Every man loves his toy, and age has no bearing on the urge in getting the most talk about “hot” item on the market

PRESCOTT 2.8E Overclocking and more

The Lazyman’s Review

2/12/2004

Last update 2-14-2004

_______________________________________________________________________________

This review is getting longer than I planned originally. It started out a simply look into the Prescott and ended up testing it with two different cases and three different mother boards. So, read on, you might be surprise to see the difference between mother boards and cases.

PART I

Every man loves his toy, and age has no bearing on the urge in getting the most talk about “hot” item on the market. In this case, I am talking about the “hot potato”, Intel Prescott 2.8E. Months before the Feb.2 2004 release, the Prescott had a long delay and rumors filled thermal disastrous controversy and the lack of performance to the current Northwood.

How it all started

On the release day, all rumors were vindicated. In deed, Prescott is a hot potato. Well, 10 days later, I am holding one of these Prescott 2.8E. Let’s take a look at what I’ve got. I could put the 2.8E in one of the three systems; IC7, IS7 or P4P800. I decided to try the least capable board, the IS7 in a not so desirable case. Here is the setup:

IS7 BIOS 11/04/03

Codegen Case 1x92 intake, 1x80 exhaust, 1x80 top blower

Codegen 400 Watts PSU

Western Digital 60 GB HD

NEC DVD

NEC 12x RW

Corsair PC2700C2 2x256

Swiftech MCX478-V 80x80x25 mm 3000 RPM

Arctic Silver-AS3

Win XP Pro Sp1

I was going to use the Intel copper base stock heatsink and fan, but decided to use the Swiftech fear of the heat rumor. With the side panel close, Window booted up the first time at default V-core and speed, no sweat. Quickly glanced over the MBM5, the idle temp was 49C and system temp was 30C while ambient temp was 22.3C. For comparison, the 2.4C @3 ghz I took out has an 34C idle and 27C system.

The next immediate agenda for me was to reboot and kick up the FSB to 250 still at default V-core. Window started to boot, but did not reach log on screen. I backed down to 245 FSB, restarted, and I sensed a relief. I ran Sandra burn-in for a short 5 loops just to see if the system could stay on, it did. I ran MetaBench, it finished without a hiccup. It now idles at 50C with 31C system. For whatever reason, Sandra does not report CPU wattage and therefore no C/W with the 2.8E. Sandra burn-in time again, 10 loops set off the preset 70C alarm, and it reached 70C when it got done. This is HOT man, and the PWM shot up to 58C.

Koolsolutions Chill Vent II time; I let the system idle for 15 minutes and snap on the Chill Vent II still at 245FSB and default vcore. It now idles at 46C. Here is what I got from 10-loop Sandra; no alarm this time, 63C and the PWM topped 57.5C. This is a 7C reduction.

Bench marking using Sandra at 2.8Ghz was disappointing, as all numbers fell behind 2.8C. The same applies to overclocked speed of 3.43 Ghz, which falls behind 3.2C.

Well, the Prescott is indeed very hot and it is definitely behind Northwood clock for clock. It is highly overclockable as this sample shows a 640 mhz increase at default voltage. I will do more tweaking in the heat department in the next few days, may be a better case and louder faster CPU fan. The Chill Vent II knocks down the very much needed 7C.

PART II

Prescott 2.8E and ABIT IC7

The CASE

The Aspire XDreamer ATX case was well received by many after reading some of the reviews. This kind of case is well liked by people who perceive the number of case fan dictates the “performance” of the rig; I mean the number of LED fans too.

This is how it looks after I assembled the unit.

NICE LOOKING!!

Briefly about the case, since I am not giving it a review. I find it well assembled with no sharp edges so you could save the band-aid. The automotive paint finish like black painted metal case has a glossy look with an about 2.5 mm thick transparent plastic front cover. It has 2x80 mm fan slots in the front with some holes pre-molded to let in fresh air. There are 2x80 exhaust fans, 1x80 top blower and 1x80 mm side panel fan. Only the side panel and top blower fans come with the case. The window side panel is clear without tint and held by 8 screws with chrome plated facing. There is a temp probe and digital readout mounted next to the power and reset switches. See the “blue light” on the front panel? That’s it. No worry, it did not work. Mine reads 77 degrees F, and stayed unchanged the whole time I did the following test.

There are two front USB ports and a blank FireWire port. There is room for 10 bays; 4 5.25”, 2 external 3.5” and 4 internal. There is a nice touch having two black CD face plates that cover my beige color CD’s. They work really well. It is a bit loud with all 8x80 fans running. The CEM DT-8850 sound level meter registered 50.7 dBA 24 inches away.

I replaced the 350 Watts PSU that came with the case with a 450 Watts. So, I can’t comment on it, although it is one of those that weight less than pound and a half type.

The TEST bed

ABIT IC7 Bios date 1/24/2004

Prescott 2.8E

Swiftech MCX478-V + 80x80x25 mm fan @3000 rpm

Arctic Silver – AS3

SIMENS PC3200 2x256

Western Digital 80 GB

MSI CD RW

Samsung DVD 12X

ATI Radeon 9000 Pro 128

I did not want to go through the default numbers again because I thought it would be similar to my previous test. Mind you that the default voltage is 1.375 volt. I started right to 245 FSB at default voltage and memory set to 5:4 ratio. Test was done at 22C ambient temp.

Intel Stock HSF is the latest with rectangular square copper base.

