Page 1 of 2

Wpcredit and the ABIT VA6

Posted: Thu Oct 19, 2006 6:00 am
by Stedman5040
The VA6 board features the Via apollo pro chipset and has a slot 1 that can use fcpga pentium cpuz with an adapter.

Most reviews damn the board as having poor memory bandwidth.

Anyway I got hold of one of these boards and a slocket with a celeron800 cdo stepping.

As booted up with no wpcredit tweaks the memory testing results on the board at 100fsb are dismal.

Celeron fcpga 800 @8x100
3x Hynix sdram 256mb pc133 cl2 memory (2-3-3-6)

Memory read 248
Memory write 112
Memory latency 165

With wpcredit tweaks and running the celeron at 1066mhz there is quite an improvement.

Celeron fcpga 800 @ 1066 (8x133)
3x Hynix 256mb cl2 memory (2-2-2-6)

Memory read 1004
Memory write 206
Memory read 99

Not too bad for an fcpga celeron with the tweaks and there maybe some room for a little bit more improvement.

Stedman

RE: Wpcredit and the ABIT VA6

Posted: Sun Oct 22, 2006 7:55 am
by Jim
@ Stedman : More details please. Is this board an AT or ATX? (I suppose I could find that out for myself). What WPCredit tweaks did you use?

I just found a "baby" AT board, (which I will be picking up later today when the store opens), which has :

An Intel 550 P3 Slot Processor w/ 512k cache, & 100Mhz FSB. (Comes w/ it).
Via Chipset w/ VT82C691 Northbridge & VT82C596 Southbridge
3 Dimm Slots
1 AGP Slot
3 PCI Slots
FSB settings for 66.6 Mhz, 75Mhz, 83Mhz 100Mhz, & 103Mhz.
EDIT : Bit of research reveals that there are 2 undocumented FSB settings that actually do work : 112Mhz & 133Mhz.
Multiplier Settings for 3, 3.5, 4, 4.5, 5, & 5.5

Not sure about voltage settings, or Make / Model of board; but it is cheap, and if CPU Cool works w/ it, it could concievably get an AT up to 733Mhz.

EDIT : Board is an FIC KA-6100. It has 4 undocumented FSB settings which will result in who knows what; and 10 undocumented Multiplier settings which will also result in who knows what. There do not seem to be any voltage settings, (instead it has an autosensing routine which can supply voltages ranging from 2v to 3.2v), which is likely to restrict overclocking to a very limited range. There are however settings for clocking the RAM at various speeds separately from the processor. Also, in an oddity that I have not seen before, there is a jumper to set the size of the Bios ROM; and another one to set the "Flash ROM Voltage".

Apart from that the board has "Onboard Sound", which I may wind up using if I am short a PCI slot; and also has 4 Onboard USB Ports i.e. 2x2. All in all not too shabby for an AT.

Posted: Sun Oct 22, 2006 5:06 pm
by Uranium235
Stedman, I would also be interested in knowing what tweaks you used to get those results. I also have a Socket 370 board, an Epox EP-3VWB+. It's the fastest AT board I could find for an old AT case but performance lags behind other 370 boards.
Here's a link:
http://www.epox.com/usa/product.asp?id=EP-3VWBplus

Posted: Sun Oct 22, 2006 7:48 pm
by Jim
Had to download the manual to find out what that Epox is all about. VERY NICE!! Oughta blow the doors off my FIC; but how are you ever going to get the "Display" to show upwards of 1 Ghz, (8 x 133Mhz)? I gather you are another affectionado of ATs; - most people would just pitch the case and get another one.

Posted: Sun Oct 22, 2006 10:31 pm
by Uranium235
I do like AT's. I've got two AT cases. The first AT case, a very nice Aopen mid-tower, which I traded someone a FIC PA-2013 for, houses the EP-3VWB+. The second case, a generic mid-tower with plastic standoffs, a leftover from an upgrade, houses a Tyan S1590. I use ATX powersupplies in both and installed momentary contact switches for the power switches.
http://www.pccables.com/03201.htm

I'm not sure how the 3VWB+ will display anything over 1Ghz but I have seen reports that the board will run a 1.4Ghz Tualatin with a socket 370 adapter.

The FIC KA-6100 looks like it has possibilities. Might take a Tualatin with a slotket.

I'm just an affectionado of old hardware in general. :)

Posted: Sun Oct 22, 2006 11:56 pm
by Jim
I ran across 1 post that said the FIC would take a Tualatin; but I didn't know whether to believe it. Seems the limiting factor will be the 5.5 multiplier.

EDIT : Have also found this link which indicates that there is a 7x multiplier on the FIC KA-6100.

http://www.kmscomponents.com/htm/suppor ... umpers.htm

What I find odd about it is the multiplier jumper settings are same for 5x, 5.5x, 6x, 6.5x, & 7x. I also found another page that mentions a 7.5x setting. Seems that what you are setting w/ the jumpers is just some sort of ballpark setting; and the board then "autosenses" the processor to determine what the multiplier should be. Makes it tricky to guess what setting to use for newer processors that weren't around when it was designed. e.g. Tualatins.

any thoughts on this?

