VOGONS


Tillamook 266MHz and working L2 cache?

Topic actions

Reply 280 of 502, by kixs

User metadata
Rank l33t
Rank
l33t
Necrodude wrote on 2020-10-08, 20:33:

A friend of mine helped me mod my Gigabyte GA-5AX board with ALI Chipset.
My 266mhz tillamook now runs at 400mhz (4x100) witg working L2 cache

What revision is your GA-5AX motherboard?

Requests are also possible... /msg kixs

Reply 282 of 502, by kixs

User metadata
Rank l33t
Rank
l33t

Nice... In this case you can also increase the FSB speed as it supports up to 140MHz. Lower the multi and increase the FSB. This way you can also test the limits of this chip. I'm yet to play with mine 😀

Requests are also possible... /msg kixs

Reply 283 of 502, by Necrodude

User metadata
Rank Member
Rank
Member

I have not had any luck with high fsb:s on this board. The board wonks out when I go above 120.
The manual inclines that there are dividers on the PCI and agp that starts to work at 125. Those dividers doesnt seem to work at all.
I will be very happy if I can get it to run at 105x4=420 at a safe voltage. I dont want to fry the ship. Everyting above 2.1 volts starts to degrade the ship.
I can probaby run it at above that for many years because I dont use the computer often. I have a lot of different retro PC:s .

Reply 284 of 502, by RichB93

User metadata
Rank Member
Rank
Member

Can anyone else give input re: high voltages on this chip? My board with its so called 'magic auto voltage' function runs the chip at 3.3v. I've run it at this for a few hours now and it seems absolutely fine. The chip is only very very slightly warm; it runs cooler than my 233MMX suggesting that it is indeed running inside the lower <10W TDP. This suggests that the chip is running at a higher voltage but pulling less amps (because V*A=W). As it is, running these chips on a desktop board already increases the I/O voltage from 2.5 to 3.3v.

Can anyone else support what I'm saying or attest to running their chip at a high voltage for extended periods of time?

Whilst the higher voltage is out of spec, I just wonder if at least some boards do current limiting so that the chips don't fry.

Reply 285 of 502, by H3nrik V!

User metadata
Rank Oldbie
Rank
Oldbie
RichB93 wrote on 2020-10-09, 13:43:

This suggests that the chip is running at a higher voltage but pulling less amps (because V*A=W). As it is, running these chips on a desktop board already increases the I/O voltage from 2.5 to 3.3v.

Unfortunately, it doesn't work that way with CPUs. They act closely like resistive loads, as in, if you increase voltage by 10%, the current will also rise roughly 10%, increasing power consumption by 21% (plus 10% of a number, that is already increased by 10%). Current consumption rises exponentially with voltage increase ..

Please use the "quote" option if asking questions to what I write - it will really up the chances of me noticing 😀

Reply 286 of 502, by RichB93

User metadata
Rank Member
Rank
Member
H3nrik V! wrote on 2020-10-09, 16:21:
RichB93 wrote on 2020-10-09, 13:43:

This suggests that the chip is running at a higher voltage but pulling less amps (because V*A=W). As it is, running these chips on a desktop board already increases the I/O voltage from 2.5 to 3.3v.

Unfortunately, it doesn't work that way with CPUs. They act closely like resistive loads, as in, if you increase voltage by 10%, the current will also rise roughly 10%, increasing power consumption by 21% (plus 10% of a number, that is already increased by 10%). Current consumption rises exponentially with voltage increase ..

Hmm. That makes sense, but I'm just so confused as to why my chip isn't barely getting warm. I would've thought that at such a high voltage, the thing would've been staggeringly hot but it isn't the case - if it's not being dissipated as heat, where is it? There's a linear regulator on my board with a heatsink on it that *is* getting very hot however.

Reply 287 of 502, by H3nrik V!

