It would make it much easier to identify its size if you can take pictures of the cache module front and back sides or the link to the original sale.

Roger that-

how to upload an image to the internet

It reads on the sticker, 32k addresses times 64 bits = 2048 kilobits = 256 kilobytes.

So I think it's not 512KB cache stick but 256KB.

Jepael wrote:

It reads on the sticker, 32k addresses times 64 bits = 2048 kilobits = 256 kilobytes.

So I think it's not 512KB cache stick but 256KB.

As I feared. I was desperate for a compatible cache stick, and jumped on the item which was advertised as "Samsung KMM764V41AG2-15 512Kb Cache for Socket 7" - Guess I should have inspected the images and done some math first... but I guess 256k is better than nothing. Beware of Ebay user "andrey_temnyy" in that case.

As for my initial question- is enabling an L2 cache in a Win95/DOS environment liable to cause HIMEM address errors in general, or is it likely just my jumper being set to 512? I installed both OSs before the cache was installed or enabled. I will switch jumpers over and update with my findings- as soon as I ask for a 50% refund!

L2 cache shouldn't cause any errors unless it's faulty or configured incorrectly. Set the jumpers to 256k and see what happens

Could have been an honest mistake on the seller's part. I would message the seller and let them know what is up. You will probably get a good discount or possibly even a full refund.

keropi wrote:

L2 cache shouldn't cause any errors unless it's faulty or configured incorrectly. Set the jumpers to 256k and see what happens

To my regret and shame, I must report that I *already* had the jumpers set to 256kb to begin with. So now I'm not sure what the trouble is.

Are these modules write-back/write-through specific? At the moment I'm awaiting a new DALLAS clock/battery chip, since my external battery mod is not stable enough to reliably save CMOS settings, so I can't get far enough into boot to confirm it, but I think my BIOS is a fixed "Write-back" distribution.

cyclone3d wrote:

Could have been an honest mistake on the sellers part. I would message the seller and let them know what is up. You will probably get a good discount or possibly even a full refund.

I considered that in my message, but I've looked at his other items, and he sells a LOT of Socket 7 era motherboards and cards, with well written descriptions of the hardware involved, so he doesn't seem to have trouble with English or correctly sourcing hardware. I'll see how he responds.

(in case this helps, these are the configurables I have access to with this BIOS, and the usual settings I had before the CMOS started to go soft on me)

UPDATE: I put some new batteries in the external DALLAS mod and got stable CMOS control back. When I load DOS/95 with EXTERNAL CACHE ENABLED, the System Config flashes for a few moments and indicated "Pipeline Burst" for L2 cache type. However, it still fails in DOS/95 with a HIMEM error, always popping back a different memory address in the error report. With EXTERNAL CACHE DISABLED, the Sys Config readout before OS boot states "Write Back" for L2 cache type.

Maybe this stick is just bad? Yet sys config recognizes it for what it is, or at least , recognizes the setting in BIOS for what it is.

So I search th OP's past posts and found this one
Assuming the OP is talking about the same board here... The rightmost motherboard has onboard L2 chips and adding another 256KB coast will make it 512KB.
Another thing that could go wrong is the setting of tag ram. Some boards come with 8KB tag ram onboard with 256KB cache and the onboard tag must be disabled by jumper after installing coast.
The last possible culprit is overclocking... the coast can't work at any bus speed higher than 66MHz.

lazibayer wrote:

So I search th OP's past posts and found this one Assuming the OP is talking about the same board here... The rightmost motherbo […]
Show full quote

So I search th OP's past posts and found this one
Assuming the OP is talking about the same board here... The rightmost motherboard has onboard L2 chips and adding another 256KB coast will make it 512KB.
Another thing that could go wrong is the setting of tag ram. Some boards come with 8KB tag ram onboard with 256KB cache and the onboard tag must be disabled by jumper after installing coast.
The last possible culprit is overclocking... the coast can't work at any bus speed higher than 66MHz.

Well shucks... you went the extra mile and I thank you! I had no idea the system possessed an onboard L2 chip anywhere... I will set jumpers to 512 and try right now. For reference, though, I have cataloged every single jumper on this board, and I can find no jumper for tag ram enable/disable, just the 256/512 option, and the silkscreen doesn't mention anything like "No jumper - Tag ram off". Also, my FSB and Multiplers have been left in the default 66/2.5 positions. I have not attempted to OC at all.

UPDATE: No joy. Switched to 512, and I receive the same error. The board you mentioned in my pics looks EXACTLY like mine, down to every chip type and jumper location, so I dunno.

