Dell Dimension XPS R Processor Upgrade Information

compiled by Robert Hancock

What can I gain from this upgrade?

Once your system is upgraded to a new processor, it will be about as up to date as any system using that processor at least as far as the CPU, motherboard, and memory performance are concerned. The motherboard and chipset in this system shares a great deal with the XPS T systems

I've upgraded my XPS R350 system to a Pentium III-700. I've listed some details on parts I used and some before and after benchmarks here.

I've received some benchmarks as well as details on the parts used from one upgrader who moved from the original Pentium II-400 MHz processor on his system to a Pentium III-700. Click here to view.

So, how do I go about doing this?

BIOS & Software:

For reference, the list of changes in all Dell BIOS versions for this system is available here.

You will probably want Dell BIOS version A09 or later when upgrading the CPU (actually, earlier versions may work, but I would advise upgrading because there are some possibly important fixes included in later versions anyway). I would recommend you upgrade to at least version A12, as there are a number of potentially important fixes in that version.

A note about upgrading to the A13 BIOS: Your system will probably redetect a bunch of the motherboard resources when your system boots up again after upgrading, don't be alarmed. You may have to reinstall the video and/or sound card drivers later on if they don't reinstall themselves properly. If you don't want to have to worry about this, and don't think you need any of the fixes in A13, version A12 is still available on Dell's FTP site here, and doesn't seem to have these complications.

You should not use the Intel BIOS, as it will not allow you to use anything over 450 MHz.

I've seen one report that when someone attempted to upgrade to a 450 MHz Pentium III processor (the only non-Coppermine Pentium III that this board supports) with the Dell BIOS version A13, it displayed a warning message and refused to boot. Going back to an earlier BIOS version seemed to prevent this. This doesn't seem to be an issue with Coppermine processors, however, for some reason.

Finally, if you have a Sound Blaster Live! sound card, you should upgrade the drivers for it to the latest available version before upgrading the processor, or you may experience system lockups and/or blue-screen errors after upgrading the CPU. Apparently old versions of 3dfx Voodoo3 video card drivers can cause problems with newer processors as well, if you have one of those cards you should get the latest drivers from 3dfx.

The processor & slocket - Pentium III upgrade:

The most common processor upgrade that's been done is to a newer Pentium III CPU. Only the newer Coppermine processors will work, the older Katmai processors will not (except perhaps the 450 MHz one) because they consume more power than this board's CPU voltage regulator can supply. See the main page for info on how to tell which kind a processor is. I believe that upgrades to CPUs up to 750 MHz should be quite safe to perform, upgrades to 800 or 850 MHz may entail some element of risk - see the technical discussion "How are you able to do these upgrades with these systems?" later on for more info on this, especially the summary at the end.

Processors which work at 133 MHz FSB will NOT work on these machines. See the main page for info on how to know which ones these are.

You can use two types of processors with these systems. The first is FC-PGA, this is in a socket format, as opposed to the slot format (SECC) that these machines normally use. Because of this, a slot-to-socket adapter, commonly called a "slocket" is needed to install one of these processors. You can also install a regular Slot 1 CPU like the ones these machines came with originally, see the section "What about using a regular Slot 1 (SECC2) processor instead of a socketed one?" below for more details on this.

This page on Intel's site lists slocket adapters which they have tested and found to meet or not meet basic compatibility requirements for the Pentium III FC-PGA processor.

