MVP Vs. ALi chipset 3-way shoot-out

[tazwegion]

After much debate with fellow K6plus enthusiast Jim, I began benchtesting 2 VIA MVP and 1 ALi chipset m/boards looking for the most versatile SS7 platform chipset.
I can only hope that the format in which it presented is acceptable, as the alternative was 27 individual screenshots

The contenders...
- Epox EP-MVP3C2 (512kb L2) revision 1.0
- Aopen AX59pro (512Kb L2) revision 2.36
- Gigabyte GA-5AA (512Kb L2) revision 3.2a


Static system components

MSI MS-4420 ATI rage 3D IIc (AGP)
Nova PC-133 128Mb
CXT core AMD K62-500

TEST 1: CPU mark99



TEST 2: 3Dmark 2000



TEST 3: Highest stable OC


Strangley enough the MVP3 motherboards had in the past both run @ 600Mhz, that would no longer seem to be the case here


Everest
TEST 4: Memory Read benchmark



TEST 5: Memory Write benchmark



TEST 6: Memory Latency benchmark




SiSoft Sandra 2003
TEST 7: CPU Arithmetic



TEST 8: Multimedia



TEST 9: Chipset & Cache benchmark



Summary
Interestingly enough the Epox platform came out on top of it's rivals, with the Aopen a close runner-up in the benchmarking, also worth pointing out is that the everest software caused random 're-boots' of the Aopen m/board

Epox EP-MVP3C2 5/9 benchmarks
Aopen AX59pro 2/9 benchmarks
Gigabyte GA-5AA 2/9 benchmarks

[Jim]

Interesting. Could be that your Aopen will be the next candidate for capacitor surgery.

My own tests are less complete, (Don't have 3D Mark, or CPU Mark 99); and less scientific, in that they were conducted with differing equipment and differing operating systems. However for what they are worth these were my results:

Sandra 2004 Multimedia Benchmark:

ASUS P5A-B Rev 104 w/ K6-2 550 AGR @ 5.5 x 100 w/ 768 Meg PC 133 SDRAM
Running Windows 98
Integer MMX 3022 it/s Float 3D Now! 3587 it/s

ASUS P5A-B Rev 104 w/ K6-2+ 550 @ 5.5 x 100 w/ 768 Meg PC133 SDRAM:
Running Windows 98: (Note Unofficial Serice Pack not installed)
Integer MMX 3088 it/s Float 3D Now! 3675 it/s
Running Windows XP:
Integer MMX 3112 it/s Float 3D Now! 3697 it/s

ASUS P5A-B Rev 104 w/ K6-2+ 550 @ 6 x 100 w/ 768 Meg PC133 SDRAM:
Running Windows 98: (Using Power Tweak)
Integer MMX 3458 it/s Float 3D Now! 4115 it/s
Running Windows XP:
Integer MMX 3397 it/s Float 3D Now! 4037 it/s

DFI K6BV3+/66 Rev B2 w/K6-3+ 450 @ 6 x 100 w/ 512 Meg PC133 SDRAM:
Running Windows 98: (W/ WPCredit Tweaks & Unofficial Service Pack)
Integer MMX 3433 it/s Float 3D Now! 4084 it/s
Running Windows XP: (Depending on WPCredit settings)
Integer MMX 3389 it/s Float 3D Now! 4025 it/s
Integer MMX 3388 it/s Float 3D Now! 4026 it/s
Integer MMX 3393 it/s Float 3D Now! 4031 it/s

Fic Va_503+ Rev 1.2 w/ K6-2+ 550 @ 5.5 x 100 w/ 512 Meg PC133 SDRAM
Running Windows 98:
Integer MMX 3138 it/s Float 3D Now! 3734 it/s

Sandra 2004 Arithmetic Test:

ASUS P5A-B Rev 104 w/ K6-2 550 AGR @ 5.5 x 100 w/ 768 Meg PC133 SDRAM
Running Windows 98: (Depending on bios tweaks)
Dhrystone ALU 1294 MIPS Whetstone FPU 627 MFLOPS
Dhrystone ALU 1297 MIPS Whetstone FPU 625 MFLOPS

