H3nrik V! wrote on 2021-10-27, 04:33:

So, a bit faster few places with 64 vs 128 megs? But as for what I can see, there is nothing being slower with 64 megs? It looks like then, that it might be that the L3 actually only can cache the first 64 megs ..

3dmark 99 scored highest in the CPU score section with 128MB and cache turned on.

This weekend I'm going to test out some other graphics cards that have VGA connections, I had this machine at my main desk which has a DVI/PS/2 KVM so I needed DVI for now. Ultimately I'll be using it with a Lenovo LCD screen that has both DVI and VGA.

EDIT - Added a pic of the screen I'm going to be using, it has a neat little set of speakers that attach to the bottom of the screen, they don't sound awesome but they are acceptable. I probably paid way too much to get those on ebay.

pete8475 wrote on 2021-10-27, 01:04:

Awesome! I'm sure it was the same guy if it was in the box, mine was too. […]
Show full quote

Brawndo wrote on 2021-10-26, 23:09:

Funny, I bought the same case on ebay, probably from the same seller. 🤣 I also paid more for the case than I probably should have, but how often are you going to find a NOS AT case? Mine is still in the box, no idea what I'm going to use it for yet. I have an ASUS TX-97 mobo laying around not currently in service, guess I'll use that.

Awesome! I'm sure it was the same guy if it was in the box, mine was too.

Onto the 64meg benchmarks:

64MB with motherboard Cache On:
3dmark 99 - 1430 3dmarks, 4837 cpumarks
3dmark 2k - 805 3dmarks, 41 cpumarks
Quake 3 - 10.0 fps! TEN!
3dbench 1.0c - 348.3
Chris's 3d Benchmark - 466.8 (280.0 fps)
Chris's 3d Benchmark 640x480 - 103.1 (61.8 fps)
Quake timedemo - 66.6 fps
Landmark System Speed Test - 585 MHz AT with a 721 MHz 80287
Topbench - 282

64MB with motherboard Cache Off:
3dmark 99 - 1203 3dmarks, 4429 cpumarks
3dmark 2k - 625 3dmarks, 35 cpumarks
Quake 3 - 7.8 fps
3dbench 1.0c - 347.8
Chris's 3d Benchmark - 430.0 (258.0 fps)
Chris's 3d Benchmark 640x480 - 84.5 (50.7 fps)
Quake timedemo - 59.2 fps
Landmark System Speed Test - 585 MHz AT with a 721 MHz 80287
Topbench - 279

EDIT - I bolded all the values that scored highest of all in this post, please keep in mind all tests were only run ONCE in each config though. Really good testing should be 3 or more runs and averaged but I just don't have time for that.

Made more a difference than I figured it would. Looks like motherboard cache on is better regardless. I have a K6-3+ build that I run 128mb of ram on a PC Chips M571, I also keep my cache enabled but hadn't done a ton of A/B testing.

pete8475 wrote on 2021-10-27, 01:04:

Onto the 64meg benchmarks: […]
Show full quote

Onto the 64meg benchmarks:

64MB with motherboard Cache On:
...
3dmark 2k - 805 3dmarks, 41 cpumarks

Your scores seem to be on the low side, although I can only compare the 3DMark2000 one. My similar system at 450MHz gets some ~1550 points currently, give or take. Are you running the benchmark in 32-bit color?

I know you have a K6-II+ - but I doubt the difference can be explained by the difference between that and a K6-III+. The Radeon 7000 should also be quite a bit faster than a Voodoo3.

leonardo wrote on 2021-10-27, 18:36:

pete8475 wrote on 2021-10-27, 01:04:

Onto the 64meg benchmarks: […]
Show full quote

Onto the 64meg benchmarks:

64MB with motherboard Cache On:
...
3dmark 2k - 805 3dmarks, 41 cpumarks

Your scores seem to be on the low side, although I can only compare the 3DMark2000 one. My similar system at 450MHz gets some ~1550 points currently, give or take. Are you running the benchmark in 32-bit color?

I know you have a K6-II+ - but I doubt the difference can be explained by the difference between that and a K6-III+. The Radeon 7000 should also be quite a bit faster than a Voodoo3.

