Powerleap PL-iP3/T & Celeron 1.2 GHz Upgrade Review

written by Robert Hancock

Also see the updated PL-iP3/T version 2.0 and Celeron 1.4 GHz upgrade review page.

Note: I replaced the Sandra screenshots on this page with text charts to reduce bandwidth usage.

I recently received this upgrade package from Powerleap for evaluation. Below is some information on the adapter, the CPU, some installation notes and benchmarks, and my conclusions.

PL-iP3/T Product Information:

You can find official info and pictures for the PL-iP3/T on Powerleap's site here.

The PL-iP3/T is a new version (version 3) of Powerleap's PL-iP3 upgrade adapter. The PL-iP3 is much like any other slocket adapter, converting a Socket 370 chip to work in a Slot 1 motherboard, but unlike a normal slocket, it has an onboard regulator to supply power to the CPU. A typical slocket simply transfers the voltage ID the CPU requests to the motherboard, which supplies the voltage through the slocket to the CPU. This means that the only voltages that can be supplied to the CPU are those that the motherboard supports. In contrast, the PL-iP3 allows installation of CPUs in a system whose motherboard cannot supply the voltage that the CPU requests, because its onboard regulator bypasses the motherboard's regulator.

Power to the regulator is supplied via a power cable which connects to a 4-pin disk drive power connector. This connector has a pass-through plug on it, so that plug can still be used to connect to a drive. The adapter also has a 3-pin connector on it for connection of the CPU fan.

The PL-iP3/T, or PL-iP3 version 3, has another new feature: It allows the new Tualatin-core CPUs to be used, even on motherboards that don't support them. There were 2 main changes to the Tualatins from the Coppermine CPUs that make them incompatible with older boards. First of all, they take a core voltage of 1.475 volts. This core voltage, being on a 0.025 volt boundary, is not supported by older voltage regulators. Most systems with which this FAQ concerns itself support core voltages of 1.45 or 1.50 voltages, which would probably be close enough. However, the onboard motherboard regulators might not be able to respond fast enough to the large jumps in current that the newer CPUs are capable of generating (they can go from near zero current to maximum in a fraction of a microsecond, and the regulator must respond fast enough to prevent the voltage from dropping too low). Also, for the XPS R machines in particular, the CPU might draw too much current for the motherboard's regulator (of course, I said this before about the 850 and 1 GHz Pentium IIIs, and those turned out to work perfectly fine..) The PL-iP3/T's regulator conforms to Intel's latest VRM 8.5 voltage regulator requirements.

There is another change with the Tualatins, however, that is more problematic: Intel changed the signaling voltage used on the CPU's Gunning Transceiver Logic (GTL) bus, which it uses to communicate with the system, from 1.5V to 1.25V. This change makes the CPU incompatible with older motherboards (more specifically, older chipsets like the 440BX). Powerleap got around this problem, however, by adding circuitry to the PL-iP3/T to convert the signaling voltage to allow it to work in older motherboards.

Other than these features (onboard regulator and bus voltage conversion), the PL-iP3/T is much like other slocket adapters such as the Iwill Slocket II. It does not have core voltage selection jumpers, it just sets the voltage based on what the CPU requests. (For overclockers and other hardcore users, Powerleap says there is some way to set the voltage manually, but apparently it requires some rewiring.) It also has an LED to indicate that the adapter is receiving power.

Celeron 1.2 GHz Product Information:

The Celeron 1.2 GHz is based on the previously mentioned Tualatin core. It uses a 0.13 micron manufacturing process, compared to the 0.18 micron process of the Coppermine chips, which allows lower heat production and power consumption at a given clock speed. It runs on a 100 MHz bus, so it will work in BX chipset systems with the PL-iP3 adapter. (It also apparently works in older motherboards which support only 66 MHz bus, but of course it will only be running at 800 MHz in that case.)

The Tualatin CPUs have a new feature called Data Prefetch Logic, which is designed to predict what data in main memory will be needed next, and fetch it into the L2 cache before it is requested. This may help to reduce the performance loss from these chips' 100 and 133 MHz buses, compared to the equivalent 266 MHz bus of the current Athlon chips and the 400 MHz equivalent bus of the Pentium 4 (although the new Athlon XP CPU and Pentium 4 also have similar prefetch logic..)

