Review: ABIT VP6 Writeup

ABIT VP6 Writeup

Introduction

The arrival of a new motherboard from Abit usually brings with it much trumpeting and fanfare from all corners of the web. New motherboards from Abit are anticipated almost as much as the next game from Epic or id Software.

So what's different about the new VP6, Abit's latest dual processor offering for the masses? I've been anticipating this board since I upgraded from Abit's now legendary BP6, it's last dual processor board. So why can I only find a preview of a pre-release board by Kyle over at the [H]ardOCP?

I usually buy a new computer part after reading many reviews from the respected tech sites across the web. I'll check that the majority are in agreement that a product is good before handing over the hard earned. But since my last 2 motherboards have been Abit boards, I thought I'd take the plunge and grab the VP6 anyway. I yearned for the good old days of the BP6 and the power of dual CPU's!

So my housemate and I took the plunge and each ordered a VP6 and a pair of matching Intel FC-PGA P3-700's (SL45Y's). They arrived as promised a couple of days later, and for the past 3 days, we've been testing and retesting to see exactly what these boards are like.

Anyway, enough rambling and on with the good stuff :)

Specifications

I don't need or wish to provide you with every single one of the boards specs. Abit's site has the full specification. However I will provide you with the juicy bits!

These are taken directly from the VP6 manual:

CPU

• Supports Intel Pentium III 500~1GHz (Based on FC-PGA package, both dual or single processors)

Chipset

• VIA chipset (VT82C694X and VT82C686B)
• Supports Ultra DMA/33, Ultra DMA/66 and Ultra DMA/100 IDE protocol
• Supports Advanced Configuration and Power Management Interface (ACPI)
• Accelerated Graphics Port connector supports AGP 1X/2X/4X mode (Sideband) 3.3V device

HighPoint Technologies, Inc. HPT370 chipset

• Supports ATA-100 specification
• Supports RAID levels 0,1,0+1

Memory (System Memory)

• Four 168-pin DIMM sockets supporting up to 2Gb memory
• Supports ECC memory

• Abit VP6 dual processor motherboard
• 2 x Intel Pentium III FC-PGA 700Mhz CPU's (SL45Y) running at 933Mhz (7x133)
• 2 x GlobalWin FOP-38 heatsink and fan combos
• 384Mb Cube PC133 CAS2 memory
• 2 x IBM Deskstar DTLA-307045 46.1Gb 7200rpm UDMA/100 disks
• Creative Labs Graphics Blaster TNT2 Ultra
• Soundblaster Live! Platinum

A fresh install of Win2K was done using the current official v4.25 VIA 4-in-1 drivers which work great on the VP6. This was due to us having problems with the shipped v4.26 drivers on the VP6 CD. We ran all benchmarks 3 times to get an average score. Then we installed the latest 4.26 drivers from the CD and repeated the benchmarks to see what impact the new drivers would have on performance.

We must admit that installing Win2K and using the 4.25 drivers, then upgrading to the 4.26 drivers gave us no problems. However a fresh install of Win2K and the 4.26 drivers gave us problems in almost everything we ran, from Counter Strike to 3DMark. Your mileage may vary however.

Windows 2000 Professional was used for all tests. 6.18 of the Detonator 3's were used with the TNT2 Ultra. Onto the numbers.

Benchmarks

Firstly, a quick shot of WCPUID to verify that we're running 2 CPU's successfully at 933Mhz.



In the tables below, you'll see 3DMark2000 and Quake3 (v1.17) scores for both 4.25 and 4.26 driver sets, to highlight the performance differential between the two. No screenshots, since we're going to use our screenshot budget for this review with Sandra scores further down the page!

3DMark2000

VIA 4-in-1 v4.25	VIA 4-in-1 v4.26
3392	3392

So no change going from the 4.25's to the 4.26 set. Nothing to see here, move on.

Quake3 Arena (v1.17)

VIA 4-in-1 4.25 drivers
Resolution	r_smp 0 (average over 3 runs)	r_smp 1( average over 3 runs)
320x240	115.4	152.5
640x480	104.4	112.5
800x600	74.7	74.7

VIA 4-in-1 4.26 drivers
Resolution	r_smp 0 (average over 3 runs)	r_smp 1 (average over 3 runs)
320x240	114.5	152.4
640x480	104.0	112.5
800x600	74.7	74.7

There's not much to choose between the two driver sets. The 4.26's are slower across the board except in the 800x600 tests where the graphics card is holding the system back. Given that the TNT2 Ultra is not the best card to test with, we ran the tests at 320x240 to show you the effect when the graphics card isn't a bottle in the system and the CPU's are pushed. We would have used my ATI Radeon 64Mb if the Win2K drivers for the Radeon hadn't sucked so much!.

I'm regretting buying the card now due to poor Win2K driver support, but that's for a whole new review. Suffice to say, the Win2K performance of the Radeon wasn't that much greater than the TNT2 Ultra, and the Detonator drivers were far more stable than what ATI has to offer.

SMP definitely makes a difference, and with a high end graphics card (with good driver support in Win2K :) ), the dual processors should show their muscle and give a handy increase over a similarly clocked single CPU system.