Load temp was obtained using Sandra burn-in 15-runs. Temps were recorded from MBM 5.3.5. The temperatures are slightly higher than previous test using IS7 board. I could attribute this to the mod Codegen case front intake using 92 mm fan. The Koolsolutions Chill Vent II found its way into the side panel fan parallel to CPU location, although they are not in perfect alignment. I believe using a more powerful Delta fan might well decrease another few C’s knowing the MCX478-V is very capable. But, I just can’t stand the noise, may as well get rid of the Prescott. Again, Chill Vent provides a few more degrees reduction.

Conclusion:

The Prescott is hot, in fact too hot. I wonder how the mass market will put Prescott on their shelf. How many customers will be complaining or returning the brand new computers they brought home only a day or two earlier. Both tests were carried out using cases with minimum 3 to 6 case fans. Swiftech is top of the line, and quite expensive at $50 landed without the fan. I have tried the Thermalright SLK800U using the same 80 mm fan and found temperature deviation of less than 1C with Swiftech. The Koolsolutions Chill Vent is $19.

Intel thermal specification of 69C tcase for the 2.8E is simply ridiculous. If the XDreamer case with 8 case fans would not do much in keeping the CPU temp below 61C at default settings while the ambient temp is 22C in the middle of Chicago winter, how could Intel expect a maximum Tcase temp of 69C in the dog days of summer? With Chill Vent at default speed, the load temp idles at 44C and top just below 60C use in conjunction with the MCX478-V.

Besides being hot, it does not perform to the current Northwood. Intel should not have released this processor until they resolve the heat and performance issues. Looking impressive on papers does not mean anything, telling customers things will get better are simply irresponsible, Intel sucks this time.

PART III

Prescott 2.8E and P4P800 Deluxe

Well, 3.5 GHz (250 FSB) is very stable, passed Prime95 required a 0.05 volt increase using the IC7 and Xdreamer setup. My temp probe broke for quite sometime, so I went and got replacement today.

Some of us old hands know that mother board report temperatures are at the very least suspects. For AMD, a thermister is installed at the middle of the socket and may be in different locations within. Intel P4 does not make it any easier, most BIOS temp readings are from diode with “calculated or computations” compensation factor in the BIOS. There you have it. Some people jump over joy seeing a low CPU temp, others cry over the high temps. The same reason all OEM systems from Compaq, Dell, Gateway, Sony and eMachines do away with any temperature reporting. Customers will not ask or complain what they don’t know.

Using the new probe and using a +10C factor (adding to the probe reading-due to location of the probe is not at center of the CPU heat-spreader as recommended by Intel), I found a 8 degrees Celsius lower than my review has shown. Using Swiftech published c/w, Intel CPU wattage specs, ambient temp, temp at CPU fan intake etc, I was surprised at the temps listed on the review. Subtracting 8 degrees makes the calculated number more in line with the external temp probe readings.

So, it is a safe assumption that 8 degrees could be subtracted from the IS7 and IC7 temps. Okay, time to take my P4P800 Deluxe resting in the box for two months and put the Prescott 2.8E in. The P4P800 was in the box because I did not like the low reporting temps after I had it installed two months ago. My thought is that it will showcase the Hot Potato to a lesser “degree”.

Here is the set up (you must know the board by looking at this).

I added a 40 x 10mm on the passive P4P800 deluxe NB heatsink.

While the IC7 was capable running at 250 FSB at 1.475 Vcore the P4P800 is not. I believe the problem came from the MOSFETS, as the Vcore dropped from as high as 1.500 BIOS to 1.32 under load while the 12 volt rail was at no less than 11.70 volt. The best I could settle for now is 245 FSB at 1.485 volt in BIOS.

Using the identical setup with only motherboard change, the surprise is 38C idle and 57C load temp at ~21.50 C ambient. Don’t tell me all motherboards are created equal. And, please don’t ask me why a motherboard could change CPU temp when you think heatsink, case ventilation, thermal paste, type of CPU fan and Power Supply are the players.

The Idle 38 C and load 57 C.

The not so useful side panel fan and the versatile Chill Vent. The light effect.

Another CPU-Z shot to illustrate the difference of Vcore from the same 1.475 BIOS load. The Asus shows up 1.504 here.

I am leaving the Koolsolutions Chill Vent installed and not planning on doing any comparison this time, I’d believe the reduction from the Vent would be no difference than in Part I and II. Three different boards, two cases and the same fan, here is the summary:

The P4P800 required slightly higher Vcore than both the IS7 and IC7 at 245 FSB yet with lower temps. Both IC7 and IS7 could reach 250 FSB and the P4P800 could not at the same Vcore setting. Which motherboard temp is to be trusted? I’d suspect they all are inaccurate. The ABIT owner may go for water-cooling only to find out having the same 100% air cooling temp of the Asus owner. The Asus owner brags about the low temp only to find out the system is unstable. This is the very reason Mass Market computers have no temperature reading whatsoever, what you don’t know won’t hurt you, or you won’t ask for what you don’t know.

Final Thought

The Prescott is hot no matter which way you look at it. Would Intel keep the 478 socket after they release the 775 socket to improve on the current Prescott setup, I don’t think so. It may be in the Celeron line. Prescott 478 will be the repeat of the 423, a 6-month product cycle. When you must get water-cooling for processor, I believe there is something wrong with it to begin with.

Until next time, cheers!

Note: Both P4P800 Deluxe and IC7 have new BIOS released soon after the tests were completed.

Be sure to check out Aspire X-Dreamer plus Chill Vent II.