The only downside to Asrock from the mid-00's (04 to 06) are really shitty cap's.
Don't know if they changed that after 06/07 or something.
I have recapped my K7s41 once. The 3300 caps became bulgy after 10 year's.

Sutekh94 wrote:

ASRock definitely did make some interesting boards... I still remember a socket A board they made one time that had a slot for a socket 754 processor upgrade card.

And yeah, I'd also recommend 775 stuff if you're wanting to do a Prescott build. I've had some rather interesting experiences with my 3.2GHz 478 Prescott in the past.

This?
http://geizhals.de/asrock-939dual-sata2-dual- … dr-a160578.html

+

http://geizhals.de/asrock-am2cpu-board-a203571.html

I know this isnt the Ebay thread but this board should handle Prescotts depending on revision*.

I use one of these boards for testing newer "untested" AGP cards from Ebay as the motherboard seems to handle fires and shorts well. Note that you might need a Northwood CPU to be able to flash a Prescott capable BIOS but this motherboard model was later sold with a Prescott 3.0*. I have read about cap failures but I do not think it's wide spread.

Fujitsu Siemens D1625-A21 Socket 478 ATX Motherboard WORKING & TESTED £14.99 with free shipping in the UK, £4 shipping in the EU.
http://www.ebay.co.uk/itm/Fujitsu-Siemens-D16 … GEAAOSwCQNWdDcS

*Edit: I just checked, only the C revision was sold with Prescotts and only the C revison seems to have a Prescott capable BIOS avalible. The linked board seems to be an A revision board but I can not see any physical difference. In worst case they all work with the Northwood 3.4 and even the Gallatin and those are faster than the Prescotts anyhow.

The P4 chips on socket 478 aren't too bad and "only" go up to around 90W or so. It's the hotter 775 chips that can use up to 130W that be a challenge 🤣

PhilsComputerLab wrote:

The P4 chips on socket 478 aren't too bad and "only" go up to around 90W or so. It's the hotter 775 chips that can use up to 130W that be a challenge 🤣

Pentium D 805 @ 4ghz or higher and be cooking meat in a hurry 😎

stuvize wrote:

shamino wrote:

I've gotten the impression that mPGA478 motherboards generally have a hard time with safely handling the power load of Prescotts. The mosfets get very hot when running them, some boards are known for catastrophic failures with them (MSI), and Dell released a BIOS patch for some (all?) of their boards that forced an underclock of Prescotts faster than 3.0GHz. It seems that for some reason the industry underestimated the requirements imposed by those CPUs.
I don't have a specific board to recommend, I'm interested in the answer myself. But I think the VRM section's current capacity and the quality of it's cooling is the most serious consideration. The VRM is under much more stress with a Prescott than it is with a Northwood.

Same reason some 478 motherboards would not support Gallatin P4s even with a bios update the boards where just not designed for a CPU with that high of a TDP. Dell put heatsinks on the VRMs of many of their late 478 boards most likely to accommodate the power load of Prescotts

The MSI 875P Neo was purportedly designed for Prescott and advertised support for that CPU at launch, but ended up having problems with them. I'm not sure if MSI failed to comprehend the Prescott specs or if the specs changed at some late date.
In case anybody has considered that board, I want to offer a side note - it has lousy performance. The "Turbo" preset in the BIOS provokes a warning message of instability, and that message is truth. Yet for the loss of stability it only manages to bring 3DMark scores up to par with a Dell. Without that preset, it's slower.

The Dell GX270 board that I have has heatsinked mosfets, but it still has limitations with the Prescott. While researching upgrades for that board, what I've put together is that the later BIOSes were updated to refuse to run a Prescott of D1 or older stepping any faster than 3.0GHz. If you try, it will underclock the FSB to get it below 3GHz and display a warning message at POST.
I'm not sure if the GX270 was originally intended to run Prescotts though, so maybe it doesn't count as a design fail. I'm looking at a short Dell advertisement PDF for the GX270 and it only mentions a cache size of 512KB, which implies the systems were Northwood only at launch.
[edit: Somebody apparently found that the underclocking behavior doesn't happen if using an E0/E1 stepping CPU. It's triggered only when the BIOS detects a higher powered older stepping.]

I think the GX280 will run faster Prescotts, but it's an LGA775 board.

