retro games 100 wrote:

a) how do I determine the chipset revision number, for my GA-5AX mobo? Both of the Alladin5 chipset chips do not have heatsinks on them. One is marked A1, and the other is marked B1.

I think it's the last character in the second-to-last line (above "TAIWAN"). Somebody please correct me if I'm wrong...
There might also be ways to find out the revision via software.

retro games 100 wrote:

b) how did you obtain the results output from CTCM17A, in order to produce your "code text block" for your Vogons post?

I redirected the output into a text file (e.g.: ctcm17a > ctcm.log). Then it was simply a matter of getting the file to my linux box, where I could copy&paste the output into the browser window.

retro games 100 wrote:

The ebay price on an Asus P5A rev. 1.06 has risen by about 10 euros! 😉 🤣

At least. 🤑 This board was quite dear to begin with. I remember paying 30 € for the board and a vanilla K6-III several years ago, when ebay was swamped with ALI SS7 boards.

And yeah, the L3 cache is really insignificant (maybe it has more impact on MVP3 boards). If you want to go for speed, turn it off and raise the FSB by 20 MHz.

^ Thanks a lot! BTW, it looks like the rev. num. on my GA-5AX mobo is H.

Well, seems my notes were not so wrong after all. Seems that with the right chipset and mobo, the Ali5 can cache up to 1GB of memory. The table would now look like this, thanks to 5u3 results:

----------------------------------
ALi V

256 kB 2nd-LC - 256 MB
512 kB 2nd-LC - 512 MB
1024 kB 2nd-LC - 1024 MB

This chipset can also use the external TAG-Ram additional to the internal to extend the area even furthermore (needing more cache memory too), but with this setup, the chipset cannot run 100 MHz anymore.
On early revisions (at least up to Rev. E) the internal TAG-RAM did not work correct. It was disabled and an external TAG-RAM was used instead (at full speed).
-------------------------------------

^ Do you have any problems running CPUs with 100Mhz FSB ? Maybe your mobo falls in the upper category ?

Also remember that some boards happily would run the FSB overclocked AND the L3 cache enabled. You'll have a MINIMAL performance increase by having it enabled, but, oh well, it's something.
Being able to have the L3 enabled when overclocking the FSB depends more on the TAG-RAM chip than on the cache chip itself.

Old Thrashbarg wrote:

It depends on the chipset, but yes, Aladdin V boards will only cache 128MB RAM with 512K L2. And yes, performance will suck if you install more than 128MB, especially since Windows tends to use the highest memory addresses first.

I would really like to test a "malfunctioning cache Ali 5 board", against a non-malfunctioning cache Ali 5 board. My GA-5AX with 256MB of RAM appears to only cache 128MB. OTOH, 5u3's Asus P5A rev. 1.06 appears to cache all 768 MB of RAM. If either 5u3 or any P5A 1.06 owner is interested, what benchmarks can be run inside Windows 98, to determine each board's performance? Both boards would need to be configured so that they are as similar as possible:

* Same amount of RAM. Is 1 stick of 256MB a good idea?
* Same video card. Would a Radeon 9800 Pro be OK? Mine is 128MB. I have other video cards, including some PCI cards.
* Generally speaking, same BIOS settings. Would "best performance" settings be OK?

Obviously people are busy with other projects, life and stuff! 😉 If this "concurrent test" is of interest to anyone, I'm willing to give it a go.

For such a benchy to be consistent, we'd need to be sure we use identical motherboard settings. It would be preferable to use, for instance a P5A 1.04 and P5A 1.06. I only have the 1.04's but for me the problem is, my attic is a mess right now as I'm busy in my house decorating and my attic's the temporary storage area right now

Tetrium, you're right. A test involving a P5A 1.04 and P5A 1.06 would be preferable.

Last time I checked, all my ali boards come with the bugged version of the chipset. Actually, I should go and check to make sure. Anyhow I'm pretty sure I don't have 2 different chipset revisions of the same board anyway.

Edit:Also it would be preferable to have identical chips, or atleast chips with the same core (not 1 CTX and 1 non-CTX chip) and identical bios settings and the same amount of memory with the same cas latency

Has anyone ever experience issues with CF drives on these Super Socket boards? ALI 5 and VIA chipsets?

Both my boards give me issues, though the Slot 1 board has no issues whatsoever.

Issues are that some games take very long to load. Like pause, pause pause, load, pause, pause, pause load and I sometimes get DOS read errors (retry, abort) messages...

It's not a deal breaker, but I might switch back to my Slot 1 system which was super stable and never had any drive issues whatsoever 😀

Mau1wurf1977 wrote:

Has anyone ever experience issues with CF drives on these Super Socket boards? ALI 5 and VIA chipsets? […]
Show full quote

Has anyone ever experience issues with CF drives on these Super Socket boards? ALI 5 and VIA chipsets?

Both my boards give me issues, though the Slot 1 board has no issues whatsoever.

Issues are that some games take very long to load. Like pause, pause pause, load, pause, pause, pause load and I sometimes get DOS read errors (retry, abort) messages...

It's not a deal breaker, but I might switch back to my Slot 1 system which was super stable and never had any drive issues whatsoever 😀

Are both your boards VIA boards?
Anyway it seems SS7 boards always have issues with drivers. Have you tried updating to a newer (unofficial) bios or tried other drivers?

I put the CF card into the secondary controller and haven't had a read error yet. So maybe that was it. Will keep an eye on it!

Got the latest bios, but no drivers. It's DOS 6.22 and using FAT16. No ontrack or stacker or tools like that.

You can try lowering the IDE transfer mode in BIOS too

h-a-l-9000 wrote:

You can try lowering the IDE transfer mode in BIOS too

Coolies I will keep that in mind.

So far so good with the secondary controller. Maybe the first one was busted?

I'm a VIA hater and thus my opinion here is a little biased 😉

h-a-l-9000 wrote:

I'm a VIA hater and thus my opinion here is a little biased 😉

That's cool. I normally insist on Intel, but many options.

The chipset on this board is ALI V. But yea so far not a single read error. I think that was it!

Old Thrashbarg wrote:

It depends on the chipset, but yes, Aladdin V boards will only cache 128MB RAM with 512K L2. And yes, performance will suck if you install more than 128MB, especially since Windows tends to use the highest memory addresses first.

I decided to test my bugged L2 cache Ali V Gigabyte GA-5AX mobo. I put an AGP Voodoo5 in it, and used a Pentium 233 MMX CPU. I ran 3DMark 99 Max. Test #1 used 1 stick of 256MB CL2 SDRAM. Test #2 used 1 stick of 128MB CL2 SDRAM.

#1 927 (3D), 1575 (CPU)
#2 1053 (3D), 1649 (CPU)

I noticed that these scores changed a bit, as I reran them. The scores above aren't averages, I just picked the last ones I ran. OK, sorry this isn't the best way to do things, but the values above do show that the 256MB test ran slower than the 128MB test. Pity. Also, the scores are fairly bad anyway, for both tests. I wonder, can someone please run 3DMark 99 Max using a SS7 mobo, with an AGP Voodoo5 and a P233 MMX CPU? Thanks a lot if possible.