ASUS P5A-B Rev 104 w/ K6-2+ 550 @ 5.5 x 100 w/ 768 Meg PC133 SDRAM:
Running Windows 98: (Depending on bios tweaks)
Dhrystone ALU 1323 MIPS Whetstone FPU 638 MFLOPS
Dhrystone ALU 1326 MIPS Whetstone FPU 638 MFLOPS
Running Windows XP:
Dhrystone ALU 1299 MIPS Whetstone FPU 685 MFLOPS

ASUS P5A-B Rev 104 w/ K6-2+ 550 @ 6 x 100 w/ 768 Meg PC 133 SDRAM:
Running Windows 98: (Run using Power Tweak)
Dhrystone ALU 1479 MIPS Whetstone FPU 713 MFLOPS
Running Windows XP: (Run with just bios tweaks)
Dhrystone ALU 1418 MIPS Whetstone FPU 748 MFLOPS

DFI K6BV3+/66 Rev B2 w/ K6-3+ 450 @ 6x100 w/ 512 Meg PC133 SDRAM:
Running Windows 98: (Not sure if I had the Unofficial Service Pack on or not)
Dhrystone ALU 1474 MIPS Whetstone FPU 708 MFLOPS
Running Windows XP (Results varying w/ different WPCredit settings)
Dhrystone ALU 1409 MIPS Whetstone FPU 746 MFLOPS
Dhrystone ALU 1414 MIPS Whetstone FPU 746 MFLOPS
Dhrystone ALU 1416 MIPS Whetstone FPU 747 MFLOPS

FIC VA_503+ Rev 1.2 w/ K6-2 550 @ 5.5 x 100 w/ 512 Meg SDRAM
Running Windows 98:
Dhrystone ALU 1347 MIPS Whetstone FPU 647 MFLOPS

CPU Mark99 Result
DFI K6BV3+/66 Rev B2 w/ K6-3+ 450 @ 6 x 100 w/ 512 Meg SDRAM
Running windows XP: 56.9 - 57.6
Note: Also of interest is the MVP4. Have heard of them, but never seen one. "Q": Are they worth going on an all out hunt for?

[tazwegion]

Unfortunately it's a little like comparing peachs & pears, as the K62-500 I used doesn't have the luxury of on-die L2 cache that your PLUS units enjoy, in fact to be totally fair it's had a really hard life... de-capped, 2.5v passed through it, run @ 600Mhz 24/7 12 months & damage to the cores' edges due to H/sink mounting experimentation (thankfully sorted out now)



I deliberately chose this mediocre processor & Vcard combination because I felt they would make the m/boards strengths and weaknesses more evident, oh and as far as the whole Aopen 'reboot' thing... it may be because it doesn't support such system intrusive software, as nothing else causes it to act so poorly (24/7 SETI)

Lastly when benchmarking (or any kind of scientific experimentation ) removing the variables IS what allows us to clearly see/test/measure that which is constant.

Here's a link (if you're interested) for CPUmark99 @ Downloads, which will also hopefully be hosted on my site at a later date

[xroyal1]

I've got the MVP3G2, and just ran CPUMark99 for fun = 59.4. Swell Epox version chipset. Recall when I had the FIC 503+ Epox was always scoring higher.

Not having cleaned my cache or defragged in some time, I just got the following in Sandra:

mem bandwidth - 222/220

arithmetic - 1291/667

m media - 3120/3817

[Jim]

Quite agree about the removing variables. Those results were not the outcome of an effort to compare the boards, so much as a series of tests to see what I could do in the way of bios settings, (with the ASUS), and WPCredit settings, (with the DFI), to get more performance out of built machines.

The interesting thing about them, (to me at least), is that with all else identical, the ASUS did better running WINXP than it did running WIN98 at 550 Mhz; but not at 600. ( At the time I was not using the "Unofficial WIN98-Se Service Pack" so 98 would have been limited to 256 Meg of Ram).