Later Prescott steppings reportedly brought down the power usage a bit, so they should be easier/safer to run. IIRC it's the E0 stepping that everybody raved about as being a worthwhile improvement. Some are listed as E0 and some as E1, I have no idea if there's any meaningful difference between those. There's also a rare G1 stepping but it seems they came out so late that not much was ever posted about them.

I have a ATI chipset socket 478 motherboard in storage, think it is MSI but been a good while, I used that with a 3GHz prescott for several years reliably back when it was fairly new. The on-board video was even pretty decent, played QUAKE3, Unreal tournament 3 etc just fine. Only problems were the common prescott ones, the case was too small, heatsink too weedy and fans too loud. But that pretty much sums up intel PC's from 2004/5 doesn't it??? 🤣 I later replaced the board with a desktop Pentium M 740 setup & geforce 6800 graphics, it was faster and way, WAY cooler & that is how the system still is today.

kanecvr wrote:

Asus i865 socket 478 boards suffer from "sudden inexplicable death syndrome" I have like 5-6 dead boards in my parts box.

"ASDS" - Asus Sudden Death Syndrome.
..........

Socket 478 may be the toughest socket to find a good board for. Lots going on at the time.

Vcore and Signaling voltages had dropped lower so power needed to be 'cleaner' than ever before (less ripple) to ensure good signal integrity while at the same time the needed power had gone up. In other words capacitors went from a light/medium duty job to a heavy duty job (so caps ran at higher internal temps than before) and the Chinese/Taiwanese or lower end Japanese capacitors the industry was accustomed to using simply weren't good enough anymore. The industry was still going through a learning process to get these things right.

Between 2001 and 2005 two of the Capacitor Plagues were in full swing.
1: The faulty stolen electrolyte fiasco. This affected some Chinese/Taiwanese brands only.
2: The defective (manufacturing error) Nichicon HM and HN series. A Japanese brand.
These both resulted in the same problems but the causes were completely different.

This is also when problems with Chemicon KZG and KZJ started showing up. These don't like heat and often fail without bloating.
Very similar in failures to OST caps. Unlike the previously mentioned two "Plagues", this never got fixed.

This is also when CPU power and heat hit new high points and both are hard on VRM caps and MOSFETs. Previously solid reliable parts were no longer good enough for the higher watt CPUs. If the board wasn't built to handle the CPU power (watts) from the get-go then just doing a micro-code update to allow Prescott (or Extreme Editions) could result in running too much power through the VRM (and traces) and I imagine some manufacturers tried to do that anyway.

........
I don't know of any brand that managed to avoid all those problems on all of their socket 478 boards.
You would have to look at them board by board, not brand vs brand.
.

Intel D865PERL is a great board. Supports the 3.4 and 3.4 EE officially from Intel, so I doubt there will be issues. If you use an EE chip, the boot screen changes with a different Logo 😀

I run a Gigabyte GA-8IPE1000-G with a 2.8Ghz hyperthreaded Northwood. Board supports Prescotts as well, but go for Revision 4.x boards. It came out late enough that it should have all the bugs worked out. Just watch for bad caps like the rest of the hardware from this era.

http://www.gigabyte.com/products/product-page.aspx?pid=1863

Nice board. Now isn't Northwood faster clock for clock than Prescott?

PCBONEZ wrote:

I don't know of any brand that managed to avoid all those problems on all of their socket 478 boards.

Exactly.
My main retro rig is ASUS P4P800 SE. Still alive after 11 years, last 3 years works 24/7.
But - I have replaced 2 Gigabyte (GA-8IPE1000 variants), 1 Asus P4C800E-Deluxe and 2 Asus P4P800 SE before I got this example.

PhilsComputerLab wrote:

Nice board. Now isn't Northwood faster clock for clock than Prescott?

Northwood has a 21 stage pipeline Prescott has a 31 stage pipeline so yes "clock for clock" Northwood is faster but the Prescotts 1MB L2 cache helps compensate for that and it has SSE3 commands. Its really hard to say which one is truly faster Prescotts almost always score more 3DMarks but Northwoods have a reputation for being faster.

I am fan of Pentiums 4 (mainly for nostalgic reasons) and from my experience Northwoods feel more responsive but seem inconsistent in their performance the Prescotts do seem sluggish comparatively but as a overall they tend to give a better experience at least in stuff that's really pushing them, if your running more era appropriate applications its really hard to see a difference. When the Prescott was released I think it was more about Intel getting a 64bit desktop CPU rather than making a faster Pentium 4

The branch miss-prediction penalty is huge with such a long pipeline. Intel was looking at 50 stages for Tejas. Madness!

