New Thunderbird System Produces Lackluster SETI Performance-Advice Requested

[paulcs]

I just put together a new Thunderbird-based system today and just got the result for it's first completed WU: over eight hours.

Now, I've unfamiliar with Athlon platforms and haven't used an Abit motherboard since my old BH6/Celeron 300A days. I pretty much know my way around Asus and AOpen BIOS's, but these new Abit BIOS's are a bit of a mystery to me. I suspect I'm missing something.

Here are some specs:

Athlon Thunderbird 700 MHz (100 MHz CPU Bus)
Abit KA7-100
PC133 RAM (Timings: 2-Turbo, whatever that means, Memory Bus: 133 MHz [HCLK+PCICLK])
Soundblaster Live
IBM Deskstar 75GXP DTLA-307015 HD (on the standard IDE interface)
Linksys 10/1000 NIC
Windows 98se

I haven't really installed much in terms of software, so there isn't a whole lot running in the background.

I'm using NT Commandline 2.0.

I can complete a WU on a PIII 600 MHz Katmai-based system in about six and a half hours, and my overclocked Coppermine-based systems are almost always between five and five and a half hours.

I'm sure I screwed up somewhere. Any help would be appreciated.

Thanks.

Paul
paulcs@flashcom.net

[Sir Hitech]

My Athlon Classic 700mhz on a MSI K7pro 100mhz cas2 mem does avg. 6hr 25min with nti386 or nti486 client. My Athlon Classic 1026mhz(9.5X108FSB) 144mhz/Turbo/CAS2 mem 1/3cache on a Abit KA7 doing 5hr15min-30min/WU.

Athlon T-Birds has 256k on-die L2 cache running at full speed. Classic version has 512k off-die cache running at 1/2,2/5,1/3 speed. Since Seti works better with larger cache(less swap), a Classic would be faster at the same speed cpu even it has slower cache. But I could be wrong.

Even I think a T-Bird is slower vs Classic in Seti, 8 hr/WU doesn't seem right.
[*]Use RY bios which optimizes mem better and enable cache speed change. [*]increase FSB to the max(overclock )(note: if you are at 112 you are running 148mhz mem with HostCLK+PCI. you are hitting the ceiling of your memory limitation. somebody explain the formula here)[*]disable video shadow/mask and bios shadow/mask[*]try Turbo/4way interleave/fast r/w turnaround/CAS2 with your mem setting[*]Download and Use the latest Via KX133 drivers. I will send you the files if you can't find them.

Go for it!

[This message has been edited by Sir Hitech (edited 01 July 2000).]

[paulcs]

Thanks.

I understand the formular. My CPU is now a 100 MHz. (I'm probably the only one. ) Clock + PCI Clock. I'll start overclocking as soon as I update the BIOS.

I haven't been able to access Abit's FTP sites. They're having some problems with their gateway. I do have that RAID capable BIOS Andy Drake (BXBoards) used.

I installed the latest VIA drivers right from the start (4in1 4.22, AGP 4.03).

I'm going to update the BIOS and enable 4 way interleave this morning. I learned about this feature at 4:00 AM this morning.

What's fast r/w turnaround? (Read/write, I'm sure; I'm looking for a technical explanation if possible.)

Sir Hitech, if you could email me the RY bin file, I'd appreciate it. Thanks.

Paul
paulcs@flashcom.net

[MegaManX4]

Im sorry Paul, but I think i've bad news for you. With this motherboard you'll probably never get good Seti results. My brother has the same combo except for the ATA 100. He bought his K7A for Thunderbird compatibility, but he has a copper Thunderbird from Dresden. It never runs stable in any way, due to chipset limitation. Even if it runs stable long enough to get a result crunched, it has poor performance. The same Processor in my Board(K7M with AMD Irongate Chipset) which is compatible with Thunderbird, only need somewhat about 6 hours. Clocked at 700Mhz of course.
The problem lies, as all of you probably know, In the way the chipset comunicates with this shiny Thunderbird. The Thunderbird needs something like 210ns to send and respond signals to the Chipset. The KX133 is only capable to send/recieve signals with 250ns. The Irongate AMD Chipset only needs 200ns. This lies within the specs. I read the whole explanation somewhere at the heise.de labyrinth, but can't remember the site.
So I think, your lost with this very fine board. It is really sad that it doesn't work well with the thunderbird.

Mega

PS: Paul, can you please make a little test for me? Could you start seti CLI and then start a mp3 file? If it is choppy, than my guess is right. Please don't use Winamp but the Media Player. Thanks.