You got me to wondering, so I went back to my own thread where I had ran the same benchmarks and sure enough I was getting roughly 1500ish points in 3DMark 2K as well.
I've got a K6-3+ as well as a Voodoo 3 (PCI) but my board isn't very fancy and only runs at 83mhz FSB.
I would say he must be running it at 32bit color depth, otherwise that is some terrible performance for a Radeon 7000 and has to be down to driver overhead.

My results:

3DMark 99 Max Default Run 800x600 16bit:
2750 3D Marks
6501 CPU 3D Marks

3DMark 2000: Default Run 1024x768 16bit:
1509 3DMarks

leonardo wrote on 2021-10-27, 18:36:

Your scores seem to be on the low side, although I can only compare the 3DMark2000 one. My similar system at 450MHz gets some ~1550 points currently, give or take. Are you running the benchmark in 32-bit color?

I know you have a K6-II+ - but I doubt the difference can be explained by the difference between that and a K6-III+. The Radeon 7000 should also be quite a bit faster than a Voodoo3.

I simply left it on the default settings, so presumably 32bit.

I'll boot it up and run 3dmark in both color depths to see what the performance difference is.

EDIT - It defaults to 1024x768 16 bit, running it now on that setting.

EDIT 2 - So after running 3dmark 2k on default this time it scored LOWER, 788 points. In 32bit there was almost no difference, 772 3dmarks.

Decided to test an MX4000 in this PC now that it's got a screen with VGA input (no DVI on this card).

3dmark 99 - 1881, 4931 cpumarks
3dmark 2k - 1508, 107 cpu 3dmarks (first run after installing this card got 1809, 119 cpu 3dmarks for some reason)

pete8475 wrote on 2021-10-28, 00:03:

Decided to test an MX4000 in this PC now that it's got a screen with VGA input (no DVI on this card).

3dmark 99 - 1881, 4931 cpumarks
3dmark 2k - 1508, 107 cpu 3dmarks (first run after installing this card got 1809, 119 cpu 3dmarks for some reason)

Dang. Is that Radeon a 32bit memory version or something weird?
It seems crazy to me that it is THAT much slower than the Geforce card.

Jasin Natael wrote on 2021-10-29, 17:13:

Dang. Is that Radeon a 32bit memory version or something weird?
It seems crazy to me that it is THAT much slower than the Geforce card.

It uses a 64-bit memory interface like the GeForce MX4000. Seems to be related to 3DNow! optimizations which only nVidia and 3Dfx did bother to implement.

Jasin Natael wrote on 2021-10-29, 17:13:

Dang. Is that Radeon a 32bit memory version or something weird?
It seems crazy to me that it is THAT much slower than the Geforce card.

64 bit according to Everest.

Interestingly 3dmark 2001 reports it as being "software T&L" though, so perhaps it's a "budget" or crap version of the card.

pete8475 wrote on 2021-10-30, 06:13:

64 bit according to Everest.

Interestingly 3dmark 2001 reports it as being "software T&L" though, so perhaps it's a "budget" or crap version of the card.

The lack of hardware T&L can also affect negatively the performance if that's the case.

Gmlb256 wrote on 2021-10-30, 13:28:

The lack of hardware T&L can also affect negatively the performance if that's the case.

That's exactly what I was thinking.

Looking at the wiki page on the 7000 series cards it would seem this is the "entry level" aka trash level card. https://en.wikipedia.org/wiki/Radeon_R100_series

The Radeon 7000 definitely lacks Hardware T&L, had one a long time and it was a slow card.

Jasin Natael wrote on 2021-10-29, 17:13:

pete8475 wrote on 2021-10-28, 00:03:

Decided to test an MX4000 in this PC now that it's got a screen with VGA input (no DVI on this card).

3dmark 99 - 1881, 4931 cpumarks
3dmark 2k - 1508, 107 cpu 3dmarks (first run after installing this card got 1809, 119 cpu 3dmarks for some reason)

Dang. Is that Radeon a 32bit memory version or something weird?
It seems crazy to me that it is THAT much slower than the Geforce card.

...or my Voodoo3 for that matter. Then again the GeForce is currently tied with the V3, so probably that's about as much as you can squeeze out of any 3D-accelerator with these K6-II/K6-III CPUs? At least now he's got something on-par with what you'd.expect. Half the score of a V3 with a newer video card just seemed wrong.

leonardo wrote on 2021-10-31, 21:37:

Jasin Natael wrote on 2021-10-29, 17:13:

pete8475 wrote on 2021-10-28, 00:03:

Decided to test an MX4000 in this PC now that it's got a screen with VGA input (no DVI on this card).

