To CMonster : Wintune results

[TidalWave]

Hi CMonster!

You asked for Wintune results for K6-III on the ASUS TX97-E.

Here's my setup:

K6-III 400 AFH @ 450 /2,5 volt (6X75)
ASUS TX97-E with BIOS version 112.1B
64MB 100 MHz SDRAM
Matrox Millennium II 4MB PCI
Helios Voodoo II 12 MB PCI
Quantum Fireball 4,3 GB UDMA 33

This are the test results with Wintune 98:

CPU Integer :1225,029 MIPS (Sandra says 1383)
CPU Floating Point:537,0571 MFLOPS (sandra says 539)
Video (2D) :78,78307 MPixels/s
Direct3D :35,52218 MPixels/s
OpenGL :9,520682 MPixels/s
Memory :898,8679 MB/s
Cached Disk :102,6838 MB/s
Uncached Disk :2,096456 MB/s

**I found out that with the 512KB level 3 cache on the mobo switched off, I get better results. When switched on, memory speedbenchmark drops about 15%

This are the "Tweak"-progams I use:

- MXK6OPT.EXE:
VGA access Optimizer for K6/Matrox combination. Gives me 73% speedup it says.
Benchmark 2D gives me about 11% more MPixels/s

- MGATWEAK:
Matrox MGA chipset overclocker.
Lets you handle the speed of the MGA chipset.
Gives me another 3% more MPixels/s (2D) when set to 73 instead of 68 MHz

- Voodoo2 Overclock:
Utility that lets you handle the speed of the voodoo2 chipset.
It runs at 107 MHz instead of 90 now.

When running only SETK6 with any kind of parameters, (no other optimizers running)I don't get any improvement, neither does the Powerleap Control Panel.

So I think the Write combining/Allocation is supported by the new BIOS version for the TX-97

Btw: Not bad for a 3 year 'old' setup what started out as a P75 with 8MB, HUH!!

[CMonster]

That is very good news indeed. I dont see the memory write speed posted, but I can see by the other results that it is quite good. I wonder if the it is the onboard L2 or the BIOS responsible for that. I wonder what a regular K6-2 would look like.

[TidalWave]

Yo CMonster,

The K6-III came only a week ago. Maybe you remember my story of killing my K6-2 by lapping the cap. That K6-2 was a K6-2 300 AFX @450 running at the same setup, and the story about the optimizers is the same, so I think its caused by the BIOS.

I will benchmark the separate memory speeds for you and post them again (next week).

I will also disable all optimizers so you can see what the plain system does okay?

[CMonster]

Yes, I remember your plight with the de-"lappedated" CPU.

[Amorpheous]

For comparisons sake, I thought I'd post my wintune results.

AMD k6-3 400 2.4v AHX @ 450 2.6v (4.5x100)
Tyan S1590 mobo w/ 1mb L3 cache
128MB PC100 8ns CAS2 sdram
VooDoo3 2000 PCI
SB Live
8.4GB 5400rpm Maxtor HD

MIPS: 1228.02 (sandra reports as 1406)
MFLOPS: 538.77 (sandra reports as 540)
2D Video: 104.04 Mpixels (960x720x16)
Direct3D: 170.55 Mpixels
OpenGL: 75.49 Mpixels
Memory: 916.92 MB/s
Cached Disk: 105.35 MB/s
Uncached Disk: 3.16 MB/s

I didn't try disabling the L3 cache on mobo, but it doesn't seem to make a difference in my system.
http://www.sysopt.com/forum/smile.gif

[TidalWave]

Yeah, quite the same, but my video subsystem sucks. I know. But its the best the old TX can do, and I'm happy with it (for now).

[jad1097]

This is what I got with their online test.

CPU (1) AMD K6-III @ 401 MHz / Soyo 5EMA+ V1.1 mobo
Video Board NVidia Riva TNT2
Video Mode 1024x768@32 bits/pixel
RAM 96 MB
OS Windows 2000 5.0.2195
Browser Microsoft Internet Explorer 4.0 (compatible; MSIE 5.01; Windows NT 5.0)

Area Tested Value % of Ref* % of Avg+
CPU Integer 1091 MIPS 56 % 104 %
CPU Floating Point 482 MFLOPS 62 % 102 %
CPU Advanced Integer 98.7 58 % 105 %
CPU Advanced Floating Point 53.3 32 % 104 %
CPU Advanced MMX 93.8 56 % 105 %
Video (2D) 42.7 MPixels/s 58 % 90 %
Direct3D Windowed 18.3 MPixels/s 21 % 117 %
Direct3D Null 109.6 35 % 114 %
Direct3D Primary 11.7 5 % 43 %
OpenGL 960 MPixels/s 1297 % 5454 %
Memory 771 MBytes/s 56 % 108 %
Cached Disk C:\ 183 MBytes/s 140 % 200 %
Uncached Disk C:\ 4.96 MBytes/s 141 % 177 %

Is this good or bad? What could I do to improve it.


[This message has been edited by jad1097 (edited 03-03-2000).]

[TidalWave]

JAD1097,

Try to run this 400 MHz chip @ 450!

Like ours!

[jad1097]

Can't do it http://www.sysopt.com/forum/frown.gif I tried and faid several times. I could get it to boot but it is not stable. When it was running @ 450 my benchmarks were about the same in Sandra.

I may get an K6 2-533 soon and see what happens.

[TidalWave]

CMonster, as promised, here are the memory benchmarks:

Plain system (w/o any 'tweaking')

Memory read speed: 1041 MB/s
Memory write speed: 1024 MB/s
Memory copy speed: 581 MB/s

'Tweaked' system:

Memory read speed: 1016 MB/s
Memory write speed: 1085 MB/s
Memory copy speed: 578 MB/s

So, the tweaking doesn't really affect the memory speed.

Video plain: 64 MPixels/s
Video Tweak: 78 Mpixels/s

kind of improvement here.

[CMonster]

That means your motherboard BIOS is either enabling the write allocate of the CPU or the K6-3 has a work-around for enabling the write allocate. If not too much trouble - do you have a K6-2 (CXT core) that you could test in there for comparrison? It would be very valuable to know if the K6-3 can work around the write enable limitation of some older BIOS.

Thanks,

BTW, your system looks great.

[TidalWave]

Sorry CMonster,

I can't check the K6-2 300 (it's a CTX-core) anymore, but I have a benchmark print from wintune 98.

Overall memory speed is 615,8997 MB/s. That is about 280 MB/s slower than the K6-III.

I think that is caused by the level-2 cache, because on that setup I didn't use setK6 either, since it didn't make any difference then.


However, P. Missel adviced me to run CTCM16R.exe to analyze and test the memory/cache functioning. It also measures the speed of transferring memblocks.

Since this program runs in DOS real mode, it is more accurate than a Windoze benchmark application.

It also shows whether if write allocation and write combining is enabled or not.

It's a small progam, and if you like to recieve it by E-mail, just let me know.