VOGONS


Reply 20 of 31, by amadeus777999

User metadata
Rank Oldbie
Rank
Oldbie
ph4nt0m wrote:
Show quote

There's one thing about 256K PPro's. While the core is 350nm indeed, the cache die is still 600nm. So it may be a limiting factor. Many 256K PPro's don't even POST at 266MHz, though 1M chips don't do it either. The 512K PPro comes with a single 350nm cache die. It takes two of these for the 1M chip with a lot of power consumption. I have measured 11.7A @ 233MHz on the VRM while running Memtest.

I have recently overclocked one 200MHz/256K to 266MHz at 3.4V. It was unstable in benchmarks and couldn't compile a damn thing, so I modded it by soldering 10 1.0uF capacitors on the top.

ppro_caps_239.jpeg
ppro266_756.png

It's fine now, passes all benchmarks and can compile a Linux kernel. What's really interesting, this is an early SY013 with the A0 core revision and A cache revision. It isn't supposed to do 266MHz, but it does.

That's an awesome mod!

Only PPro machine that I saw in real life was back in '99 at a friends house - 24" CRT included.

The machine was a few years old and he must have spent a fortune on it. He wanted to sell it to me for a few hundred and I refused as I had no use for it.

Reply 21 of 31, by alvaro84

User metadata
Rank Member
Rank
Member

Cool to see it! I won't probably do it (though I can't rule it out!) but today I certainly learnt something new.

Shame on us, doomed from the start
May God have mercy on our dirty little hearts

Reply 22 of 31, by ph4nt0m

User metadata
Rank Member
Rank
Member

For those who are considering this mod for themselves, a great source of such 1.0uF capacitors is Socket 370 Coppermine chips whether Pentium 3 or Celeron. These are very inexpensive especially if broken because gold refiners don't want them usually. It's much more expensive to buy new capacitors from distributors. A set of 10 costs $5 to $10 there.

My Active Sales on CPU-World

Reply 23 of 31, by mrau

User metadata
Rank Oldbie
Rank
Oldbie

could someone explain please why the reading is so slow on the cpu graph? its the slowest operation on cache here;

also - have You tried resetting the bios? whats with other games, are they slow too?

Reply 24 of 31, by ph4nt0m

User metadata
Rank Member
Rank
Member

That's a Speedsys bug. The speed is fine. AIDA64 also doesn't get the L2 cache speed right under Windows 98.

vA6iRqt.png

RAMspeed (DOS) v2.5.0 by Rhett M. Hollander and Paul V. Bolotoff, 2002-09

1Gb per pass mode

INTEGER & READING 1 Kb block: 1055.79 MB/s
INTEGER & READING 2 Kb block: 1034.43 MB/s
INTEGER & READING 4 Kb block: 1046.53 MB/s
INTEGER & READING 8 Kb block: 1035.43 MB/s
INTEGER & READING 16 Kb block: 606.29 MB/s
INTEGER & READING 32 Kb block: 606.98 MB/s
INTEGER & READING 64 Kb block: 607.32 MB/s
INTEGER & READING 128 Kb block: 607.32 MB/s
INTEGER & READING 256 Kb block: 600.19 MB/s
INTEGER & READING 512 Kb block: 249.65 MB/s
INTEGER & READING 1024 Kb block: 249.65 MB/s
INTEGER & READING 2048 Kb block: 249.71 MB/s
INTEGER & READING 4096 Kb block: 249.71 MB/s

INTEGER & WRITING 1 Kb block: 935.32 MB/s
INTEGER & WRITING 2 Kb block: 924.05 MB/s
INTEGER & WRITING 4 Kb block: 934.50 MB/s
INTEGER & WRITING 8 Kb block: 933.69 MB/s
INTEGER & WRITING 16 Kb block: 565.72 MB/s
INTEGER & WRITING 32 Kb block: 566.02 MB/s
INTEGER & WRITING 64 Kb block: 566.32 MB/s
INTEGER & WRITING 128 Kb block: 560.99 MB/s
INTEGER & WRITING 256 Kb block: 528.42 MB/s
INTEGER & WRITING 512 Kb block: 90.99 MB/s
INTEGER & WRITING 1024 Kb block: 90.98 MB/s
INTEGER & WRITING 2048 Kb block: 90.94 MB/s
INTEGER & WRITING 4096 Kb block: 90.86 MB/s

