VOGONS


a Choice between 2 P4's

Topic actions

Reply 21 of 39, by ODwilly

User metadata
Rank l33t
Rank
l33t

Watch upping the Voltage to high on those Northwoods. Northwood Sudden Death Syndrome is/was a thing back in the day and there is a Wikipedia article on it.

Main pc: Asus ROG 17. R9 5900HX, RTX 3070m, 16gb ddr4 3200, 1tb NVME.
Retro PC: Soyo P4S Dragon, 3gb ddr 266, 120gb Maxtor, Geforce Fx 5950 Ultra, SB Live! 5.1

Reply 23 of 39, by shamino

User metadata
Rank l33t
Rank
l33t
obobskivich wrote:
Sutekh94 wrote:

IIRC, the Dell was actively trying to downclock the Prescott chip, almost as if it knew how crazy hot those things can get. Even if those systems worked properly with that Prescott chip, the performance just wouldn't match an equivalent Northwood chip.

No curse, nothing creepy, etc. Prescott supports SpeedStep and will clock itself down (like any modern CPU) when it isn't under heavy load, or to improve cooling if it's overheating. A lot of older thick-client SFF machines shipped with 478/939 CPUs that they really weren't equipped to cool (~90W+ TDP chips), so the CPU usually ends up pretty aggressively throttled when put under load. Dell and HP were probably the absolute worst about this ca. ~2004-2007, and I think it's mostly those machines that end up perpetuating the stereotype of Pentium 4 as "awful."

I was searching the web somewhat recently for info on CPU support on the Dell GX270. I remember seeing a couple separate discussions which seemed to indicate that the GX270 will downclock the bus when it detects Prescotts faster than 3.0GHz. It also gives the user a warning message at POST. I'm not sure if that only applies to the SFF model, or all of them. I haven't tested this myself because I don't have the chip to do so.

Meanwhile, MSI 875P overclockers widely reported their motherboards dying with Prescotts, a chip which was apparently not yet released at the time the boards were made (but it was advertised to support it). MSI reportedly disabled Prescott overclocking in later BIOSes.
On my own MSI 875P, it didn't kill itself (yet) but it was emanating an incredible amount of heat compared to a similar clocked Northwood. Way more than the TDP difference would have led me to expect.

My interpretation of all this is that perhaps the motherboard industry failed to anticipate the actual power consumption of Prescott chips, and many motherboard VRMs weren't able to handle it safely. Dell apparently recognized this and patched the BIOS to keep the VRM within safe limits. I don't know if Intel made a late change to the spec, or board makers were just careless, but it seems board makers were taken by surprise.

Reply 24 of 39, by candle_86

User metadata
Rank l33t
Rank
l33t
shamino wrote:
I was searching the web somewhat recently for info on CPU support on the Dell GX270. I remember seeing a couple separate discus […]
Show full quote
obobskivich wrote:
Sutekh94 wrote:

IIRC, the Dell was actively trying to downclock the Prescott chip, almost as if it knew how crazy hot those things can get. Even if those systems worked properly with that Prescott chip, the performance just wouldn't match an equivalent Northwood chip.

No curse, nothing creepy, etc. Prescott supports SpeedStep and will clock itself down (like any modern CPU) when it isn't under heavy load, or to improve cooling if it's overheating. A lot of older thick-client SFF machines shipped with 478/939 CPUs that they really weren't equipped to cool (~90W+ TDP chips), so the CPU usually ends up pretty aggressively throttled when put under load. Dell and HP were probably the absolute worst about this ca. ~2004-2007, and I think it's mostly those machines that end up perpetuating the stereotype of Pentium 4 as "awful."

I was searching the web somewhat recently for info on CPU support on the Dell GX270. I remember seeing a couple separate discussions which seemed to indicate that the GX270 will downclock the bus when it detects Prescotts faster than 3.0GHz. It also gives the user a warning message at POST. I'm not sure if that only applies to the SFF model, or all of them. I haven't tested this myself because I don't have the chip to do so.

Meanwhile, MSI 875P overclockers widely reported their motherboards dying with Prescotts, a chip which was apparently not yet released at the time the boards were made (but it was advertised to support it). MSI reportedly disabled Prescott overclocking in later BIOSes.
On my own MSI 875P, it didn't kill itself (yet) but it was emanating an incredible amount of heat compared to a similar clocked Northwood. Way more than the TDP difference would have led me to expect.

My interpretation of all this is that perhaps the motherboard industry failed to anticipate the actual power consumption of Prescott chips, and many motherboard VRMs weren't able to handle it safely. Dell apparently recognized this and patched the BIOS to keep the VRM within safe limits. I don't know if Intel made a late change to the spec, or board makers were just careless, but it seems board makers were taken by surprise.

yea my Abit IC7 had a Prescott 3.2 in it, that did die suddenly, thats why I replaced it. It ideled at 56C and would load up to 79c. Now with this northwood at 3.6 1.60V stable i load at 63C 🤣

Reply 25 of 39, by TELVM

User metadata
Rank Oldbie
Rank
Oldbie
shamino wrote:

... My interpretation of all this is that perhaps the motherboard industry failed to anticipate the actual power consumption of Prescott chips, and many motherboard VRMs weren't able to handle it safely ...

