G400MAX beatan by a TNT2U in Q3??????????

[Hex]

I dont know if this is going to be of any help. But i ran Wcpuid and have posted the results bellow. Anything not look right for an AMD Athlon system?

[ WCPUID Ver.2.7a (c) 1996-2000 By H.Oda! ]

Processor 1 : AMD Athlon (Model 1) / 6C84107A
Platform : Slot A (SC242 Connector)
Vendor ID : AuthenticAMD
Type ID : 0 (0)
Family ID : 6 (7)
Model ID : 1 (1)
Stepping ID : 2 (2)
Brand ID : 0 (0)

Clock Frequency
Internal : 704.92 MHz
External : 201.41 MHz DDR
Multiplier : 3.5

L1 I-Cache : 64K
L1 D-Cache : 64K
L2 Cache : 512K Byte
L2 Speed : 1/2 H4 (1/2 H4, 1/2 S4)

MMX : Supported
SIMD : Not Supported
3DNow! : Supported

Machine : IBM PC/AT
Version : Windows 98 Version 4.10.2222 A

----------- : --------
F.Flags : C0C1F9FF 80000006
I@PCP : 00000000 00000000 00000000 00000000
: 00000000 00000000 00000000 00000000
A@EXT : 0408FF08 FF18FF10 40020140 40020140
: 00000000 41004100 02002140 00000000
A@Cx10 : 00160604 CB02810F 60000001 5A1421D5 222402DD
A@Cx15 : 08601000 00000000 00000000 E0000018 F8000800
A@Cx1Z : 08000000 1CE20223 000A0000 00000000 00000001

[ WCPUID Ver.2.7a (c) 1996-2000 By H.Oda! ]
(Processor 1)
<< Standaed & Extended Feature Flags >>

Floating-point unit on-chip (FPU) : Supported
Virtual Mode Extension (VME) : Supported
Debugging Extension (DE) : Supported
Page Size Extension (PSE) : Supported
Time Stamp Counter (TSC) : Supported
Model Specific Registers (MSR) : Supported
Physical Address Extension (PAE) : Supported
Machine Check Exception (MCE) : Supported
CMPXCHG8 Instruction (CX8) : Supported
On-chip APIC Hardware (APIC) : Not Supported
Fast System Call (SEP) : Not Supported
Memory Type Range Registers (MTRR) : Supported
Page Global Enable (PGE) : Supported
Machine Check Architecture (MCA) : Supported
Conditional Move Instruction (CMOV) : Supported
Page Attribute Table (PAT) : Supported
36-bit Page Size Extension (PSE36) : Not Supported
96-bit Processor Number (PN) : Not Supported
MMX Instructions Extensions (MMX+) : Supported
MMX Instructions (MMX) : Supported
Fast floating-point save and restore (FXSR) : Not Supported
Streaming SIMD Extensions (XMM) : Not Supported
3DNow! Instructions Extensions (3DNow!+): Supported
3DNow! Instructions (3DNow!) : Supported

[ WCPUID Ver.2.7a (c) 1996-2000 By H.Oda! ]
(Processor 1)
<< Cache Info. >>

[L1 Instruction TLB]
2-Mbyte/4-Mbyte Pages, fully associative, 8 entries
4-Kbyte Pages, fully associative, 16 entries

[L1 Data TLB]
2-Mbyte/4-Mbyte Pages, 4-way set associative, 8 entries
4-Kbyte Pages, fully associative, 24 entries

[L1 Instruction cache]
64K byte cache size, 2-way set associative, 64 byte line size, 1 line par tag

[L1 Data cache]
64K byte cache size, 2-way set associative, 64 byte line size, 1 line par tag

[L2 Unified cache]
512K byte cache size, 2-way set associative, 64 byte line size, 1 line par tag

[L2 Instruction/Unified TLB]
2-Mbyte/4-Mbyte Pages, Off, 0 entries
4-Kbyte Pages, 4-way set associative, 256 entries

[L2 Data TLB]
2-Mbyte/4-Mbyte Pages, Off, 0 entries
4-Kbyte Pages, 4-way set associative, 256 entries

[ WCPUID Ver.2.7a (c) 1996-2000 By H.Oda! ]
<< AGP Chip set Info. >>

Chip-set Device : AMD-751
Chip-set Revision : 23
AGP Revision : 1.0
AGP Operations : Enable

Data Transfer Rate
Supported : 1x 2x
Current : 2x

Side band addressing : Enable
Aperture Size : 256MB

-------- -------- -------- --------
70061022 22100006 06000023 00804000
00000000
00100002 0F000203 00000302 00010005
00020000 00000000 00000000 00000000


------------------
"There's no such thing as a stupid question, just stupid people"


[This message has been edited by Hex (edited 09 January 2000).]

[Fyrespray]

Ok we have both taken Trilinear off and gone to bilinear the results are now

Me :59.6 FPS
Hex :54.0 FPS

well we are certainly getting closer but my system in only a celery 550 whereas his is an Athlon 700 so it still looks like something isn't right, lets face it I should not be beating his scores

Fyrespray

------------------
Celery 366 @ 550,Abit BM6, 128Mb PC100 Ram, G200 8Mb, AWE64 Gold, Win98SE.