Also, that the ASUS did better with a K6-2+ and 768 Meg of Ram, than than the DFI did with a K6-3+ and 512 Meg of Ram, with both running at the same speed. I don't have a third stick of Ram that is compatable with the DFI, and the ASUS machine was torn down and rebuilt as the DFI so I can't very easily run 2 stick tests with it now.

Additionally, (now that Ihave dug out the results from my 98 side), that the FIC VA_503+ (which was only tested at 550 Mhz), outscored both the DFI and the ASUS at that level.

The question raised by my results being: Does the ASUS outperform the DFI? or is the addition of an extra 256 Meg of Ram sufficient to put it in front despite running an inferior processor?

PS: Tried link, - not working , - at least not now; but thanks for trying.
EDit: Working now, got it, and thanks again!

[xroyal1]

FYI: cleaned out a big load of files, defragged, cleaned registry and ran benches again. Sandra mem bandwidth and multi media scores slightly lower, arithmetic the same.

[Jim]

I always got the best results with 98 by first, (once only), running Norton disk optimizer, (which moves the swap file to the front of the drive), then by defragging first with Norton speeddisk, then again with the Win 98 defragger. (Note all the test results shown in my previous posts were taken from tests run with pretty fresh defrags). The XP side, I can't run Norton, so I would run diskeeper 8 repeatedly to get the best defrag I could. Also worth noting , is that according to Sandra, you have to be running the same version of Sandra to get comparable results.

[tazwegion]

I've got the MVP3G2, and just ran CPUMark99 for fun = 59.4. Swell Epox version chipset. Recall when I had the FIC 503+ Epox was always scoring higher.

Yeah that seems about right, the site/forum I originally got CPUmark99 from had threads/posts accounting for Athlon 700's being outrun by OC'ed K63+'s @600Mhz anyways here's a graph denoting some average scores on various platforms



For the record my Barton @ 2.2Ghz pulls a 202, while the T'bird @ 1.2Ghz is a mere 101

[xroyal1]

taz, thanks. This little K6 III 450+ I picked up for $30 never ceases to amaze me relative to how many basic tasks one can achieve. Coupled with my Epox with the latest bios options, it's a hummer. Having it cooled with an 80x80 + Rain, t'is cool and very quiet.

My 2400 Barton mobile overclocked to 2400 gave me a cpumark score of about 240.

[tazwegion]

Those Barton 'mobiles' are a sweet number! and having lower V/core is always an advantage

[Guest]

You bet.

Truth is I've since sold the 2400 mobile, and temporarily relying on my K6 rig. I've got a Duron 1600 that is very peppy which I may mate to a Biostar M7NCG 400 matx MB in a very small case. I'm still holding off K8 while that arena is changing so much. If the Biostar works out, I may just drop in a 2600 mobile, and ride the K7 wave even longer. In the meantime, my trusty K6 rig rules, especially while I use our summer to get some household projects done. Winter is my rig tinkering time.

[xroyal1]

The "Guest" was me who dumped this cookie by mistake.

Expect it's chilly downunder about now.

[tazwegion]

Indeed it is... 0-5c overnight, up to 15c during the day with thunderstorms recently etc. we cop it bad due to our location atop a mountain range

In fact... the 'PC room' is the warmest in the house (except for the lounge with it's wood combustion fire) 4/6 PC's on 24/7

[Guest]

Sounds you're settled in for some winter fun. I live near the foot of a 5-6K high mountain range with a 10K peak, another 14K peak 75 miles away in California. I'm a mountain/lake man at heart.

Our nights are cool and dry most every night, but I just got a room a/c unit for sleeping when we have 100F temps. Thought of putting one of my rigs in that bedroom, but concerned about power draws when the a/c kicks in, not wanting to lock up the rig. It's only a 5050 btu/hr unit with a 4.8 amp rating. Pretty sure that room is on a 15 amp circuit, so theoretically it should work.

Happy winter on you!

[tazwegion]

And happy summer to you!