Fascinating stuff.

The P4 EE also moved from Gallatin to Prescott at some stage.

There's a lot of benchmarks showing Northwood faster than Prescott, but those articles were written at a time when the Prescott was brand new. The results might be different with newer applications, especially if they utilize SSE3. I'm curious about that myself. Gaming-wise, I wonder if later games like the Prescott better than the games of ~2004 did.

I'm pretty convinced that the Prescott is better than the Northwood for decoding H.264 video, at least in VLC player anyway. It matters because those videos are marginal on a P4. The Prescott can manage to play things smoothly that will stutter on a Northwood. Maybe it's the cache, but I think it's more likely that the reason is the player/codec/whatever is programmed to benefit from SSE3.

I've tried 3 CPUs for H.264 playback on a Dell GX270 (865G based):
Northwood 2.6/800 Hyperthreaded - dropped a lot of frames on more demanding videos. Runs cool.
Prescott 2.8/533 non-HT D0 stepping - dropped a bit less. Runs hot.
Prescott 3.0/800 HT E1 stepping - Major improvement, almost perfect. The only way I got it to drop frames was with a high bitrate, fast moving 1080p video with the H.264 deblocking filter enabled. Runs noticeably cooler than the D0 Prescott did.

If you want a P4 to be able to play H.264 video, then I think the combination of Prescott + Hyperthreading is what it needs.

I'd still like to try a 3.4GHz Northwood HT, but they're a little pricey for something I might not end up using. Using the 2.6/800 as a guideline, I don't think it will scale to the same results as the 3.0 Prescott in video playback. Other applications might still prefer the Northwood though.

.
All of these (mentioned) boards were hit and miss, luck of the draw.
Sometimes you got good caps, sometimes you didn't.
Examples of all of those boards have been across my workbench numerous times.
.

PhilsComputerLab wrote:

Intel D865PERL is a great board.

Many had bad Nichicon HM/HN. KZG/KZJ are possible too. If you were lucky you got Rubycon.
Back in the day Intel published a notice to vendors about the HM/HN problems that listed all their boards that might have them installed.
I think all their socket 478 boards were affected.
I agree it's a great board.

Gamecollector wrote:

1 Asus P4C800E-Deluxe and 2 Asus P4P800 SE

P4Cxxx-x, P4Pxxx-x: Sometimes bad Nichicon HM/HN, sometimes KZG/KZJ, sometimes OST.
If you were lucky you got all Rubycon. Asus was using them, but not a lot.

NJRoadfan wrote:

I run a Gigabyte GA-8IPE1000-G.

GA-8IPE1000 (series): Some with bad Nichicon HM/HN and/or KZG/KZJ.
I've seen these with all the same caps or mixed caps (HM, KZG, MBZ) in every combination, even in the same board revision.
My assumption is that at the time Gigabyte was having trouble sourcing caps in sufficient quantities.
If you were lucky you got all Rubycon.
I used two of these myself. Liked them.
.
.
.
During that time (2001-2005)
Nichicon HM, Rubycon MBZ, Chemicon KZG, Sanyo WG, and those damned OST were the common VRM grade caps.
Of those only Rubycon MBZ and Sanyo WG were completely reliable. (Meaning 2001-2005, Nichicon fixed their problems after that.)
Supermicro was the only one heavily using WG so for most other brands what you wanted to see was the pretty blue MBZ.
.
The industry never recognized KZG/KZJ as problem series' but a little googling will show you different.
https://www.google.com/#q=bad+KZG
.

^My Asus P4C800-E has Rubycon Primaries and a mix of Rubycon and OST secondaries 😜

shamino wrote:

I'd still like to try a 3.4GHz Northwood HT, but they're a little pricey for something I might not end up using.

You should go for a 3.2ghz Northwood. For some reason they are all over the place on ebay for like $5-10

ODwilly wrote:

shamino wrote:

I'd still like to try a 3.4GHz Northwood HT, but they're a little pricey for something I might not end up using.

You should go for a 3.2ghz Northwood. For some reason they are all over the place on ebay for like $5-10

It's a common trend, the second best is affordable, the top product commands a premium.