SiSoft Sandra 2000

SiSoft Sandra gave us a few problems while testing with this board. Not with the actual running of the test, which all completed fine ,100% stable, but with the widely varying results it seemed to give us.

We first noticed it during tests to determine what effects memory tweaking would have on the system. The memory scores returned were lower than what we'd originally tested without tweaking the memory! Obviously not right, we investigated further and came up with the following, quite disturbing result. Sandra would return a higher memory benchmark score depending on whether we had previously run the drive benchmark beforehand.

If we didn't run the drive benchmark after a reboot and bios memory tweak, the returned score would be lower than our original memory benchmark scores, which we obtained after a drive benchmark. To obtain the scores you are about to see, we ran the benchmarks in the following order:

• CPU Benchmark
• CPU Multi-Media Benchmark
• Drive Benchmark
• Memory Benchmark

I can't reproduce this effect using my Abit KT7-RAID motherboard and AMD Duron processor. On that system, the scores returned are identical regardless of whether the drive benchmark was run. I don't know whether it's chipset, Sandra or VP6 specific, but it casts an ugly shadow over using Sandra as a benchmark. If this effect can be reproduced on other systems, using other chipsets or motherboards, many hundreds of Sandra benchmarks may be worthless! Try it on your own system and let us know the results.

Anyway, here are the results we obtained for the VP6 using the 4.25 drivers. First off, the CPU Benchmark:

It gets close to the dual Intel 1Ghz system, but falls short as you would expect due to it's lower CPU speed. Nothing unusual here. Next, the CPU Multimedia Benchmark:



Again, it's within striking distance of the dual 1Ghz setup. Notice how it demolishes the single 1Ghz Athlon which is to be expected. Next, the Drive Benchmark:



In our test setup, the drives were connected to the HPT370 controller on the motherboard. There was no RAID array setup using the disks and you would see a significantly higher score if that were the case. We briefly tested the onboard UDMA/100 support from the VIA chipset and found a similar score to the one shown. It does nothing except show that the UDMA/100 support of the HPT370 controller is working! Finally using this 4.25 driver set, the Memory Benchmark:



The score falls short of both dual 1Ghz Intel and AMD setups and way short of the monster dual Xeon setup using Rambus RDRAM. The score reflects no memory tweaks in the BIOS and running at 133Mhz.

Next, the 4.26 driver set. Again, firstly, the CPU Benchmark:



Here we see a very slight increase over the 4.25 results. I think we can attribute this to slight differences in testing conditions. I don't think we can attribute this to the driver. I may be wrong though! Next, the CPU Multimedia Benchmark:



Again, a very slight increase over the 4.25 results. Again, I think it's down to very slight differences is testing conditions. Still the score approaches the 1Ghz i820 based Intel setup, but falls short due to the clock speed difference. Maybe the VP6 would pull ahead running dual 1Ghz CPU's? Next, the Drive Benchmark:



This score is interesting due to the fact that the 4.26 set installs a new Busmastering disk driver under Windows 2000. VIAHardware.com are reporting that the driver *may* increase performance. However, our Sandra numbers don't support this. We see a 1% decrease in score between the 4.25's and the 4.26 set. While not earth shattering, it's by no means a significant increase. Finally the Memory Benchmark:



Again, a slight decrease in performance using the new drivers. Not that much (1.3% if I did my sums right!), but maybe enough to attribute it to the new driver and not slight differences present in the testing conditions?

Conclusions and Notes

Can we now answer the question we posed at the top of our review? Is performance bad enough for Abit to plead with sites not to review it yet? Nope, I think performance is adequate. How about this? During testing, while trying to tweak memory settings, we came across a significant brick wall. Whenever we tried to set the CAS memory latency to 2, from 3, Win2K would blue screen upon booting and warn us that the BIOS in the system was not ACPI compliant! The screen told us to check hardware-support.com for an updated BIOS and gave us a method of disabling ACPI support while installing Win2K.

Bear in mind that I have personally witnessed the very same 384Mb of Cube PC133 CAS2 memory run at CAS2 at 150Mhz FSB in an Asus P3V4X! Rest assured, the memory is the good stuff and would do CAS2 at 133Mhz without breaking a sweat. So maybe BIOS problems are halting the expected slew of reviews. Setting to CAS2 in the BIOS let the machine complete the POST test, but would report the wrong memory size every time. Sometimes bang on 128Mb, sometimes a weird value that didn't match up with any known memory configuration supported by the board. We even thought we'd killed the 256Mb DIMM at one point and I nearly sh*t myself!

So maybe the board is incompatible with some DIMM modules? Who knows? Whatever the reason, we hope Abit fix the problem soon. The board is a great overclocker. Every FSB speed imaginable is available, and both boards we have run our personal P3-700's at 933 no problems. We've even hit 980 with one pair (7x140Mhz) and the same pair POSTS at 1001 (7x143Mhz) but doesn't quite get into windows. With some tweaking, I'm sure both may hit the magic 1Ghz with 2 CPU's.

We're just happy to bring you one of the only reviews of the VP6 currently out there.