INTEGER Copy: 109.51 MB/s
INTEGER Scale: 108.29 MB/s
INTEGER Add: 116.16 MB/s
INTEGER Triad: 113.03 MB/s
---
INTEGER AVERAGE: 111.75 MB/s

My Active Sales on CPU-World

Reply 25 of 31, by doaks80

User metadata
Rank Member
Rank
Member

Cpu cooler is a time traveler 😜

k6-3+ 400 / s3 virge DX+voodoo1 / awe32(32mb)
via c3 866 / s3 savage4+voodoo2 sli / audigy1+awe64(8mb)
athlon xp 3200+ / voodoo5 5500 / diamond mx300
pentium4 3400 / geforce fx5950U / audigy2 ZS
core2duo E8500 / radeon HD5850 / x-fi titanium

Reply 27 of 31, by Atom Ant

User metadata
Rank Newbie
Rank
Newbie
Cga.8086 wrote:

Zalman CNPS90F on a socket8, wow, nice too see that fit on the old socket

Yeah, I had to fit, because all my old cpu coolers were way too loud.

My high end of '96 gaming machine;
Intel PR440FX - Pentium Pro 200MHz 512K, Matrox Millenium I 4MB, Creative 3D Blaster Voodoo II 12MB SLI, 128MB EDO RAM, Creative Sound Blaster AWE64 Gold, 4x Creative CD reader, Windows 95...

Reply 28 of 31, by Atom Ant

User metadata
Rank Newbie
Rank
Newbie
ph4nt0m wrote:

For those who are considering this mod for themselves, a great source of such 1.0uF capacitors is Socket 370 Coppermine chips whether Pentium 3 or Celeron. These are very inexpensive especially if broken because gold refiners don't want them usually. It's much more expensive to buy new capacitors from distributors. A set of 10 costs $5 to $10 there.

What is this mod actually? Why the CPU will run faster if those capacitors are soldered?

My high end of '96 gaming machine;
Intel PR440FX - Pentium Pro 200MHz 512K, Matrox Millenium I 4MB, Creative 3D Blaster Voodoo II 12MB SLI, 128MB EDO RAM, Creative Sound Blaster AWE64 Gold, 4x Creative CD reader, Windows 95...

Reply 29 of 31, by ph4nt0m

User metadata
Rank Member
Rank
Member
Atom Ant wrote:
ph4nt0m wrote:

For those who are considering this mod for themselves, a great source of such 1.0uF capacitors is Socket 370 Coppermine chips whether Pentium 3 or Celeron. These are very inexpensive especially if broken because gold refiners don't want them usually. It's much more expensive to buy new capacitors from distributors. A set of 10 costs $5 to $10 there.

What is this mod actually? Why the CPU will run faster if those capacitors are soldered?

Because low ESR/ESL bypass capacitors are very important in high frequency designs. There are bulk electrolytic capacitors in the VRM, some tantalum and/or ceramic capacitors near the CPU socket, but none on the CPU itself. Internal CPU power traces, socket pins and mainboard power traces introduce a lot of impedance. So what seems to be a quality power supply isn't actually at the point of load.

My Active Sales on CPU-World

Reply 30 of 31, by Atom Ant

User metadata
Rank Newbie
Rank
Newbie

It is really imteresting an 1995 cpu can run at 266MHz with that modification. I woonder a 96' 512k version might could hit 300MHz?

Anybody can adwise hot to turn off vsync on Voodoo 2 SLI for Quake benchmark? I don't find Vsync as option to turn on/off.

My high end of '96 gaming machine;
Intel PR440FX - Pentium Pro 200MHz 512K, Matrox Millenium I 4MB, Creative 3D Blaster Voodoo II 12MB SLI, 128MB EDO RAM, Creative Sound Blaster AWE64 Gold, 4x Creative CD reader, Windows 95...

Reply 31 of 31, by ph4nt0m

User metadata
Rank Member
Rank
Member
Atom Ant wrote:

It is really imteresting an 1995 cpu can run at 266MHz with that modification. I woonder a 96' 512k version might could hit 300MHz?

300MHz not going to happen unless on liquid nitrogen or so. 266MHz is already extreme for the 350nm core and cache die. Besides there is no support for multipliers higher than 4x anyway.

My Active Sales on CPU-World