PDA

View Full Version : Evergreen Spectra III CPU


Godfodda
07-23-01, 11:18 PM
These CPUs were being discussed in the previous thread "Unlocked PIII? Yes!" It was touted as being a *multiplier* unlocked CPU. Well, you gotta have one of these. I set myself up as guinea pig and bought one. I had hoped for enough of a surprise to warrant writing an article, but, as everyone suspected, the only surprise was the terrible performance. I'll at least pass on some information here. If the moderator feels this belongs elsewhere, or should be shared elsewhere, please move or copy at will.

I ordered the Evergreen Spectra III 550 from an online vendor for $49 + shipping (actually not too bad a price for a toy I've never played with :-) ). Arrived today.

Cracked open the box and went straight for the CPU. As had been discussed, it was a VIA Cyrix III. Sticker on the plastic case says 128K L1, FSB max 100, CPU core speed max 550, CPU volts 1.9.

The core was about the size of my son's Celly 400 core. The ceramic was also about as thick. A notable difference, though, was the pin board. "There is no pin board." Well, there is on this thing. Attached to the bottom of the CPU there is a plastic green board like a plain old circuit board, but thinner and with no traces. The board is stamped Evergreen Technologies, Inc." The pins come through holes in this board. But there aren't 370 pins. Two pins are gone... not talking about the two at the corners. These are in the first column (looking at pin side with pin 1 upper right) and the next column over. One of these two empty spots is soldered. In my ignorance on these matters, I'm assuming this is the unlocking mechanism.

I tested the CPU in my Abit VH6-2 with 768M RAM. For reference, I used my P3 733 at default speed (133x5.5), and my son's Celeron 400 also at default speed (66x6.0). The only test I used on all three was Sandra's CPU Benchmark. You'll understand why I used only one when you see the numbers. I tested the reference CPUs first. The 733 gave 1971 MIPS, 981 MFLOPS. The 400 gave 1062, and 528.

At the advertised 550 for the Evergreen / Cyrix, Sandra showed 653 and 175. 400 (66x6) was 470, 126. 700 (100x7, and the nearest I could get to the 733 without wasting more of my time) showed 835 and 223. Wow :-D

I did pull out a few interesting things from this, though. Point 1, the multiplier is, indeed, unlocked. I went from 4 to 7 in my simple and hurried tests. Are the missing pins the key to unlocking the Cyrix? Would a P3 work like this, too?

Point 2, I discovered that my Abit VH6-2 supports higher voltages than 1.85. This is the max voltage shown with my 1.65V 733 in socket. Both the Celeron 400 and the Evergreen / Cyrix were 1.9 or higher. The voltage options increased accordingly when these were in.

Point 3 still needs investigating. The installation guide that came with the CPU (a floppy was also included), says that on the floppy there is a program called Eticlock that will allow the multiplier to be software adjusted (the way that SoftFSB allows the FSB to be adjusted). As I said, this still needs investigating. I didn't make it past "Holy s*** that thing is slow!" tonight. I'll test out the Eticlock program soon and post back if something looks interesting.

Don't know to what to attribute the poor performance. Some of you tech guys can probably answer those questions. I'll try to snag a digicam within the next couple of days so you all can see this pin board I'm talking about. Maybe that'll give someone a spark to break their P3 using a similar pin chopping technique. |-) All I know is that I at least have a backup processor now... even though it was a $49 P1. :-)

Endeavor
07-24-01, 01:09 AM
is it slow compared to a PIII of the same speed or just slow in general?

dozier768
07-24-01, 02:27 AM
the cyrix has always been a lowend low performer, last thing i saw from evergreen as a pentuim upgrade for a 486 lol i didnt even know they were still around

Godfodda
07-24-01, 12:47 PM
Endeavor (Jul 24, 2001 01:09 a.m.):
is it slow compared to a PIII of the same speed or just slow in general?

Yes and yes. That's why I used two processors for comparison. Stepped it up for one and down for the other. Even at the default max speed, it still benchmarked slower than the Celeron 400. Just slow.

outhouse
07-24-01, 07:15 PM
The missing pins may very well be for CV if you look at one of the vidpinning links you will be able to confirm my guess, and as far as the VH6-2 ive got mine up to 2.05 for CV with a vidpinned p3. hope you get some better performance out of it but at that price its not too much of a loss if its not what you desire.

Godfodda
07-24-01, 10:15 PM
outhouse (Jul 24, 2001 07:15 p.m.):
The missing pins may very well be for CV if you look at one of the vidpinning links you will be able to confirm my guess.

Can you point me toward these links? I must be looking in the wrong direction.

vimal
07-26-01, 12:03 PM
The Spectra 550 is a Cyrix 550 repackaged. All Cyrix chips are inherently unlocked.

Here is a link to the datasheet for the PIII-FCPGA version. Somewhere in that document is a pin diagram telling you which pins have what names. I would assume (as in 95% sure, but havent done extensive research) that all FCPGA processors have the same pinouts (except for the Tualatin-S, but we won't worry about that)

http://developer.intel.com/design/pentiumiii/datashts/245264.htm

I doubt the VCore ID pins were clipped, otherwise how would your motherboard know what voltage to give it? Likewise, I doubt that the multiplier pins are clipped otherwise the mobo can't set the multiplier.

If you want to try pin-clipping on other processors, use nailpolish or some other thin coating material to cover them instead of actually breaking it. That way if it doesn't work, you can always scrape off the coating without too much fuss.

As for the whole SoftFSB type program that comes with the proc - well, I suppose there may be a way for the program to control what multiplier the processor is set to (some special instruction outside of the x86/MMX/SSE/3DNow! set) but bus control depends entirely on the motherboard.