Here's a summary of what slocket adapters are known to work or not work:

• Iwill Slocket II: Many reported successes, by far the most common slocket used. A few people have said that they had bootup problems, and that one or more of the jumpers on the slocket didn't seem to fit securely on the pins, reseating the jumpers or replacing them with different ones fixed the problem. Note that Intel reports the IWill Slocket II versions 0.9 and 1.0 as failing their tests, however most of them out there now are apparently later versions (1.1 or 1.2) which probably fix whatever the problem was. They list an IWill Slocket III as passing - this would presumably be their voltage-regulating slocket they have hesitated to release. See the main page section on voltage-regulating slockets for more info on this.
• Abit SLOTket III: Confirmed not to work. System would not power up at all. Did pass Intel's tests (version 0.82), what versions the people who reported it not to work were using is not known.
• Epox EP-S1C: Confirmed not to work. System would not power up at all. Did pass Intel's tests (version 1.2), what versions the people who reported it not to work were using is not known.
• AOpen FC-PGA: 1 report that it worked fine, 1 report that it didn't work at all. Failed Intel's tests.
• Asus S370-133: Found to work, although attachment mechanism doesn't fit securely in processor retention clips so may come loose. From pictures it seems like this is made to fit the newer SECC2-style retention clips instead of the SECC ones these systems have. I suppose if you wanted to you could replace the clips with newer SECC2-style ones if you found some that would fit. Passed Intel's tests (version 1.01).
• MSI MS-6905 Master 2.0: Found to work (1 reported upgrade with this slocket). This one, like the Asus, looks like it might not fit the retention clips properly from images I've seen, but reportedly it actually does fit OK. Passed Intel's tests (version 2.0, versions 1.0 and 1.1 did not).
• Soltek SL-02A++: 2 reported successful upgrades. Passed Intel's tests (version 4.5).

The slocket definitely must be designed to work with FC-PGA Coppermine CPUs. Older slockets which support only the older PPGA Celeron chips will NOT work. You should use one that uses a plastic case around it so that it fits properly in the processor retention clips. Some slockets don't have this, so they're more susceptible to coming loose from the processor slot.

Some more slockets which may work but have not been tried to my knowledge:

• Gigabyte GA-6R7Pro / GA-6R7+: The Pro version apparently has voltage selection jumpers, while the + version doesn't (I guess it just uses the automatic setting). GA6R7+ passed Intel's tests (version 1.1).
• PowerLeap PL-iP3: This slocket has an onboard voltage regulator which bypasses the motherboards'. See my notes in "Is there a way I can safely use a processor faster than 750 MHz.."
• Magic-Pro MP-A8++: Passed Intel's tests (version 4.5). May really be the same as the Soltek SL-02A++?

The slocket's voltage and FSB (front-side bus) settings should be set to the automatic position. The jumper that selects between the FC-PGA and PPGA must be set to the FC-PGA position, and the CPU setting must be set to Intel (as opposed to Cyrix). Note that for the Iwill slocket, the instruction page may confusingly label the voltage jumper positions for automatic as "No CPU". This page on their web site has a more accurate description. Also, please note that the jumpers on the IWill are NOT in numerical order 1-9 as they go down, they go 6 to 9 and then 1-5 from top to bottom. Please make sure that you look at the markings next to the jumpers so that you set the proper ones.

The processor & slocket - Celeron Processor Upgrade:

Except as noted, this is basically the same as the Pentium III Coppermine upgrade.

Both PPGA and FC-PGA Celerons should work, although most people have used FC-PGA chips.

The most conservative option with a Celeron is to set the voltage and FSB jumpers on the slocket to automatic, this will cause it (and your system) to run at a 66 MHz FSB speed.

Or, if you are more adventurous, you can use the slocket's FSB selection jumper to force a 100 MHz FSB. This will cause the Celeron to run overclocked, at a speed 50% faster than its rated speed (i.e. a 566 will run at 850 MHz). (Of course this will void the warranty on the processor, + all the usual other disclaimers about overclocking...) To make it run at this speed, you'll likely have to increase the voltage setting on the slocket (i.e. don't use the auto setting).