3dmark 99 - 1881, 4931 cpumarks
3dmark 2k - 1508, 107 cpu 3dmarks (first run after installing this card got 1809, 119 cpu 3dmarks for some reason)

Dang. Is that Radeon a 32bit memory version or something weird?
It seems crazy to me that it is THAT much slower than the Geforce card.

...or my Voodoo3 for that matter. Then again the GeForce is currently tied with the V3, so probably that's about as much as you can squeeze out of any 3D-accelerator with these K6-II/K6-III CPUs? At least now he's got something on-par with what you'd.expect. Half the score of a V3 with a newer video card just seemed wrong.

Yeah I can buy roughly the same speed as a V3 with that CPU, but half seems off. But looks like it is mystery solved.

So I've had an Asus TXP4 sitting around for a while now and decided why not compare the speed of these parts in that board vs. this one.

So I did some quick benchmarks after swapping over everything except the RAM. The RAM in the TXP4 is a single 256MB stick of PC100 CL2 Kingston memory.

Results:
VX97
3dmark 99 - 1881, 4931 cpumarks
3dmark 2k - 1508, 107 cpu 3dmarks
Quake timedemo - 65.2 fps
Quake 3 (Radeon VE) - 10.0 fps
Topbench - 282
3dbench 1.0c - 348.3
Chris's 3d Benchmark - 466.8 (280.0 fps)
Chris's 3d Benchmark 640x480 - 103.1 (61.8 fps)
Speedsys 4.78 Processor Benchmark - 509.67
Speedsys 4.78 Memory Bandwidth - 119.87 MB/s

TXP4 256MB PC100 CL2 SDRAM
3dmark 99 - 2262, 5887 cpumarks
3dmark 2k - 1981, 133 cpumarks
Quake timedemo - 78.0 fps
Quake 3 (Radeon VE) - 12.2 fps
Quake 3 (MX4000) - 27.8 fps
Topbench - 304 (451 was incorrect and is generated when restarting into DOS from 98)
3dbench 1.0c - 241.8
Chris's 3d Benchmark 478.8 - (278.3 fps)
Chris's 3d Benchmark 640x480 106.1 - (63.6 fps)
Speedsys 4.78 Processor Benchmark - 516.76
Speedsys 4.78 Memory Bandwidth - 218.23 MB/s

Almost across the board graphics and processor improvements in these tests. That 3dbench program reports a lower number for some reason though.

I've also ordered an MX4000 with DVI from ebay, presumably otherwise it's the same as the current one.

EDIT - Eventually I am going to try single sticks of 64mb and 128MB SD and the EDO memory from the VX97.
EDIT 2 - I was also hoping some of the PCI cards (various USB 2 cards, several ATI and Matrox graphics cards) I have that don't work in the VX97 would work in this board but none do.

Trying to find my bag of small sdram modules right now and not having much luck so far, but when I do I'll do some 64mb testing.

EDIT - Memory found, running some benchmarks now.

TXP4 64MB PC100 CL2 SDRAM
3dmark 99 - 2356, 5952 cpumarks - higher than 256
3dmark 2k - 2200, 144 cpumarks - also higher than 256
Quake timedemo - 76.8 fps
Quake 3 (MX4000) - 27.8 fps
Topbench - 304
3dbench 1.0c - 241.5
Chris's 3d Benchmark - 479.9 (287.9 fps)
Chris's 3d Benchmark 640x480 - 106.1 (63.6 fps)
Speedsys 4.78 Processor Benchmark - 516.76
Speedsys 4.78 Memory Bandwidth - 218.23 MB/s

TXP4 128MB PC100 CL2 SDRAM
3dmark 99 - 2315, 5961 cpumarks
3dmark 2k - 2150, 141 cpumarks
Quake timedemo - 77.5 fps
Quake 3 (MX4000) - 27.8 fps
Topbench - 304
3dbench 1.0c - 241.6
Chris's 3d Benchmark - 479.9 (287.9 fps)
Chris's 3d Benchmark 640x480 - 106.1 (63.6 fps)
Speedsys 4.78 Processor Benchmark - 516.76
Speedsys 4.78 Memory Bandwidth - 218.23 MB/s

EDIT 2 - I figured out what's up with that high Topbench score on my 256MB original run, if I boot into Windows 98 and then restart in MS-DOS mode it gives a score of 451 regardless of what ram is in the system. If I boot straight to DOS it gives the lower and probably more accurate numbers.

