4x00 Unit CPUs

[Phoenix32]

Okay, so I started a new topic because I could not seem to find the old one...


Alright, so what was the end results on the swapping of the CPU in a 4x00 unit for a faster CPU? Did it help it any? Did things speed up? I do not recall ever hearing if it helped (any benchmarks).

The 4000 uses a Pentium 233 CPU. RAID 5 can be demanding (during writes) on the CPU if you are doing mass xfer and so can the 100baseT.

For example, on my 4000 with the 4 x 250 Seagate drives and 256 megs of CL2 SDRAM, I get an average of 25MB transfer speed while writing (on a gigabit network with smart switches, so no bottlenecks there). This means about 75 minutes or so to xfer 15 Gigs to the SNAP.

Another example. On the same SNAP 4000, a format of those drives takes a little over an hour, but the RAID 5 parity build (where it says it is building the backup drive) takes about 17 hours 45 minutes.

While there is a limit on how fast you can transfer/build due to drive speed and the 100baseT, I would bet the CPU at 233Mhz is a limiting factor here (heavy XOR parity creation load). Now other than the mobile processors, the fastest Intel Socket 7 CPU was the 133Mhz. However, there are AMD K6-2 processors with faster speeds. I do not know if the 4000 main board would support a 100Mhz FSB, it might, but there are K6-2 models still with 333 and so forth speed on a 66Mhz FSB. It "MIGHT" make things a little faster on a SNAP 4000.

I do not remember what the CPU is on the 4100, but I bet it could be speeded up also.

Any ideas? Comments? Examples? Benchmarks?

Come on people let's not just let this forum be a place where people pass through to get some quick fix or plea for an OS update or whatever. Let's see if we can learn some new things? Maybe even improve on the units we have and can afford. Who's with me?

(Jontz, here would be the place for you to tell me what CPU the 4100 has in it)

[re3dyb0y]

My Snap Server 2200 had nearly exactly the same chipset (or was it the same...?) as my old Intel PII 233mhz i had about. The CPU was the same speed, just socket, rather than slot on the desktop

[dgoodrich]

I don't suppose that the Snap 2000 would benefit from a CPU upgrade due to the lack of gigabit ethernet, and the slow drives? I happen to have a K6-2 CPU lying around is why I am asking.

[re3dyb0y]

It would have to be the same pins and stuff....

[Phoenix32]

Most, if not all, I can't remember back that far these days, of the K6-2 CPUs were of the same pinouts (socket 7 and super 7). Those questions are exactly what this thread is for. Do some benchmarks, put the CPU in, do some more benchmarks, and let's see what the results are. I plan to do the same with a 4000 as soon as I can get my hands on a couple test CPUs (assuming this thread doesn't die out).

[dgoodrich]

It is a socket 7 CPU. It originally ran at 350mhz (3.5X100) in an old Compaq notebook I had (which I upgraded to a K6-3+ running at 550 (5.5X100).

Do we know how the clocking works on the Snap 2000? It is user configurable at all? If we can't configure the FSB speed or the multiplier, I'd bet it would probably run at the same speed as the Intel chip thats in it (I'm guessing 66x3.5).

If it is configurable, I'd give it a try and report back.

[blue68f100]

Some upu only need 1 pin brought up to the High State. Which can be done using a simpler jumper.

[Phoenix32]

I do not know of any way to configure. You just brought out the very one thing I have been wondering myself. Back in the old socket 7 days, there were some motherboards that required you to set jumpers for the mulitplier and/or for the FSB. On some, you had to set it in BIOS. But there were also some that would auto detect and switch according to the CPU installed.

It may very well be that the multiplier on the SNAP 2000/4000 Main board is hard wired to the 3.5 x 66 and nothing we do is going to change it. It may also be hard wired to 66 FSB but might allow other mulipliers based on the CPU installed. It may be hard wired to 3.5 multiplier, but might allow either 66 or 100 FSB depending on the CPU installed. Or if we are real lucky, both might be auto selected by the CPU installed. I seriously doubt we are going to find this info anywhere, or at least be real hard to come by. I suspect the only way we are going to know is to install a few other socket 7 CPUs and find out.