Indeed, Preshott can trigger 'criticality excursions' on the unsinked VRMs/power MOSFETs of the day 😵 .

thermo_mosfet-b.jpg ····· cpu.gif

http://www.xbitlabs.com/articles/cases/displa … nn500af_11.html

Location of the VRMs on the ABIT IC7:

VzV4U4bX.png

http://hwbot.org/competition/hoc_mar12/stage/ … reference_clock

Let the air flow!

Reply 26 of 39, by alexanrs

User metadata
Rank l33t
Rank
l33t

In the first case wouldn't adding MOSFETs to the pads nearby (which seem to be shorted to the ones installed) kindda trivial and also help with the temperatures? (Sure, adding a heatsink afterwards would still be recommended).

Reply 27 of 39, by TELVM

User metadata
Rank Oldbie
Rank
Oldbie
alexanrs wrote:

... wouldn't adding MOSFETs to the pads nearby (which seem to be shorted to the ones installed) kindda trivial and also help with the temperatures? ...

Indeed!

vrm.jpg ····· power_supply.jpg

Let the air flow!

Reply 30 of 39, by 2fort5r

User metadata
Rank Member
Rank
Member

The IC7-G is a nice board but beware that the integrated gigabit Ethernet is buggy. You'll have to connect at 100 Mbps. Tx is OK but Rx at gigabit speeds will cause the computer to lock up. This is with board v1.1. I don't know if the problem was fixed in later revisions.

Edit: see the following:

http://discuss.extremetech.com/forums/4/37509 … ShowThread.aspx
https://lkml.org/lkml/2006/2/19/163
https://web.archive.org/web/20070702205315/ht … ead.php?t=18707

Last edited by 2fort5r on 2015-04-16, 14:06. Edited 2 times in total.

Account retired. Now posting as Errius.

Reply 31 of 39, by candle_86

User metadata
Rank l33t
Rank
l33t
2fort5r wrote:

The IC7-G is a nice board but beware that the integrated gigabit Ethernet is buggy. You'll have to connect at 100 Mbps. Tx is OK but Rx at gigabit speeds will cause the computer to lock up. This is with board v1.1. I don't know if the problem was fixed in later revisions.

I've done alot of file transfers with it, no issues at all.

Reply 32 of 39, by 2fort5r

User metadata
Rank Member
Rank
Member

The bug only kicks in with large volumes of data moving at high speed (c. >300 Mbps IIRC). You won't notice it if you're just copying small files.

Last edited by 2fort5r on 2015-04-15, 16:06. Edited 1 time in total.

Account retired. Now posting as Errius.

Reply 33 of 39, by obobskivich

User metadata
Rank l33t
Rank
l33t
RacoonRider wrote:

What's the milk-like substance around IC7 socket?

It looks like dielectric grease or something similar - most likely that machine was being prepped for cryogenic cooling (like LN2) and the grease prevents condensation from forming on the board.

Reply 34 of 39, by dr_st

User metadata
Rank l33t
Rank
l33t
2fort5r wrote:

The IC7-G is a nice board but beware that the integrated gigabit Ethernet is buggy.

Interesting. I wonder if it is a problem with the LAN itself, or just with the Abit implementation. I'll have to try with my P4C800-E one of these days...

https://cloakedthargoid.wordpress.com/ - Random content on hardware, software, games and toys

Reply 35 of 39, by 2fort5r

User metadata
Rank Member
Rank
Member

It's an ABIT problem. I have the P4C800-E as well and it works flawlessly. However it doesn't support 2 floppy drives, so instead of one annoyance you have another.

Account retired. Now posting as Errius.

Reply 36 of 39, by dr_st

User metadata
Rank l33t
Rank
l33t

Well, that depends on your needs. 😀 I can barely find any justification for one floppy drive on a P4 system that supports USB2.0 and Gigabit LAN. Much less two. 😉

https://cloakedthargoid.wordpress.com/ - Random content on hardware, software, games and toys

Reply 38 of 39, by candle_86

User metadata
Rank l33t
Rank
l33t

Well I transfered 230gb over that nic the other day, no issues at all, plus streamed media between the computers, need data on a bluray 50gb disk in main machine on my P4 no issues at all.

Reply 39 of 39, by candle_86

User metadata
Rank l33t
Rank
l33t

well good news a local guy that I buy stuff from often just ran across a P4 3.2 Northwood and asked if I wanted it, so yay my 3.0 is upgraded to a 3.2 🤣