This Celeron, and other Tualatin chips, use the same Socket 370 connector as the Coppermine FC-PGA chips, but come in a new FC-PGA2 package. The main difference is that there is now a metal slug covering the CPU core. Therefore the heatsink sits on this slug instead of the core itself. The core itself is quite fragile - Intel probably got tired of people chipping the core and wrecking the processor when installing heatsinks on FC-PGA chips. It also helps transfer heat from the CPU to the heatsink more efficiently, Intel says. This is one feature that AMD should copy for its CPUs..

This Celeron CPU has a 256K level 2 cache onboard, the same as the Coppermine Pentium IIIs. However, the Tualatin Pentium IIIs have 512K of L2 cache, so the Celeron still has half as much cache as the Pentium III of the same class. Also, the Tualatin Pentium IIIs run on a 133 MHz bus, compared to the Celeron's 100 MHz bus, which boosts the Pentium III's performance compared to the Celeron - but also renders it incompatible with BX motherboards which support only 100 MHz bus speeds.

The upgrade package I received originally included a Powerleap-brand heatsink and fan unit. Powerleap later sent me the Intel heatsink/fan unit for the Celeron CPU, I put that one on at that time.

Installation & Testing:

I tested this upgrade on my Dell Dimension R-series machine. It originally came with a Pentium II-350 MHz processor, which I later upgraded to a Pentium III-700 FC-PGA using an Iwill Slocket II adapter.

The evaluation PL-iP3/T and 1.2 GHz upgrade package I received from Powerleap came with the CPU already inserted and the heatsink already attached. (Updated info: Powerleap says that they intend to ship the final units preassembled like this as well.) So, basically all installation required was unplugging the previous CPU and slocket adapter, plugging the new adapter and CPU in, and connecting the power cable to a drive power connector.

The XPS R BIOS identifies this CPU as a Pentium Pro 500, same as it does with the previous Coppermine Pentium IIIs.

After installation, the system booted up without incident. I performed some before and after benchmarks to judge the performance improvement of the new CPU. My system's specs are shown below:

Dell Dimension XPS R-series
MSI Starforce 816 (GeForce2 MX) 32MB video card
IBM Deskstar 60GXP 40GB hard drive
Maxtor 91360D8 13.6GB hard drive (original, still installed)
Turtle Beach Montego II sound card
Operating system: Windows XP Home Edition

Here are the results of the SisSoft Sandra 2001te CPU benchmark:

Benchmark Pentium III-700 Celeron 1.2GHz Pentium 4 1.2 GHz Athlon 1.2 GHz Pentium 4 1.6 GHz
Dhrystone ALU (MIPS) 1843 3362 2300 3366 3065
Whetstone FPU (MFLOPS) 939 1605 1466 (SSE2) 1650 1955 (SSE2)

Now there is a significant improvement: The Dhrystone benchmark increases by 82% and the Whetstone benchmark increases by 71%. Furthermore, the CPU is nearly even with an Athlon 1.2 GHz, and even bests the Pentium 4 1.6 GHz in the Dhrystone benchmark! It's quite remarkable that a system whose motherboard and RAM subsystem dates from 1998 can be upgraded to beat out one of Intel's fastest CPUs in 2001, if only on a synthetic benchmark.

Here's the results of Sandra's multimedia CPU benchmark:

Benchmark Pentium III-700 Celeron 1.2GHz Pentium 4 1.2 GHz Athlon 1.2 GHz Pentium 4 1.6 GHz
Integer (it/s) 3788 (SSE) 6516 (SSE) 4725 (SSE2) 6556 (MMX) 6300 (SSE2)
Floating Point (it/s) 4635 (SSE) 8089 (SSE) 5855 (SSE2) 8206 (3DNow) 7804 (SSE2)

Another impressive result - the SSE integer performance improves 72% and the SSE floating point performance increases 75%. Now it pulls past the Pentium 4 1.6 GHz entirely, and comes close to the Athlon 1.2 GHz.

I should note, though, that you shouldn't expect this kind of increase in real applications, because these synthetic CPU benchmarks will hardly stress the memory system at all, and this system is still using old PC100 CAS3 memory. The next benchmark will show that the memory performance doesn't go up much.

Benchmark Pentium III-700 Celeron 1.2GHz PC600 RDRAM PC800 RDRAM PC2100 CL2 DDR
Int ALU/RAM Bandwidth (MB/s) 292 302 1030 1374 644
Float FPU/RAM Bandwidth (MB/s) 324 316 1050 1400 830

Here we can see that the memory bandwidth increases only marginally for the integer benchmark, and actually drops slightly for the floating point for some reason. (Both of these tests were run twice and the second result was used, because the first run produced some swapping to disk that could have affected the results.) In both cases, the memory bandwidth is clearly higher on the newer DDR SDRAM and RDRAM machines. (However, unless all you care about is memory benchmark scores, it's the improvement in overall performance on real applications that matters.)