User metadata
Rank Oldbie
Rank
Oldbie

Maybe, and this is a long shot, your regulator gets SO hot so it actually outputs a lower than expected voltage and all the power is consumed in the regulator in stead - all though I highly doubt it.

But what is the nominal power consumption of the Tillamook actually?

Please use the "quote" option if asking questions to what I write - it will really up the chances of me noticing 😀

Reply 288 of 502, by Necrodude

User metadata
Rank Member
Rank
Member
RichB93 wrote on 2020-10-09, 13:43:

Can anyone else give input re: high voltages on this chip? My board with its so called 'magic auto voltage' function runs the chip at 3.3v. I've run it at this for a few hours now and it seems absolutely fine. The chip is only very very slightly warm; it runs cooler than my 233MMX suggesting that it is indeed running inside the lower <10W TDP. This suggests that the chip is running at a higher voltage but pulling less amps (because V*A=W). As it is, running these chips on a desktop board already increases the I/O voltage from 2.5 to 3.3v.

Can anyone else support what I'm saying or attest to running their chip at a high voltage for extended periods of time?

Whilst the higher voltage is out of spec, I just wonder if at least some boards do current limiting so that the chips don't fry.

http://www.pchardwarelinks.com/elec_pentium.htm

You could also check out cpu world
https://www.cpu-world.com/CPUs/Pentium/Intel- … FV80503266.html

Attachments

  • tillamook.jpg
    Filename
    tillamook.jpg
    File size
    79.45 KiB
    Views
    1778 views
    File license
    Public domain

Reply 289 of 502, by RichB93

User metadata
Rank Member
Rank
Member
H3nrik V! wrote on 2020-10-09, 17:54:

Maybe, and this is a long shot, your regulator gets SO hot so it actually outputs a lower than expected voltage and all the power is consumed in the regulator in stead - all though I highly doubt it.

But what is the nominal power consumption of the Tillamook actually?

Necrodude wrote on 2020-10-09, 20:51:
http://www.pchardwarelinks.com/elec_pentium.htm […]
Show full quote
RichB93 wrote on 2020-10-09, 13:43:

Can anyone else give input re: high voltages on this chip? My board with its so called 'magic auto voltage' function runs the chip at 3.3v. I've run it at this for a few hours now and it seems absolutely fine. The chip is only very very slightly warm; it runs cooler than my 233MMX suggesting that it is indeed running inside the lower <10W TDP. This suggests that the chip is running at a higher voltage but pulling less amps (because V*A=W). As it is, running these chips on a desktop board already increases the I/O voltage from 2.5 to 3.3v.

Can anyone else support what I'm saying or attest to running their chip at a high voltage for extended periods of time?

Whilst the higher voltage is out of spec, I just wonder if at least some boards do current limiting so that the chips don't fry.

http://www.pchardwarelinks.com/elec_pentium.htm

You could also check out cpu world
https://www.cpu-world.com/CPUs/Pentium/Intel- … FV80503266.html

I've measured at the coil near the CPU on my board and it's definitely measuring 3.3v. I don't know how I'd measure current however. It posted and worked fine at 3.5v at least long enough to get into Windows and for me to gracefully shut the machine down. But as of now, I've run the chip for a good few hours at 3.3v with no stability issues - 3DMark 2000 benchmarks, a good bit of Carmageddon II (which is amazingly playable with the extra 33MHz of the 266MHz chip compared to a 233MHz MMX), Final Reality etc. Everything has been absolutely fine. I would love to find a way of getting the voltage down, but it's not straightforward on my board. I tried the chip in my M550 board too but cache doesn't work properly and the board doesn't have Microcode for it (maybe related), but it does run at a healthier 2.1v core.

Reply 291 of 502, by RichB93

User metadata
Rank Member
Rank
Member
Necrodude wrote on 2020-10-09, 22:15:

A friend of mine gave the cpu 3.3 volts and did the 4x multi mod and the cachemod. The cpu died after 2 days at the end of a retro lan party.
The cpu wont last long at 3.3 volts

Thanks for the heads up - I'll stop using it until I can find a way to get the voltage down. Not sure how easy it'll be however 🙁 Just so odd because if the chip isn't getting hot, where is the energy going? It just doesn't make sense.