@Stedman : Sorry if this is too far off topic. Will move it if you'd prefer.

Posted: Mon Oct 23, 2006 3:49 pm
by Stedman5040
The ABIT VA6 is ATX and has as stated before the VIA Apollo PRO133 chipset. Not the 133A chipset so it only supports 2x AGP. Anyway here are some figures with regards to the memory bandwidth.

Set up as Celeron 800 (8x100) with 3x hynix pc133 cl2 256mb SDRAM

The bios was set up with memory timings on turbo and CAS latency on 2. You would think that would give you good memory timings, but it doesn't, it sets up as 2-3-3-5. Anyway under these conditions the Everest scores are as follows

Memory Read 237.
Memory write 101.
Memory latency 161.

Sandra2004 bandwidth gives 243/243

With the following tweaks by wpcredit the memory timings are 2-2-2-5. The cpu is kept at 800

Register 50 set to FF
Register 51 set to FF
Registers 64,65,66 are the memory timings for the 3 sticks and are set at 12
Register 70 set to 4F

With these tweaks we get on Everest

Memory Read 760.
Memory write 160.
Memory latency 134.

Sandra memory bench increases to 703/686 from 243/243 !!

Taking these tweaks and powering up the fsb to 133 and increasing the cpu speed to 1066 Everest now gives

Memory read 1017.
Memory write 214.
Memory latency 108.

Sandra memory bandwidth increases to 945/846.

I am sure that there will be other tweaks in wpcredit, but I wouldn't expect to see any more massive performance gains.

So going from no tweaks and 800MHz the memory bandwidth has come from being a super socket 7 performer to almost a KT133 chipset performer at 1066 and 133fsb.

Everest memory read scores are up by over 4x and writes by more than 2x. Sandra scores are up by 3.5 to 4x.

Stedman

Posted: Mon Oct 23, 2006 11:22 pm
by DasMan2
That's great stuff , can you equate to real life Game FPS performance increase scaling so others can preceive how much it would increase that area of computer use .... if any ?

I hope it does , instead of getting game crashes .. :oops:

Posted: Tue Oct 24, 2006 5:23 am
by KachiWachi
You should mention the version of Everest you are using, since data taken in one version is not comparable to data taken in another.

This *might* apply to other test software as well.

Thanks.

Posted: Tue Oct 24, 2006 11:18 am
by Stedman5040
Benchmarks with Everest build v2.20.405 and Sandra build 2004.10.9.133.

Stedman

Posted: Wed Oct 25, 2006 9:26 am
by Uranium235
Stedman- Thanks for the info. That was a serious performance gain. :)

Jim- This is probably what you need:
http://cgi.ebay.com/UpgradeWare-Slot-T- ... dZViewItem

AFAIK the multiplier is CPU dependant and done internally, unless the CPU is an engineering sample. The board settings will have no effect. The main concern for Tualatin support is for the voltage @1.45v which the slotket provides for. A 100FSB, 1400 Celeron might be your best bet. The BIOS may not recognize it though but BIOS Patcher should correct that.

Posted: Wed Oct 25, 2006 8:03 pm
by Jim
@ Uranium235. Thanks, I already knew about the "Slot T" adapters, actually for another $3.00 you can still get them direct from Upgradeware. What I was not sure about is whether to get a "1.4 Ghz Pentium 3 S Tualatin" or a "1.3 Ghz Celeron". I have found website info that indicates that the FIC KA-6100 will support a 1.3 Ghz Celeron, but nothing about the 1.4 Ghz Pentium 3 S, or the Celeron 1.4 Ghz, (which I had not previously heard of, - keep in mind if it isn't AT compatable I don't pay any attention to it since I will probably never build an ATX.)

What I do want to do, is build the best possible AT machines that I can . So what I will probably wind up doing is getting both the P3S and a Celeron; and if the P3S works, use it and dump the Celeron. If the P3S doesn't work, I'll use the Celeron; and hang onto the P3S until I can get a board that will support the P3S.

As for the bios patcher, I have it; but have never used it. Probably need help to use it without screwing up the board beyond redemption.

Posted: Thu Oct 26, 2006 10:40 pm
by Uranium235
Jim, I ran BIOS Patcher on the KA-6100 BIOS version 610011D.BIN to see what it would do. Here's the output:

BIOS Patcher ver. 4.23.
Attention! Advanced qualification is required!

Found 1Mbit BIOS!
Remove "otkat" (backup key)

1. New CPU Support : -> fixed.
2. P3-detect error : not found.
3. New Koeffs Support : -> fixed.
4. 32Gb-problem : -> fixed.
5. Some HDD detect-problem : -> fixed.
6. "MB"/"GB" string search : -> fixed.
7. 65Gb-problem (1-st step) : -> fixed.
8. 65Gb-problem (2-nd step) : -> fixed.
9. Error display Freq>999MHz : -> fixed.
10.Error display Koefs>9.5x : not found.
11.New Stepping Support : -> fixed.
12.Tualatin L2-init error : -> fixed.
13.New Freq in Setup open : not found.