It could be that the slight drop in memory speed with the newer CPU could be because of the Tualatin's prefetch logic, which may be using up some of the RAM bandwidth to fetch data that it thinks will be needed later. However, this feature will likely improve overall performance, regardless of its impact on the memory benchmark.

These are the 3DMark 2001 scores, before and after:

Before upgrade: 1847
After upgrade: 2078

That's only a 13% increase, which can probably be explained by the fact that so many of 3DMark 2001's benchmarks depend so much on the video card's transform & lighting (T&L) speed and fill-rate, which did not increase with the new CPU. Most games have more dependency on the CPU speed, so you will likely see a bigger increase than that in real games, even with the same video card.

Finally, here is WCPUID's output for the new CPU:

WCPUID thinks the processor is a Pentium III Xeon, for whatever reason.


Overall, I think that this upgrade would be worthwhile for owners of machines like the Dimension V, XPS R, and XPS T machines, especially if they don't already have a fast CPU like a Pentium III-850 or 1 GHz. Powerleap says the price for the package of the PL-iP3/T and Celeron 1.2 GHz will be at or below $170 US, less than the approx. $200 US that people are paying now for the 1 GHz, 100 MHz FSB Slot 1 Pentium III.

In my testing so far, I have not noticed any unusual behaviour or instability, and the performance in a number of games and other applications has noticeably improved (especially in those that use the CPU a lot).

I only have two real complaints with the PL-iP3/T: First is that the Powerleap-supplied CPU fan I originally received was a bit noisy, noticeably more so than the puny fan on the Intel stock heatsink on my old Pentium III-700. Certainly the fan seems effective - during intense gameplay, the CPU heatsink got just slightly warm to the touch.

I later received the Intel heatsink/fan unit, which I installed on the adapter. It fit fine, except that I had to remove the old heatsink support bracket (for the Pentium II CPU heatsink) from the motherboard to provide the required clearance above the motherboard surface. (I have added instructions on the slocket CPU installation page for how to do this. This is only necessary on machines that actually have such a retainer bar, like the R-series systems.) It is much quieter than the Powerleap unit, and seems to provide more than sufficient cooling. It is my understanding that Powerleap will be shipping the Intel fan with the adapters when they are offered for sale. (Update: Some units seem to be shipping with Powerleap heatsink/fan units as well, but I have been told that the current model they are using is considerably quieter than the one I received.)

Another small complaint is that the adapter has a number of capacitors located right next to the CPU socket. These might obstruct and prevent the installation of some larger heatsinks. The Intel stock heatsink for the CPU does fit (barely). To some extent, I imagine that this placement was unavoidable, considering the limited size of the adapter circuit board and the necessity of mounting the capacitors near the CPU to meet the electrical requirements of the CPU, but it's something to keep in mind.

Later Updates:

Some PL-iP3/T units shipped out before Dec. 20, 2001 which have the Powerleap-brand heatsink had a problem with heat transfer from the CPU to the heatsink, because they used a thermal pad on the heatsink which was too small to cover the entire metal slug over the CPU core. Symptoms of this problem can include system lockups after the system has been on for a few minutes, or even failing to finish booting into Windows. To see if your adapter might have this problem, look in through the gap between the CPU base and the heatsink and see if you can see yellow thermal material in between the two - if so, your unit has the tape (if it's white grease material, it should be OK). To fix this, you can:

Powerleap says that all units made starting January 9, 2002 will use the high-performance Arctic Silver 3 thermal grease. Now, there isn't that much difference in heat transfer between different thermal greases (so it's not worth pulling your heatsink off just to put Arctic Silver on), but it does give a little bit more cooling.

A few users have reported that after installing the unit, and making sure it was all the way in the slot, the system failed to boot up (most likely it will come on right away when power is applied, there will be no video display and will not turn off with the power button). This indicates the CPU isn't functioning properly, because the motherboard defaults to "on" when the power is applied, and if the BIOS setting is not to turn on after AC power loss, it will turn itself off, but if the CPU isn't working this won't happen. This is most likely caused by either the CPU or the adapter (most likely the adapter). Powerleap seems pretty good about sending out replacements in this case.

Return to Main Processor Upgrade FAQ Page
Return to Robert Hancock's Home Page