MEGA UPDATE:

I decided to perform a complete tabulated test of all possible configurations. There are two lines in System Configurations which loads after POST that discuss cache status on the machine- Cache Value (#), and L2 Cache type (type). The following are my results

WITH MODULE INSTALLED

Jumper = 512, L2 BIOS = Enabled = # "512 kb", type "Pipeline Burst", HIMEM ERROR

Jumper = 512, L2 BIOS = Disabled = # "W/B Cache", type "Write Back", HIMEM GOOD
-----------------
Jumper = 256, L2 BIOS = Enabled = # "512 kb", type "Pipeline Burst", HIMEM ERROR

Jumper = 256, L2 BIOS = Disabled = # "W/B Cache", type "Write Back", HIMEM GOOD

WITH MODULE REMOVED

Jumper = 256, L2 BIOS = Enabled = # "256 kb", type "Pipeline Burst", HIMEM GOOD!!

Jumper = 256, L2 BIOS = Disabled = # "W/B Cache", type "Write Back", HIMEM GOOD

I didn't continue testing with the module removed and jumpers at 512 because it seems clear to me from the first round with module in that the system's "Cache Value" number is responding to the total amount installed, independent of the jumper setting.

OBSERVATIONS:

I do indeed have 256kb of onboard pipeline burst cache on this mobo, which is pleasing news.
I seem to have a bad COAST module, despite sys configs recognizing it in the Cache Value entry as adding to a total of 512 in system.

QUESTIONS:

I am curious as to what would happen in one added 512kb via a COAST module to this system- clearly the jumper implies a 512 *could* be added- but I thought the VX chipset only allowed for a maximum of 512kb of L2 cache anyway...so what's the purpose of allowing a 512 stick and 512 jumper setting? Or am I mistaken, and the VX chipset is capable of addressing more than 512kb of L2 cache?

Since I have no intention of going through the riff-raff of shipping this thing back to Kazakhstan, does anybody have any socket 7 mobos lying around who want this module to plug in for testing and evaluation?

Is this your motherboard?
http://motherboards.mbarron.net/models/pcchips/m525.htm

gdjacobs wrote:

Is this your motherboard?
http://motherboards.mbarron.net/models/pcchips/m525.htm

Wow, I'm speechless...spent 2 weeks trying to find any reference to this board, which isn't branded. And now this wealth of data! Thank you sir.

According to this manual, if the coast module isn't specced for INTEL's COAST 3.0 standard or higher, it won't work. I can't tell if my coast module (a Samsung KMM764V41AG2) is is a 3.0 or later, or not. Oh well- Looks like I should set my jumper to 256, accept the onboard Pipeline Burst/Tag Ram for what it is, and move on to fixing my Sound Blaster issues.

Thanks everyone!

35mmshowdown wrote:

Wow, I'm speechless...spent 2 weeks trying to find any reference to this board, which isn't branded. And now this wealth of data […]
Show full quote

gdjacobs wrote:

Is this your motherboard?
http://motherboards.mbarron.net/models/pcchips/m525.htm

Wow, I'm speechless...spent 2 weeks trying to find any reference to this board, which isn't branded. And now this wealth of data! Thank you sir.

According to this manual, if the coast module isn't specced for INTEL's COAST 3.0 standard or higher, it won't work. I can't tell if my coast module (a Samsung KMM764V41AG2) is is a 3.0 or later, or not. Oh well- Looks like I should set my jumper to 256, accept the onboard Pipeline Burst/Tag Ram for what it is, and move on to fixing my Sound Blaster issues.

Thanks everyone!

The coast module is a 3.1 according to the datasheet.

Its a PC Chips board? I'm not surprised at all by your troubles. I wouldn't be so quick to blaim the COAST module. Did PC Chips ever make fake pipeline burst cache?

feipoa wrote:

Its a PC Chips board? I'm not surprised at all by your troubles. I wouldn't be so quick to blaim the COAST module. Did PC Chips ever make fake pipeline burst cache?

PC Chips and fake components? Was water ever wet?

gdjacobs wrote:

feipoa wrote:

Its a PC Chips board? I'm not surprised at all by your troubles. I wouldn't be so quick to blaim the COAST module. Did PC Chips ever make fake pipeline burst cache?

PC Chips and fake components? Was water ever wet?

Hah, it's true, the board could just be a sleeping giant of awful- but I will say, besides this cache trouble, it's been pretty solid otherwise. I checked the data sheet on the COAST stick I bought, and in my not-at-all-expert opinion (based on the latency information) it seems like it's a more advanced low-latency stick designed for advanced server type boards- and thus incompatible with the VX/Intel 3.0 standards.

I know I've read it about here on the boards before, but isn't there some benchmark application that can test and confirm L2 cache performance and functionality, on top of other performance parameters?

Speedsys checks memory performance across the various step sizes, so you can see were caches are in effect.
http://forum.benchmarkreviews.com/showthread.php?p=15023

I find CPUMark99 the most sensitive to L2 cache and if all RAM is cached.