Reply 293 of 502, by RichB93

User metadata
Rank Member
Rank
Member
Necrodude wrote on 2020-10-09, 22:36:

Well it is a cool running low power laptop chip.

But that's the thing though - if its running at 3.3v, naturally it would run hotter and require additional cooling?

EDIT: I can probably ditch this theory - perhaps the cooler is just doing a good job... unplugged the fan and it warms up pretty quickly. Man I wish I could figure out how to disable this damn auto voltage thing. I can only get it to run at 3.3 or 3.52v 🙁

Reply 294 of 502, by Necrodude

User metadata
Rank Member
Rank
Member
RichB93 wrote on 2020-10-09, 22:23:
Necrodude wrote on 2020-10-09, 22:15:

A friend of mine gave the cpu 3.3 volts and did the 4x multi mod and the cachemod. The cpu died after 2 days at the end of a retro lan party.
The cpu wont last long at 3.3 volts

Thanks for the heads up - I'll stop using it until I can find a way to get the voltage down. Not sure how easy it'll be however 🙁 Just so odd because if the chip isn't getting hot, where is the energy going? It just doesn't make sense.

I found this formula for calculating cpu wattage
P=V^2*C*f
where:
p=cpu wattage. Acording to cpu world the typical load is 4,5 watts

https://www.cpu-world.com/CPUs/Pentium/Intel- … FV80503266.html

V=cpu voltage V^2=means cpu voltage X cpu voltage
C=cpu capacitance ( we dont know this one)
F=Cpu frequency in mhz. The tillamook has a frequency of 266mhz.

We dont know what the capacitance is. So lets calculate it.
P=V^2*C*f

4,5=(1,9*1,9)*C*266
4,5=3,61*C*266
4,5=960*C

1/C=960/4,5
1/C=213,3911111111
1/213,39=C/1
C=0,0046862308 farad

know we know C.

We can now change the voltage and see the typical wattage with a vcore of 3.3volt
P=V^2*C*f
W=3,3^2*0,0046862308*266
W=10,89*0,0046862308*266
W=13,5747 watts typical load.

In not entirely sure if I did calculate this 100% correct. But I think it is somewhat correct
A normal pentium mmx has a typical load of 9.6 watts and a maximum load of 21 watts at 2,8volts vcore

https://www.cpu-world.com/CPUs/Pentium/Intel- … BP80503233.html

So a typical wattage at 13.57watts at 3.3 volts dont sound that far off

Reply 295 of 502, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
RichB93 wrote on 2020-10-09, 22:40:
Necrodude wrote on 2020-10-09, 22:36:

Well it is a cool running low power laptop chip.

But that's the thing though - if its running at 3.3v, naturally it would run hotter and require additional cooling?

EDIT: I can probably ditch this theory - perhaps the cooler is just doing a good job... unplugged the fan and it warms up pretty quickly. Man I wish I could figure out how to disable this damn auto voltage thing. I can only get it to run at 3.3 or 3.52v 🙁

(s)s7 boards usually only had the option to set i/o voltage to 3.3v or 3.52v. For getting lower i/o voltages, my guess is that either the board would need to be modified, or perhaps modified by using some kind of CPU adapter (kinda like how the POD83 and the POD63 lower i/o voltage from 5v to 3.3v on Socket 3 boards).

