Why is my P3 966 not acting like one?

[Viper GTS]

OK, I think I've FINALLY nailed down why my computer scores so low in 3DMark 2000 & Q3A timedemos. Found a guy over in the OC forum that reduced the speed of his system slightly to match mine, & then ran 3DMark 2000.

Both of us were running the following:

P3 700e @ 966 (138x7.0)
128 MB Mosel Vitalic Revision 2 @ 138 CAS2
GeForce 2 GTS @ 220/400 (His is a 64 MB, mine's a 32)
I have a BX based BF6, he has an i815 based CUSL2

My scores:

3DMark Result 6711 3D marks

CPU Speed 470 CPU 3D marks
Game 1 - Helicopter - Low Detail 130.05 FPS
Game 1 - Helicopter - Medium Detail 96.84 FPS
Game 1 - Helicopter - High Detail 58.21 FPS
Game 2 - Adventure - Low Detail 161.55 FPS
Game 2 - Adventure - Medium Detail 70.82 FPS
Game 2 - Adventure - High Detail 41.75 FPS
Fill Rate (Single-Texturing) 685.95 MTexels/s
Fill Rate (Multi-Texturing) 1404.27 MTexels/s
High Polygon Count (1 Light) 9178 KTriangles/s
High Polygon Count (4 Lights) 7971 KTriangles/s
High Polygon Count (8 Lights) 6069 KTriangles/s
8MB Texture Rendering Speed 505.15 FPS
16MB Texture Rendering Speed 352.29 FPS
32MB Texture Rendering Speed 167.45 FPS
64MB Texture Rendering Speed 80.77 FPS
Bump Mapping (Emboss, 3-pass) 233.85 FPS
Bump Mapping (Emboss, 2-pass) 338.44 FPS
Bump Mapping (Emboss, 1-pass) 558.35 FPS
Bump Mapping (Environment) Not Supported

His scores:

3DMark Result 7555 3D marks

CPU Speed 549 CPU 3D marks
Game 1 - Helicopter - Low Detail 146.15 FPS
Game 1 - Helicopter - Medium Detail 109.70 FPS
Game 1 - Helicopter - High Detail 59.07 FPS
Game 2 - Adventure - Low Detail 179.02 FPS
Game 2 - Adventure - Medium Detail 85.16 FPS
Game 2 - Adventure - High Detail 50.45 FPS
Fill Rate (Single-Texturing) 688.85 MTexels/s
Fill Rate (Multi-Texturing) 1400.42 MTexels/s
High Polygon Count (1 Light) 14968 KTriangles/s
High Polygon Count (4 Lights) 7945 KTriangles/s
High Polygon Count (8 Lights) 4230 KTriangles/s
8MB Texture Rendering Speed 511.62 FPS
16MB Texture Rendering Speed 384.98 FPS
32MB Texture Rendering Speed 208.23 FPS
64MB Texture Rendering Speed 104.60 FPS
Bump Mapping (Emboss, 3-pass) 250.90 FPS
Bump Mapping (Emboss, 2-pass) 333.47 FPS
Bump Mapping (Emboss, 1-pass) 551.20 FPS
Bump Mapping (Environment) Not Supported

The highlighted areas are the interesting points.

CPU Speed - This is where the main discrepancy is.
Fill Rate - These values are practically identical, eliminating my video card/drivers as the source of the problem.

What could be causing the huge variance in CPU speed? As far as I know, my BIOS is configured optimally.

Any ideas? This has been frustrating me for some time, & I may FINALLY be onto the source of the problem.

Viper GTS

 

[Syborg1211]

no one cpu is alike and it seems like you got a bad one. Some cpus take well to testing and benching and overclocking and some seem to have a lower bandwidth so to say. I have a coppermine 700e and i downclocked to 700 and benchmakred just my cpu against my sister's coppermine 700 comcrap machine and my cpu TOTALLY annihilated my sister's cpu. Also, i used the same program for cpu speed tester and it revealed mine as a 701 or 2 or something and it rated my sister's as a 694.

 

[Viper GTS]

But a 14% difference? That's ridiculous. RC5 performance is right were it should be, too.