EDIT 3 - It seems to me that 128MB of CL2 SD-RAM is the optimal config, at least for the hardware I have so I'm going to go with that.

pete8475 wrote on 2021-10-26, 05:05:

Running Some benchmarks: […]
Show full quote

Running Some benchmarks:

MB Cache On:
3dmark 99 - 1388 3dmarks, 4906 cpumarks
3dmark 2k - 805 3dmarks, 42 cpumarks
Quake 3 - 9.8 fps
3dbench 1.0c - 348.3
Chris's 3d Benchmark - 466.8 (280.0 fps)
Chris's 3d Benchmark 640x480 - 103.1 (61.8 fps)
Quake timedemo - 65.2 fps
Landmark System Speed Test - 585 MHz AT with a 721 MHz 80287
Topbench - 282

Speedsys 4.78 (don't know what I'm looking for in here tbh):
Processor Benchmark - 509.67
Memory Bandwidth - 119.87 MB/s

MB Cache Off:
3dmark 99 - 1230 3dmarks, 4403 cpumarks
3dmark 2k - 631 3dmarks, 35 cpumarks
Quake 3 - 8.1 fps
3dbench 1.0c - 347.8
Chris's 3d Benchmark - 430.1 (258.0 fps)
Chris's 3d Benchmark 640x480 - 84.4 (50.6 fps)
Quake timedemo - 59.2 fps
Landmark System Speed Test - 585 MHz AT with a 721 MHz 80287
Topbench - 279

Speedsys 4.78 (don't know what I'm looking for in here tbh):
Processor Benchmark - 509.67
Memory Bandwidth - 119.87 MB/s

EDIT - So this is just 20 or 30 minutes or so of playing around with becnhmarks and I think it's safe to conclude that in this case motherboard cache turned on is much faster. Also if there's any specific test you want me to do please let me know and I'll run it!

Interesting how on some mobos it’s faster off and some it’s faster on. So best for everyone to try it both ways and see.

pete8475 wrote on 2021-11-13, 23:53:

So I've had an Asus TXP4 sitting around for a while now and decided why not compare the speed of these parts in that board vs. t […]
Show full quote

So I've had an Asus TXP4 sitting around for a while now and decided why not compare the speed of these parts in that board vs. this one.

So I did some quick benchmarks after swapping over everything except the RAM. The RAM in the TXP4 is a single 256MB stick of PC100 CL2 Kingston memory.

Results:
VX97
3dmark 99 - 1881, 4931 cpumarks
3dmark 2k - 1508, 107 cpu 3dmarks
Quake timedemo - 65.2 fps
Quake 3 (Radeon VE) - 10.0 fps
Topbench - 282
3dbench 1.0c - 348.3
Chris's 3d Benchmark - 466.8 (280.0 fps)
Chris's 3d Benchmark 640x480 - 103.1 (61.8 fps)
Speedsys 4.78 Processor Benchmark - 509.67
Speedsys 4.78 Memory Bandwidth - 119.87 MB/s

TXP4 256MB PC100 CL2 SDRAM
3dmark 99 - 2262, 5887 cpumarks
3dmark 2k - 1981, 133 cpumarks
Quake timedemo - 78.0 fps
Quake 3 (Radeon VE) - 12.2 fps
Quake 3 (MX4000) - 27.8 fps
Topbench - 304 (451 was incorrect and is generated when restarting into DOS from 98)
3dbench 1.0c - 241.8
Chris's 3d Benchmark 478.8 - (278.3 fps)
Chris's 3d Benchmark 640x480 106.1 - (63.6 fps)
Speedsys 4.78 Processor Benchmark - 516.76
Speedsys 4.78 Memory Bandwidth - 218.23 MB/s

Almost across the board graphics and processor improvements in these tests. That 3dbench program reports a lower number for some reason though.

I've also ordered an MX4000 with DVI from ebay, presumably otherwise it's the same as the current one.