[Rags]

I don't know what else to try except what I have already suggested to you....which TurboGL are you using?

Maybe when Greebe gets online, he can help you tweak that properly.

Rags

[Hex]

Ok. Im beginning to think that maybe my system as a whole is running slower than it should be. Can someone recomend me a good bit of benchmarking software (preferably free) that i can test my system as a whole, Not just games, with and post the results back.

----Hex


------------------
"There's no such thing as a stupid question, just stupid people"

[Rags]

Stay tuned, you will get some help in a few minutes

[Hex]

OK. Cheers.

----Hex

[paulcs]

This is really weird. I just noticed that when I enter the game using the test config and the G400, the game is set to trilinear filter as it should be. However, when I use an identical copy of the config, and start the game with the TNT2 machine, it defaults to *bilinear* filtering.

I have a separate, completely segregated, clean copy of the config file I downloaded, and I have reloaded it (and my CD code!) several times, in both machines, with the same results. This makes my previous TNT2 numbers invalid, as I assumed the game would open with trilinear filter, because the config is set up that way.

Here are some new numbers with the texture slider back up to Max and bilinear filtering forced manually on the TNT2 Ultra:

G400/PIII 600/Trilinear: 40.9
G400/PIII 600/Bilinear: 58.6

TNT2U/PIII 500/Trilinear: 56.7
TNT2U/PIII 500/Bilinear: 57.5

Trilinear filtering is the killer for me. My Max takes a hit of about 30% going from bilinear to trilinear filtering. My TNT2 Ultra takes a hit of less than one frame per second.

Either the Max has a serious problem with trilinear filtering on my system, the TNT2U is particularly good with this setting, or nVidia is fudging somehow.

If the latter is true, this is not a valid test.

Paul
paulcs@flashcom.net

[Rags]

Hi Paul,

The reason the TNT2 doesn't show a huge decrease is because it doesn't really use trilinear filtering, it will run with trilinear set, and will show some improvement, but for the most part it is not working. In 800X600 with all the goodies on, the textures won't fit into the video memory with trilinear set to enabled, that is why the G400 shows a performance hit, it is using AGP texturing.

Rags

[paulcs]

Well, then you were right and the TNT2 is fudging.

As far as the science of this goes, we are measuring the performance hit the G400 takes using trilinear filtering; on the other hand, we are measuring how well nVidia fudges trilinear filtering. The comparisons are not at all valid using this setting.

NVidia bastards.

Paul
paulcs@flashcom.net

[Greebe]

OK I set my Athlon to 700 (1:2 latency) PD 5.41 TGL, MSI6167 MB, using your config file...

trilinear = 54.8
bilinear = 67.1

[Fyrespray]

We originally had that problem too, if you change the name of r_lastvalidrenderer in the cfg file between g400 and the nvidia entry (can't remem it offhand) that should solve the problem of it resetting it to defaults. We think we might have gone someway to finding out what is happening though, this is a shot of memory benchmarks from Hex's machine, and mine. My machine seems to be beating his which is definetly not right. Any Athlon owners out there have any ideas ???

we did this benchmark with sisoft sandra available <a href=http://download.cnet.com/downloads/0-10102-101-915435.html>Here</a>

<IMG SRC="http://www.fyre.nildram.co.uk/benchmarks.jpg"">

This is relly confusing us now please somebody help

ok i know the jpg looks a bit shitty but hey i figured you don't really care about eye candy and nicly lined up boxes with a benchmark screenie

Fyrespray

Also thanks for all the help you have been giving us so far its definetly keeping us sane

------------------
Celery 366 @ 550,Abit BM6, 128Mb PC100 Ram, G200 8Mb, AWE64 Gold, Win98SE.



[This message has been edited by Fyrespray (edited 10 January 2000).]

[Rags]

The memory performance of the Athlon motherboards cannot compare to the performance of the Intel 440BX chipset, that is why that particular benchmark is awry.

Rags

[Greebe]

I'll second that Rags!

[Fyrespray]

Ok but it still looks a bit close to an AMD K6-3 500, what sort of benchmarks would you suggest we take to compare to see if it is his machine that is not configured properly in some way??

Fyrespray

------------------
Celery 366 @ 550,Abit BM6, 128Mb PC100 Ram, G200 8Mb, AWE64 Gold, Win98SE.

[Greebe]

Well I decided to OC me Max and reran the tests...

bi =79.2
tri =65.2

What does his score on the CPU test in Sandra?

I'll lower mine for the same...

------------------
Designer of the first commercial Athlon mod.
Anyone up for an Eclipse'd Athlon? Email me (epc@netdirect.net) for more information

3000 mips, 1.172 gflops @ 900 mHz 1:3 latency
CPUmark99 75
FPU Winmark 4890

3Dmark 99 Max (8x6 16, 16, triple) PD5.41
8000
15181