Viper GTS

 

[utopia]

<< no one cpu is alike and it seems like you got a bad one. Some cpus take well to testing and benching and overclocking and some seem to have a lower bandwidth so to say. I have a coppermine 700e and i downclocked to 700 and benchmakred just my cpu against my sister's coppermine 700 comcrap machine and my cpu TOTALLY annihilated my sister's cpu. >>



Nothing personal, but READ your own writing! Do you have ANY idea on what you are saying? A CPU will operate at whatever frequency (Hz/MHz/GHz) you set it to (if it is capable). The score difference is because of system configurations; even small stuff like different BUS (IDE/SCSI/PCI) drivers will affect the score. I never trust 3dMark; i somehow get 4000 on a 1.1Ghz (PIII) GF2-64 256MB system... If you want it real, do Q3 and UT timedemos.

 

[Viper GTS]

I'll go do some Q3 timedemos, BRB...

They've always been EXTREMELY low, though.

Viper GTS

 

[Buddha Bart]

syborg: nothing personal man, but you're talking out your ass.

Viper: it has to be somthing else, try to download some other benchmarks. Especially since rc5 is where it should be, then you know its running fine.

bart

 

[AdamK47]

Did you set the &quot;In Order Queue Depth&quot; to 8 in the BIOS of the BF6?

 

[utopia]

Viper, Q3 will show you the right scores... try 800x600x32. Dont go over that resolution because your card is a 32MB and his is a 64MB...

 

[Viper GTS]

OK, here's some timedemo scores:

Each test was run three times, &amp; results are in the order they occured.

Q3 Arena, Demo001

Fastest: 105.4, 105.5, 105.5
Fast: 99.9, 99.7, 99.9
Normal: 96.5, 99.4, 99.2
Highest: 98.1, 98.0, 98.0

And just for the heck of it:

1280x1024 with EVERYTHING maxed (max detail, 32-bit color, lightmap lighting, etc): 59.3, 59.9, 59.9

I think it's got to be CPU, since the fastest scores are like 30% lower than comparable systems, &amp; they're CPU limited, not video card.

I'll go check in-order queue depth.

BRB, again.



Viper GTS

 

[Viper GTS]

Did you set the &quot;In Order Queue Depth&quot; to 8 in the BIOS of the BF6?

AdamK47 - 3DS, you are my hero for today. It was set to 1. Rebooted with it set to eight, things are looking A HELL OF A FVCKING LOT BRIGHTER!

After scores:

Fastest: 141.9, 141.7, 143.9
Normal: 133.3, 133.2, 133.1

I love you, I love you, I love you!!!

Now THAT'S what my system should perform like!



Viper GTS

 

[Slacker]

Hey Adam, is there a setting like that for the i815e pro? I have similar results as viper and would like to make sure my bios is setup right.

 

[BuckMaster]

Doesnt it just feel GOOD all over when someone helps out with a solution!! I know i do!!!

 

[Viper GTS]

I've been fighting this for several months, &amp; I guess I just overlooked that setting.

It's an incredible relief to have the answer.

Viper GTS

 

[TuffGuy]

Slacker - no such setting in the cusl2 that i've found.

 

[Viper GTS]

For anybody who's curious, here's my 3DMark score after the fix:

3DMark Result 7588 3D marks
CPU Speed 551 CPU 3D marks
Game 1 - Helicopter - Low Detail 151.72 FPS
Game 1 - Helicopter - Medium Detail 111.47 FPS
Game 1 - Helicopter - High Detail 59.61 FPS
Game 2 - Adventure - Low Detail 173.51 FPS
Game 2 - Adventure - Medium Detail 85.52 FPS
Game 2 - Adventure - High Detail 50.52 FPS
Fill Rate (Single-Texturing) 686.31 MTexels/s
Fill Rate (Multi-Texturing) 1405.70 MTexels/s
High Polygon Count (1 Light) 11370 KTriangles/s
High Polygon Count (4 Lights) 9963 KTriangles/s
High Polygon Count (8 Lights) 6079 KTriangles/s
8MB Texture Rendering Speed 510.14 FPS
16MB Texture Rendering Speed 363.91 FPS
32MB Texture Rendering Speed 198.39 FPS
64MB Texture Rendering Speed 94.61 FPS
Bump Mapping (Emboss, 3-pass) 256.85 FPS
Bump Mapping (Emboss, 2-pass) 341.80 FPS
Bump Mapping (Emboss, 1-pass) 560.94 FPS
Bump Mapping (Environment) Not Supported

That's a little more like it.