The default voltage for PPGA Celerons is 2.0 volts. Most FC-PGA Celerons below 633 MHz have a default voltage of 1.50 volts, and most of 633 MHz and above have a default voltage of 1.65 volts. However, FC-PGA Celerons that are the newest C0 stepping (aka stepping 6) all have a default voltage of 1.70 volts. If your Celeron is C0 stepping, it should have one of the following S-codes marked on it: SL4PC or SL4NW (566 MHz), SL4PB or SL4NX (600 MHz), SL4PA or SL4NY (633 MHz), SL4P9 or SL4NZ (667 MHz), SL4P8 or SL4P2 (700 MHz), there may be some others.

For FC-PGA Celerons that normally run at 1.50 volts (the only ones I have detailed accounts of people using) you will likely have to increase the voltage to somewhere around 1.60 to 1.65 volts to make the processor run stably. I recommend starting at about 1.60 and increasing the voltage in 0.05 volt steps until the system boots up properly and runs without crashing. (Use the lowest possible voltage that allows the chip to run properly, as increasing voltage also increases heat production and possibly shortens chip life.) Some processors may not be stable running at 100 MHz FSB until the voltage is increased to 1.80 volts or more, and some may not work at 100 MHz FSB no matter how much the voltage is increased - if that happens, you will have to either run it at 66 MHz FSB speed (i.e. not overclocked) or get a different processor. There are no in-between FSB settings on this motherboard. Note that although some slockets may have a 133 MHz FSB jumper position, this motherboard does not support that speed.

What voltages Celeron CPUs with default voltages of 1.65 or 1.70 volts need to overclock to 100 MHz FSB is not known. However, you should take the higher standard voltage into account when setting an initial testing voltage for overclocking.

Obviously to overclock these processors you need a slocket that has FSB selection jumpers like the Iwill Slocket II. Some slockets don't have these, they can't be used for overclocking.

There have been a number of people who have been able to run a Celeron-566 overclocked to 850 MHz with the processor being quite stable. One person apparently was able to use an older PPGA Celeron-366 and run it at 550 MHz.

It seems that sometimes the SisSoft Sandra utility incorrectly detects the FSB speed when you overclock one of these processors.

I do have concerns about how much power these processors consume when overclocked - obviously Intel's specs don't contain this data! It may or may not be too much for this board to handle safely. See the technical discussion later on for more info on this issue. However, I haven't heard any reports of anyone burning out their motherboard yet..

Whatever option you choose, be sure you have the correct heatsink/fan for your processor, some heatsinks have different versions for PPGA and FC-PGA. For running at 66 MHz FSB (no overclocking) the Intel boxed CPU heatsink/fan should be sufficient. If you are overclocking, however, you may want to use a larger heatsink/fan to ensure the processor stays cool. Again, as I mentioned on the main page you should make sure that it actually fits properly onto the slocket and properly contacts the processor core.

Once you have your CPU and slocket: Follow the installation instructions for "FC-PGA processor installation on Slot 1 board using slocket adapter" on the main page.

After installing the new processor, you should be able to power on your system and enjoy your now faster machine. (Note: Apparently on BIOS version A09 you may have to enter maintenance mode setup by moving the jumper (see Dell manual for details), set the processor speed to 450 MHz, save, move the jumper back to normal and reboot for it to work, though some report that it works alright without using maintenance mode. For later BIOS versions this should not be needed.) The Dell BIOS won't identify your processor properly in the bootup screen, it will display it as a Pentium Pro 500 or something silly like that. However, the processor speed should still be correct. The Dell BIOS doesn't have the ability to turn the Pentium III's processor serial number off, to do this you can use this program from Intel.

If you encounter problems, see the troubleshooting section on the main page.

How are you able to do these upgrades with these systems? (more technical stuff)

The XPS R series systems use an OEM version of the Intel SE440BX motherboard - note this is not to be confused with the later SE440BX-2 motherboard used in the XPS T series systems. With respect to the processors this board officially supports, Intel says that the only Pentium III CPU supported is the 450 MHz version, and states:

"Intel® Pentium III processors that run internally faster than 450 MHz are not supported because the maximum Icc current required is greater than what can be supplied by the motherboard's on-board voltage regulator."

