remaining 2x256, but still got the same 3dMark score of 2800, 900 less than I was getting with my 2500+ Barton.

Could I have a bad motherboard? All the other components were simply reinstalled from my other motherboard. Was the Nforce2 really that much better than the ULI that I have now?

Thanks for all the help and input.

 

[pat]

Try removing the 512mb stick and run the 2x256mb in dual channel mode.
See if this makes any difference.

Yea try that, memory bandwidth from dual to single is a big change. for socket 939 single 2500 MB/s to around 5000 MB/s for dual channel, so imagine that along with the possibility of the sound card messing with things, which i dont think that has that great of an effect, but try the memory swapping or pick up another 512 stick ~ 40 bucks for a single value stick.
So try all that and report back

I pulled the 512mb module. On reboot showed Dual Channel on the remaining 2x256, but still got the same 3dMark score of 2800, 900 less than I was getting with my 2500+ Barton.

Could I have a bad motherboard? All the other components were simply reinstalled from my other motherboard. Was the Nforce2 really that much better than the ULI that I have now?

Thanks for all the help and input.

Didi you set the Deep FIFO option in BIOS , as I told in one of my last post?

 

[Lost_in_VA]

Yes AGP P2P Deep FIFO is enabled in BIOS.

 

[azrealhk]

Just a stupid question. Is the resolution the same?

 

[Lost_in_VA]

Found the issue. It appears to be specific to the 6600 GT. That card has a 2d clock rate and a 3d clock rate. For whatever reason, my card was always running at the 300Mhz 2d clock rate instead of the 500Mhz 3d rate.

Got RivaTuner and reset the 3d clock rate to 500 Mhz and 3dMark05 scores went above 3000 again. Don't know how I would have fixed this without RivaTuner or CoolBits.

Don't know if the PCI version would have this issue.

Thanks for everyone's input.