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 296 of 502, by RichB93

User metadata
Rank Member
Rank
Member

Bit of an update on my particular situation, as per my post here (Re: Modifying Socket 7 motherboards for 1.9v (Tillamook) CPU voltage), I was able to determine that pin AL1 on the Tillamook CPUs is not connected, unlike it is on standard P55C chips - meaning that the board was treating the chip as a single rail voltage of 3.3 or 3.5v. I found a nearby undocumented jumper that connects to AL1 and controls the voltage regulation on the board. By pulling it to ground (I'll eventually solder the chip), the board now runs in dual rail mode and by mucking around with jumpers I've been able to get down to 2.5v. Just need to try and get further now. Wonder how long these chips last at 2.5v? I mean, its a good deal lower than the insane 3.3v it was being fed before.

I've also got the chip running at 75MHz FSB too, so a nice 300MHz (4*75).

Oh, and yes, do ignore all of my previous rubbish about no heat = being okay. This is running even cooler at 2.5v so at the intended 1.9v it could probably run passively 😜

Reply 297 of 502, by RichB93

User metadata
Rank Member
Rank
Member

Sorry to bump, but I did some investigation, and I figured that others would be interested too. As per the Tillamook datasheet (attached), the absolute maximum voltages are defined as such:

Screenshot 2020-10-10 at 18.25.12.png
Filename
Screenshot 2020-10-10 at 18.25.12.png
File size
195.92 KiB
Views
1694 views
File comment
Absolute maximum ratings from datasheet (pg. 42)
File license
Fair use/fair dealing exception

This means that the core (VCC2) should be able to withstand 2.8v as the absolute maximum voltage, and the I/O as 3.2V. Now, I assume that on all desktop boards, the I/O voltage is 3.3V. I wonder if this is an issue long-term.

Perhaps anyone else who has been running a Tillamook chip for any length of time can provide any further insight? This puts my mind at rest running the chip at 2.5v, but of course it would be nicer to bring it down to 1.9 or 2.0/2.1v.

Attachments

  • Filename
    tillamook.pdf
    File size
    1.1 MiB
    Downloads
    82 downloads
    File comment
    Tillamook datasheet
    File license
    Fair use/fair dealing exception

Reply 298 of 502, by RichB93

User metadata
Rank Member
Rank
Member
RichB93 wrote on 2020-10-10, 15:41:

Bit of an update on my particular situation, as per my post here (Re: Modifying Socket 7 motherboards for 1.9v (Tillamook) CPU voltage), I was able to determine that pin AL1 on the Tillamook CPUs is not connected, unlike it is on standard P55C chips - meaning that the board was treating the chip as a single rail voltage of 3.3 or 3.5v. I found a nearby undocumented jumper that connects to AL1 and controls the voltage regulation on the board. By pulling it to ground (I'll eventually solder the chip), the board now runs in dual rail mode and by mucking around with jumpers I've been able to get down to 2.5v. Just need to try and get further now. Wonder how long these chips last at 2.5v? I mean, its a good deal lower than the insane 3.3v it was being fed before.

I've also got the chip running at 75MHz FSB too, so a nice 300MHz (4*75).

Oh, and yes, do ignore all of my previous rubbish about no heat = being okay. This is running even cooler at 2.5v so at the intended 1.9v it could probably run passively 😜

Just wanted to update the diagram earlier in the post to add what I'm on about here. VCC2DET is not connected on the Tillamook chip so some boards will detect is as a single rail voltage chip and run it either only at 3.3 or 3.5v on both the core and I/O. You can connect VCC2DET to ground (as it is on the P55C chips) to enable split rail voltage and get the core voltage down (but I/O will still be at 3.3v). Maybe this is why some other boards don't like the chip?

Tillamook_chip_modifications.png
Filename
Tillamook_chip_modifications.png
File size
691.32 KiB
Views
1622 views
File license
Public domain

Reply 299 of 502, by feipoa

User metadata
Rank l33t++
Rank
l33t++

RichB93, that is an interesting discovery about the AL1 pin not being connected to GND, thereby having some MB's forcing the chip into single rail mode. It could explain why many boards don't work with the Tillamook, but does a Tillamook generally not turn on at 3.3V single-rail?

Plan your life wisely, you'll be dead before you know it.