From Intel's Pentium III datasheet for SECC2 processors, it can be seen that the 450 MHz Pentium III draws a maximum of 14.5 amps of current (see p. 26). The 500 MHz Pentium III CPU (no E) draws 16.1 amps. Obviously the maximum current output of this board's voltage regulator (at least according to Intel) lies somewhere between these two numbers.

However, the Coppermine processors are fabricated using a 0.18 micron process, as compared to the older 0.25 micron process used in the older Katmai Pentium IIIs and the Deschutes Pentium IIs these systems are originally equipped with. This means they produce less heat, and more importantly for us, consume less power. A 500 MHz Coppermine Pentium III consumes only 10.0 amps, and a 750 MHz Coppermine processor consumes 15.0 amps, slightly over that of an older 450 MHz processor (see Intel's Pentium III datasheet for PGA370 processors for this information, on p. 24). However, once you get over 750 MHz they all start to exceed the P3-450's current draw more and more, which could cause problems.

Besides current, voltage is another consideration. The Pentium IIs and older Pentium IIIs used a CPU core voltage of 2.0 volts, while the newer Coppermine Pentium IIIs require a voltage of 1.60 or 1.65 volts.

The voltage regulator control chip on my system's motherboard is a Semtech SC1182CS, its data sheet is available on Semtech's web site here. (It's a rectangular chip below the processor slot.) This data indicates that this regulator supports output voltages ranging from 1.30 volts to 3.5 volts, including the 1.60/1.65V voltages needed by Coppermine CPUs (see p. 4). The voltage ID signals that produce these voltages also match those put out by the CPU to request these voltages according to the Intel datasheets. So it would appear that putting out the proper voltage is not a problem for this motherboard. Many other older BX motherboards don't seem to support the proper voltages for Coppermine CPUs, so I guess we can consider ourselves lucky..

The voltage regulator chip is what selects the voltage and ensures it remains at that level, but it doesn't actually do the regulating itself, it controls a power transistor which is what actually controls the power. On my board, that transistor is a Philips PHB21N06LT, its datasheet is available on Philips' web site here. (It's a square-shaped device near the control chip, it has 2 legs attached to the motherboard and one stubby leg that doesn't have anything attached to it.) The important info on this datasheet is the maximum current it can handle, it is given as 19 amps at 25 degrees Celsius and 13 amps at 100 degrees Celsius. This means that 19 amps is the absolute maximum current this board can handle, no way you could ever get more than that, and usually you can't get this much anyway (there's limits from how much heat whatever heatsink they provide for it can carry away).

With the info from these datasheets we can do a bit more number crunching. The following data on processor voltage and current can be obtained from Intel's datasheets for Pentium II, Pentium III SECC2, and Pentium III PGA370 processors. The processor power is calculated using the formula Power = Voltage x Current. The voltage regulator dissipation is calculated by using some of the formulas in the Semtech datasheets (except the "body diode recovery losses" part is not included). These formulas ask for a on-state resistance for the transistor, I used 0.060 ohms as given in the transistor datasheets, although this is quite a bit higher than the values given for other similar transistors in the Semtech datasheets. This doesn't really matter that much though, the relative differences between the numbers should be the same.

(Chart color coding for regulator power dissipation: Green: should be OK, Orange: maybe OK, Red: probably not OK)

(Note: Thermal design power in watts refers only to the amount of heat the processor produces. This is not really related to this discussion, just included for reference.)

From this info, we can see that looking just from a regulator power dissipation (i.e. heat production) standpoint, the P3-750 is still as fast as you can go without exceeding the dissipation of the P3-450 which Intel says is the most it can handle. 800 and 850 aren't that much higher than that, though, however there may be other considerations that deal with current only, rather than the amount of heat the regulator produces. Intel certainly isn't very clear about it. Hey, if any Intel engineers are reading, please enlighten me a little bit on this topic :-)

So basically I'd say that a Coppermine Pentium III up to 750 MHz should work without problems, at least from the info I have.