[This message has been edited by MegaManX4 (edited 01 July 2000).]

[Sir Hitech]

Unfortunly, I have to agree with Mega. You are throwing darts when you dealing with Slot A T-Birds and KX-133 boards. That's why instead opt for a T-Bird I went for a Classic 850 for the machine I am building for my friend. If you want T-Bird, SocketA is the way to go. It sucks that people who have a perfectly fine KX-133 board like Abit KA7 are screwed when comes to upgrading.

Anyway, Paul:

Here's the bios for KA7

<A HREF="http://jabba.tlh.fdt.net/~ymc/KA7Bios/">http://jabba.tlh.fdt.net/~ymc/KA7Bios/</A>



[This message has been edited by Sir Hitech (edited 01 July 2000).]

[Assimilator1]

MM X4

Your missing out a major point here ,to allow the T bird to run properly on the Abit KA7 board you need to adjust the pullup & pulldown strength in the bios.
This should sort it out (sorry dont know settings ,but its on there website)

------------------
I guess I won't recruit for my team here TA

[paulcs]

Thanks guys.

MegaMan, I have the Austin Thunderbird (sounds like a country & western band). I did what you asked. No choppiness, but it did get noisy when running the SETI client.

Hmmmm. Maybe there's a K7M in my future.

I'm using some very aggressive settings now, which gave my Sandra scores a nice shot in the arm. They are very high. 3DMark2000 scores remain a bit iffy, despite upgrading to Ver. 1.1. Overclocking might help a bit there.

I think I'll skip the KT133 chipset altogether and hold off on a Socket A Thunderbird until AMD releases their next chipset (which, apparently, is named after a beer!).

Paul
paulcs@flashcom.net

[paulcs]

Thanks for the link, Sir Hitech.

I've adjusted CPU Pull Up Strength to 2, CPU Drive Strength to 4, as per Dan's Data's suggestion.

I have corresponded with Abit about this issue. Here's what they said:

"For T-bird to boot up with KA7-100, you'd better set "CPU Pull Up Strength" to 0 & "CPU Drive Strength" to 1."

I've tried these settings with an older BIOS and less aggressive memory settings, and saw no difference when compared to Dan's suggestion. I'll experiment a bit later with Abit's settings again and see if it makes a difference with SETI.

Paul
paulcs@flashcom.net

[This message has been edited by paulcs (edited 01 July 2000).]

[MegaManX4]

Hey Paul,

Which FSB are you running?

Mega

[Assimilator1]

Paulcs

Other than the SETI time problems ,how is the T bird running on your KA7 ? .Is it satble?

------------------
I guess I won't recruit for my team here TA

[paulcs]

I'm running at spec with the memory bus at 133 MHz.

Assimilator, it's been solid as a rock with one exception. After multiple runs of Winstone 99, the benchmarking program began crashing. I'm not sure if you have experience with Winstone 99, but the results weren't pretty. I suspect it's a software issue, however. No problems with Content Creation Winstone 2000 so far.

I've intentionally tried to stress the system since then, and I haven't had any problems, and again, I'm using some very aggressive settings.

Realworld performance is a big concern, however. It's funny, because the machine seems to boot faster than any computer I've used in a long time. Synthetic benchmark results (Sandra) are very good as well.

Paul
paulcs@flashcom.net

[MegaManX4]

After some real testing my brother found his Thunderbird slower than an Athlon Classic in Seti, both at 700Mhz. The classic needs somewhat of 6hrs 40min, the Thunderbird needs 7hrs 40 min. It's a difference of a hour in the wrong direction! It seems that we see the same performance difference like Katmai to Coppermine. We need big on die caches!

Mega

[Assimilator1]

MMX4

That's in SETI ,what about general useage?

We pretty muched guessed that the T bird would slower than the Athlon in SETI ,at least at lower clock speeds ,you bro's PC seems to confirm that.
Like you said same prob as Katami vs Coppermine ,down to the cache size.

Paulcs

A good program I find for testing for stabilty is Unreal (I guess UT would be as good) ,leave it looping bot deathmatches ,with maximium number of bots for at least 3hrs.If it's not 100% stable (or as near as is realistic) it will kick you out of the game to the desktop usually ,or just freeze up if its a heat problem

------------------
I guess I won't recruit for my team here TA

[Swing]

For those of you who have talked about stressing your computer to check for stability, looping 3DMark2000 does a good job. It will crash if the system is not stable.

[MegaManX4]

Assim,

Everything except SETI is faster with the Thunderbird. It's only a litte sad that it is so.

Mega

PS: Stability is probably best checked with prime95.