Viper GTS

 

[Lore]

What in the world is the &quot;In Order Queue Depth&quot; anyhow?

 

[Viper GTS]

According to ocworkbench.com:

In Order Queue Depth: This determines the command queue depth between the CPU and the chipset and the default is at 1 but for most people, setting it to 8 will give increased performance.

Viper GTS

 

[masterskartrg]

that setting right there makes a huge difference in overclocking its also to test stabilty i remeber on a be6-2 if i set it to 1 i can overclock way more its just like memory cas settings

 

[The Dancing Peacock]

Does this only apply to overclocked CPU's? Could I do it on my be-6 PIII 450, and expect an increase in performance with no other mods?


thanks.


late.


TDP

 

[Viper GTS]

TDP...

No, it applies to stock clocked CPUs too. One is the default option in the BIOS, but 8 obviously gives a rather large performance boost.

Viper GTS

 

[Goi]

Hmmn...does the BH6 have that option too? Can't seem to find it.

 

[GT578]

I know this may be off topic but, do you guys think the In Order Queue Depth can cause a windows installation to go bad?? I have a Cel566@901mhz on a BF6 and I put it back to 566(in user define still) with the In Order Queue Depth set to 8 and installed windows. IE sometimes stalls on me...the program doesn't respond when I click on a link sometimes...i have to click several times or wait a while until it responds...and I'm pretty sure its software-related. If I put the In Order Queue Depth to 1 at 566mhz and reinstall...do you think it'll make a difference in the stability of all the programs installed?

 

[AdamK47]

Ever since I wrote my BF6 and BE6-2 reviews, I got a lot of questions with people asking why they were getting sub-par performance. Nine times out of ten, they had the Queue depth set to 1 instead of 8.

Now that you have your 3DMark scores up there, try publishing them.

3DMark 2K Compare

It reports the speed as 1050MHz, but it's still at 1021MHz.

 

[Syborg1211]

ok, then how do guys explain the TOTAL difference between the two p3 700s? AND my friend at DELL told me i was right and he TOTALLY backed me up in that no two cpus are alike. I also have another friend at INTEL that told me that there could be slight differences between all cpus. I just thought it was also interesting how no software could clock my cpu and hers the same speed, hers always falling behind. And, on Sisoft Sandra, the scores were a huge difference apart, mine being better. This difference i think can cause the scores in benchmarking programs to be much different.

 

[Eli]

Syborg, a CPU, is a CPU, is a CPU, assuming it's the same core. Of course, your personal piece of silicon is diffrent, but that does not matter. a P3 running at 966MHz should score exactly the same in every benchmark you can throw at it, +- the benchmarks margin of error, given the exact same system configuration. That's the key. Exact same system configuration.

If you took two P3/700s, two of the same motherboards, made sure the BIOS settings were exactly the same, memory was exactly the same, HDD was exactly the same, video card was exactly the same, you get the point here, with a fresh install of Windows.. they would be.. exactly the same.

There are far too many variables in a system. Everything matters. The discrepency in the scores of your two systems could be caused by anything. Diffrent motherboard. Diffren't brand of RAM. Diffren't BIOS settings. Diffren't OS configurating, Diffren't HDD, diffrent anything, other than your CPU. Your CPU is actually one of the only things(the only?) that is consistant with diffrent systems. It could not be inconsistant. 700MHz with a certain archtecutre is going to give you the same exact output as every other 700MHz cpu with that same architecture, running the same lines of code.


I'm sorry, but your friend at Dell, and Intel, do not know what they're talking about. Is your friend at Dell one of those people that calls the whole box a &quot;CPU&quot;? If that's the case, then he's right.. No two boxes are alike.