Quite a few prople have upgraded to an 800 or 850 MHz processor in these machines, and reportedly they do work fine, in the short term at least. However, the long-term reliability with the 800 and 850 MHz processors in these machines is not known. I guess what I'd ask people planning to use one of these processors, "Do you feel lucky?" :-)

Remember, an 850 MHz processor is only about 13% faster than a 750 MHz, at most, and the difference is probably quite a bit less in real applications. You'll likely also pay quite a bit more for it as well.

Apparently Intel is now releasing a 1 GHz processor with a 100 MHz bus speed. While this would likely be technically compatible, the power consumption is considerably higher than the 850 MHz processor, so I would not recommend R-series owners attempt to install this processor, or you could damage the voltage regulator.

Is there a way I can safely use a processor faster than 750 MHz without worrying about all the garbage you talked about above?

There may be. If you use a slocket with an onboard regulator, the limitations of the motherboard regulator basically become irrelevant. See the main page section on this subject for details.

Finally, again note that processors which use a 133 MHz FSB won't work on this board (or at best they'll run at 2/3rds of the rated speed). This includes all processors with a B or EB attached to the speed rating, as well as others like the 667, 733, 866, 933 MHz, 1 GHz and 1.13 GHz processors. Currently the fastest processor with a 100 MHz FSB is 850 MHz, Intel may come out with faster ones in the future.

What about using a regular Slot 1 (SECC2) processor instead of a socketed one?

This is an option. I have received numerous reports of successful upgrades with Slot 1 (SECC2) Coppermine Pentium IIIs.

One concern is that most Pentium III CPUs are in the SECC2 format, while the original processor is the original SECC - the processor retention clips are different, so they use a different way of locking the processor down. If you use the existing clips, the processor won't be properly secured in the slot. It will probably fit in, but you may run the risk of it loosening up or falling out - of course, if your system stays in one place all the time and isn't bumped around it most likely will stay in place..

If you want to be sure the processor is secured down properly, you will need to get some different retention clips that work with SECC2 and will fit the screw setup on this board. Retention clips made for the SEPP Celeron processors should also work with an SECC2 processor. The electronics mail-order company DigiKey carries some retention clips, shown in their catalog here. The universal (URM) style they have doesn't seem like it would fit on these boards, they need a captive nut-type fastener and these use a screw. They do carry a Celeron retention mechanism set that I believe should work, they use captive nut fasteners. The part number is A15108-ND and costs $5.39 US ($8.03 Cdn for Canadian customers including duty & brokerage). They do have a handling fee of $5 (US) for orders under $25. I've dealt with this company before, they seem to have pretty good service. A local computer store may have these available as well.

Another consideration is that Intel is going to be phasing out the SECC2 package this year. Because of this, they seem to be starting to become more expensive and harder to get. It may end up cheaper and easier to use an FC-PGA processor even though you have to use a slocket.

Why can't I just overclock the original processor, or a Pentium III processor?

The only known way of doing any processor overclocking on these machines is running a Celeron processor at 100 MHz FSB as mentioned earlier. All processors made since midway through the Pentium II era (which probably includes most or all of the ones that came in these machines originally) are multiplier-locked, so the CPU will only run at the multiple of the FSB speed that it is designed and tested to run at. The only other way of overclocking a Pentium II or Pentium III would be to increase the FSB speed past 100 MHz.

However, this does not appear to be possible on this board. The clock generator chip on my system's board is a Cypress 48C101-01, its datasheets are available here. (It's a rectangular chip, mounted on the right side of the board in the area of the AGP slot, next to a small silver can which is presumably the clock crystal itself.) This data reveals that this chip's FSB setting is set based solely on the signals received from the processor slot, so it is not changeable through software, and in addition, the ONLY supported FSB speeds are 66 MHz and 100 MHz anyway.

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

Page