14.Set "Y" as default on exit: -> fixed.

UDMA for "big"-HDD on UDMA33_only_MB fix: -> fixed.

CBROM V2.07 (C)Award Software 2000 All Rights Reserved.
[CPUCODE] ROM is releaseCBROM V2.07 (C)Award Software 2000 All Rights Reserved.
Adding modul.tmp 35.8%
CBROM V2.07 (C)Award Software 2000 All Rights Reserved.
Adding start.tmp 87.5%
CBROM V2.07 (C)Award Software 2000 All Rights Reserved.
Adding cpucode.bin 53.6%

if you can`t see all messages - choose 80x50 mode or run with ">report.txt".
(c)2002-2003 apple_rom, www.ROM.by


It would appear that the BIOS would need to be patched for Tualatin support and large hardrive support. BIOS Patcher only fixes what the BIOS lacks. It looks to me like you can make the Tualatin work. I can send you the patched BIOS I made if you need it.

Posted: Fri Oct 27, 2006 6:30 am
by Jim
@ Uranium235 : What you have just posted is entirely Greek to me. I fail to see how you can tell the bios needs to be patched for Tualatin support and large HDD support from what you have posted above; though you are definitely right about the HDD at least. But the bios you used is not the latest; I have an "e" version, and also a patched version that does give support for large HDDs supposedly.

Also, the board offers the choice of setting the Bios ROM size as either 1Meg or 2Meg. (Weird) Mine came set at 1Meg; and I have not touched the jumper, though I have encountered something that would seem to indicate that I have the 2Meg chip. It also allows you to set the flash voltage at either 5v. or 12v. More Weirdness that I have not touched. Get that sort of thing wrong and you "Colonel Saunders" your mobo. Yeah, I'll definitely appreciate any help you can give me w/ this one. Thanks.

@ Stedman : Sincere appologies for outright thread piracy. Maybe when it is up and running I'll be able to post some comparison numbers to go with those you posted.

Posted: Fri Oct 27, 2006 9:52 am
by Uranium235
OK, I ran the BIOS patcher on the newer 610011E.bin BIOS. Results were nearly the same.

BIOS Patcher ver. 4.23.
Attention! Advanced qualification is required!

Found 1Mbit BIOS!
Remove "otkat" (backup key)

1. New CPU Support : -> fixed.
2. P3-detect error : not found.
3. New Koeffs Support : -> fixed.
4. 32Gb-problem : -> fixed.
5. Some HDD detect-problem : -> fixed.
6. "MB"/"GB" string search : -> fixed.
7. 65Gb-problem (1-st step) : -> fixed.
8. 65Gb-problem (2-nd step) : -> fixed.
9. Error display Freq>999MHz : -> fixed.
10.Error display Koefs>9.5x : not found.
11.New Stepping Support : -> fixed.
12.Tualatin L2-init error : -> fixed.
13.New Freq in Setup open : not found.

14.Set "Y" as default on exit: -> fixed.

UDMA for "big"-HDD on UDMA33_only_MB fix: not found.

CBROM V2.07 (C)Award Software 2000 All Rights Reserved.
[CPUCODE] ROM is releaseCBROM V2.07 (C)Award Software 2000 All Rights Reserved.
Adding modul.tmp 35.4%
CBROM V2.07 (C)Award Software 2000 All Rights Reserved.
Adding start.tmp 87.5%
CBROM V2.07 (C)Award Software 2000 All Rights Reserved.
Adding cpucode.bin 53.6%

if you can`t see all messages - choose 80x50 mode or run with ">report.txt".
(c)2002-2003 apple_rom, www.ROM.by


Here's how to interpret the results:
4. 32Gb-problem : -> fixed.
5. Some HDD detect-problem : -> fixed.
6. "MB"/"GB" string search : -> fixed.
7. 65Gb-problem (1-st step) : -> fixed.
8. 65Gb-problem (2-nd step) : -> fixed.

These lines indicate the BIOS doesn't recognize hardrives over 32GB. The patcher corrected this by indicating-"fixed".
9. Error display Freq>999MHz : -> fixed.
This line indicates the BIOS had a problem displaying CPU's clocked at 1Ghz or over and "fixed" this problem.
1. New CPU Support : -> fixed.
11.New Stepping Support : -> fixed.
12.Tualatin L2-init error : -> fixed.

These lines show new support was added to the BIOS for newer CPU recognition and corrected a problem with the Tualatin's L2 cache operation.

I don't believe a Tualatin CPU would operate properly on your motherboard without the BIOS patch. It would either run very slow or not at all IMHO. There is also a very real chance the BIOS could be FUBAR'd by this patch but I doubt it would physically fry the board. Probably the safest way to go about it is to have a second BIOS chip with the patched BIOS on it so that if it didn't work you could just swap back the old BIOS chip and clear CMOS.
I did see that there is a patched BIOS for large hardrives available:
http://wims.rainbow-software.org/index.php?count=-1

Stedman- I'm going to try some tweaks on my S370 board and post some results. Thanks.