Author: Corry S
Editor: Howard Ha
Publish Date: Tuesday, March 16th, 2004
Originally Published on Neoseeker (http://www.neoseeker.com)
Article Link: http://www.neoseeker.com/Articles/Hardware/s/XP-M/
Copyright Neo Era Media, Inc. - please do not redistribute or use for commercial purposes.
The overclocking world was dealt a blow late last year when AMD began locking the multiplier of its top selling chip, the 2500+ Barton. While it is possible to still overclock the trusty Barton core by way of the front side bus it does have limits. An OC of 2.2 on a 2500+, which is equal to 3200+ in AMD terms, is still attainable at 11x200Mhz as long you have a decent board and RAM. However those that like to play with the multiplier to squeeze the juice harder have to look elsewhere now or look at modding the actual chip.
Enter AMD's XP-M processor AKA the Mobile Barton; the Mobile Barton, which is made for the laptop market, has been around for a while now but is starting to get wide range of attention in overclocking circles. Computer forums around the net are abuzz about what a great overclocker the 2500+ Mobile Barton is. Results range from easy overclocks of 2.4 Ghz to tweaked highs of 2.7 on air cooling only! Sounds incredible and when you factor in that the Mobile Barton is on average only $20-$25 dollars higher in price that a Desktop Barton, you'll find yourself asking "how does this chip do it?" To answer this question we'll take a look at a 2500+ Desktop Barton* and a 2500+ Mobile Barton and see why the two chips are different and how that difference allows us to exploit the overclocking potential of the Mobile Barton. We'll also install both chips and see what the end overclocking results are.
*Our 2500+ XP test chip was purchased prior to October and is unlocked
When you look at these two processors side by side you'll be hard pressed to see any difference and with good reason. Both chips have the same .13 micron architecture and core; essentially they are cut from the same cloth, well same silicon wafer to be exact. The difference comes when the chip is assembled and tested; chips that become Mobile Bartons pass a voltage test showing they can run solid at 1.45V and chips that pass at 1.65V become Desktop Barton. The Mobile Bartons then are set to a default of 133 Mhz rather than the standard 166 Mhz of the Barton. AMD then designates the chip a XP-M with PowerNow technology. New laptop motherboards with PowerNow settings are able to adjust the voltage and clock speed of the CPU in order save battery power and lower heat output as CPU power requirements fluctuate. In order to accomplish this, the boards need to be able to adjust the multiplier on the chip up and down, which means the XP-M has to always be unlocked; we all know this is fantastic right?
Which is which? Hard to tell isn't it, left is the new XP-M
Now take this new gem and put it inside a desktop computer we and we have a unlocked Barton running at 1.45V rather than 1.65V. I know some of you are saying to yourselves "yeah but only running at a total of 266Mhz " which of course is bellow the 333Mhz we are accustomed to seeing in the standard XP model. What we need to take into consideration is the ability of the Barton core to be pushed past those limits. Many overclockers know full well that reaching 400 Mhz is easy with the Barton core especially when running motherboard with a chipset such as acclaimed Nvidia Nforce2. Pushing the envelope on an XP-M gets easier because as you increase the Front Side Bus and increase the unlocked multiplier you are required to add more power to the CPU by increasing the Vcore; since the default Vcore of the X-MP is 1.45V you have much more room to play with. Now that we have an understanding of the potential, we'll test the theory by running stock tests and then overclocking the XP-M and XP and comparing results.
Tests done in Sisoft Sandra 2004 indicate that both chips when overclocked, all things being equal, post very similar results. Given this we'll dispense with software testing and focus our tests on voltage, heat and maximum raw overclock. We'll use an Epox 8RDA+ with an Nforce2 chipset, 3200 dual channel Mushkin RAM and a Thermalright SLK700 heatsink with a TMD fan.
A closer look at the stepping codes
Thanks to Dave N and his new camera for providing this nice macro of our steppings
First up are the Voltage tests. As we mentioned earlier the XP-M (indicated in green on the chart) starts at a Vcore of 1.45. When we apply the common overclock of 11x200 for 2.2 you can see we are able to run at 1.60 on the XP-M but need to run 1.70 to remain stable* on the XP. Next we ramp up to 2.4 gig at 12 x 200, now the spread widens and we need a Vcore 1.925 compared to 1.75 on the XP-M; the Mobile Barton is starting to show it's worth. The final test here in comparison ends at 2.5 gig because that is all I was willing to push the XP to. I'm sure you'll understand why when you see in order to get 2.5 gig (12.5 x 200) I had to really stretch out the Vcore to a scary 2.15 while still on air cooling. Notice the XP-M comes in nicely at 1.875, meaning there is still some head room and we'll get to that later when we push into areas where the XP dare not tread.
*Sisoft Sandra and Super Pi and Prime95 were run to check for stability in all tests.
We know temperature is a major issue when overclocking so lets have a look at the chart below and compare results in celsius as we pushed the Vcore. As you can see at stock both chips come in at respectable temperatures *32c for the XP and 28C for the XP-M. Even in the next set 2.2 the temperatures don't vary widely but when we go 2.4 the XP-M's ability to use a lower Vcore shows value in keeping the temp under 40C and when reaching to the limit of the XP voltage wise the 47C idle was a full 7C above the XP-M again allowing more head room for the XP-M.
*all temps are idle measured with the Epox under CPU thermal sensor
Now as mentioned before we were limited in direct comparison to 2.5 gig. However, it was evident that there was more room to play with the XP-M so we pushed the envelope to see where we could go in terms of maximum overclock. After many configurations and testing we managed to achive a overclock of 2.6 using a mulitplier of 12.5 X 208 and a Vcore of 1.925, the temperature we reached was 43C at idle.
I need to say that I don't feel the chip limited out at 2.6 but feel that our test kit limited us. The Epox board being a good performer on its own was somewhat limited when combined with the 3200 Mushkin RAM. This RAM is good quality but I had hoped to have some 3700 or 4000 DDR in order to push the FSB but as of the test date it had yet to arrive. Cooling also factors in, while the SLK700 is good with a TMD fan, another high peformance fan would prove better. Also one has to consider where you might go with water cooling factored in. However, all things considered, we still managed 2.6 gig which fully demonstrates the potential of the XP-M. Given this, I have no doubts about the claims circulating that 2.7 gig on air alone is in fact attainable under the right circumstances.
The engineers at AMD might not have planned it this way, but they've put out the best overclocking work horse since the famed 1700+ Thoroughbred
A star is born and overclockers can rejoice again.
Please do not redistribute or use this article in whole, or in part, for commercial purposes.