EDIT - Eventually I am going to try single sticks of 64mb and 128MB SD and the EDO memory from the VX97.
EDIT 2 - I was also hoping some of the PCI cards (various USB 2 cards, several ATI and Matrox graphics cards) I have that don't work in the VX97 would work in this board but none do.

That memory bandwidth increase is massive.

Jasin Natael wrote on 2021-11-15, 16:34:

pete8475 wrote on 2021-11-13, 23:53:

So I've had an Asus TXP4 sitting around for a while now and decided why not compare the speed of these parts in that board vs. t […]
Show full quote

So I've had an Asus TXP4 sitting around for a while now and decided why not compare the speed of these parts in that board vs. this one.

So I did some quick benchmarks after swapping over everything except the RAM. The RAM in the TXP4 is a single 256MB stick of PC100 CL2 Kingston memory.

Results:
VX97
3dmark 99 - 1881, 4931 cpumarks
3dmark 2k - 1508, 107 cpu 3dmarks
Quake timedemo - 65.2 fps
Quake 3 (Radeon VE) - 10.0 fps
Topbench - 282
3dbench 1.0c - 348.3
Chris's 3d Benchmark - 466.8 (280.0 fps)
Chris's 3d Benchmark 640x480 - 103.1 (61.8 fps)
Speedsys 4.78 Processor Benchmark - 509.67
Speedsys 4.78 Memory Bandwidth - 119.87 MB/s

TXP4 256MB PC100 CL2 SDRAM
3dmark 99 - 2262, 5887 cpumarks
3dmark 2k - 1981, 133 cpumarks
Quake timedemo - 78.0 fps
Quake 3 (Radeon VE) - 12.2 fps
Quake 3 (MX4000) - 27.8 fps
Topbench - 304 (451 was incorrect and is generated when restarting into DOS from 98)
3dbench 1.0c - 241.8
Chris's 3d Benchmark 478.8 - (278.3 fps)
Chris's 3d Benchmark 640x480 106.1 - (63.6 fps)
Speedsys 4.78 Processor Benchmark - 516.76
Speedsys 4.78 Memory Bandwidth - 218.23 MB/s

Almost across the board graphics and processor improvements in these tests. That 3dbench program reports a lower number for some reason though.

I've also ordered an MX4000 with DVI from ebay, presumably otherwise it's the same as the current one.

EDIT - Eventually I am going to try single sticks of 64mb and 128MB SD and the EDO memory from the VX97.
EDIT 2 - I was also hoping some of the PCI cards (various USB 2 cards, several ATI and Matrox graphics cards) I have that don't work in the VX97 would work in this board but none do.

That memory bandwidth increase is massive.

Indeed, however I get 175.57 MB/s for the memory bandwidth in SpeedSys using the exact same CPU with just EDO RAM (still lower than the ASUS TXP4 using SDRAM though). I think that the EDO RAM used on the ASUS VX97 may not handle the optimal memory configuration with 75MHz FSB set.

Gmlb256 wrote on 2021-11-15, 17:06:

Jasin Natael wrote on 2021-11-15, 16:34:

That memory bandwidth increase is massive.

Indeed, however I get 175.57 MB/s for the memory bandwidth in SpeedSys using the exact same CPU with just EDO RAM (still lower than the ASUS TXP4 using SDRAM though). I think that the EDO RAM used on the ASUS VX97 may not handle the optimal memory configuration with 75MHz FSB set.

Very likely.
My humble little PC Chips m571 with the SiS 5598 doesn't do that well even with the 83mhz FSB it runs at.

Superscape 1.0 Low Res:
Superscape 1.0 High Res: 587.5
Chris' 3D Bench Low Res: 627.6 (376.5FPS)
Chris' 3D Bench High Res: 116 (69.6FPS)
PC Player Bench Low Res: 162.3FPS
PC Player Bench High Res: 41.2FPS
DOOM Minimum: 2134 in 140 realtics
DOOM Maximum: 2134 in 510 realtics
Quake Low Res: 88.5
Quake Medium Res: 38.0
Quake High Res: 22.2FPS
Topbench: 519
Speedsys CPU: 564.46
Speedsys Memory Bandwidth - 186.69 MB/s
Speedsys Throughput - 122.57MB/s