Fact is, chances are, it is all hard wired to 3.5 x 66 and we are stuck with it, but we will never know without trying. As long as we use socket 7 CPUs, I doubt we can do any harm.

Now, if Douglas (Snap-Tech) can come up with some info for us and/or has a schematic of the main board being used, maybe a modification can be done to change the multiplier or even FSB if it is hard wired. Who knows? But as I have repeated here several times, we are not going to know without trying.

[Morpheus256]

there was a trick we use to use and still do on cobalt raq's which are socket 7 based, we used a single strand of wire, made a horseshoe out of it, and put it into the cpu socket then placed the chip in, it would change the multiplyier and pulled a signal to ground as a result casuing the cpu to run at a higher speed then the motherboard multiplyer specified.

[Phoenix32]

Yup

[jontz]

Ususally if a motherboard is set to a certain speed/multiplier, it is done with low value soldered resistors....perhaps just de-soldering and moving said resistors to an adjacent spot on the mobo would change the clock or the multiplier. I don't know anything about the 4000 mobo, but I have done this on other mobo's that were permanetly "set"...

[Phoenix32]

Yes, I have done mods like that myself. I did look this board over, and did not see any multiplier resistors anywhere on the board. This does not mean they do not exist, I just didn't see them anywhere.

Maybe Snap-Tech can come up with a Schematic for me...

[jontz]

Originally Posted by Phoenix32

Maybe Snap-Tech can come up with a Schematic for me...

US you mean Heck, we'll even include Blue, just for kicks...

[re3dyb0y]

Originally Posted by Morpheus256

there was a trick we use to use and still do on cobalt raq's which are socket 7 based, we used a single strand of wire, made a horseshoe out of it, and put it into the cpu socket then placed the chip in, it would change the multiplyier and pulled a signal to ground as a result casuing the cpu to run at a higher speed then the motherboard multiplyer specified.

Some of the Socket A boards needed this...

[Phoenix32]

Originally Posted by Morpheus256

ahhh but Phoenix if YOU read that thread you'd notice i also posted in there

(we can both have fun with eachother )

Then why didn't you use this thread sicne you knew about it?


Well, anywise,

Originally Posted by Morpheus256

digging through intels database, the 233mhz intel part had a core voltage of 2.7-2.9
the K6-2's ran at 2.2 core (ive got a stack sitting here)
thats why i suggested the powerleap device as it has built in voltage regulators to overcome this problem on regular pc's AND multiplyer hurdles on old pc's.

All of the Pentium 233Mhz I have pulled from 4x00 units has been 2.8V. If memory serves, the K6-2 CPUs did run at 2.2V as you indicate, but I do seem to recall some people got away with overvolting them. I do agree however that this is a pretty serious overvolt (18%). I do not usualy recommend more than 10% overvolting, and even then do not like doing it.

I see two problems with your suggestion of the powerleap adapter (or any CPU adapter). #1) Availability. Even on ebay, these are hard to even find now days. #2) It won't work for a 4100, which more people have. There is barely 1/4 inch clearance between the CPU fan and the top of the case in a 4100 as it is.

This does not mean it is a bad idea. In fact it may be worth investigating. Just go in with eyes wide open and look for the flaws. I know people were using those K6-2 processors in standard socket 7 motherboards without voltage control and without proper mulipliers etc. I just can't remember what they did anymore. It was more than likely an overvolt of the CPU and then some jumper wires wire-wrapped on the CPU pins. Been way too long ago for me to remember.

I was going to do some playing around with it, but I don't have any K6-2 CPUs laying around and no spare cash to go buying them, thus I never got to it.