Thinkpads Forum

These processors are all 1066 FSB and obviously require the 1066 mod to work in a T61.

Interesting that the X9100: http://www.cpu-world.com/CPUs/Core_2/In ... 0836M.html

and the T9900: http://www.cpu-world.com/CPUs/Core_2/In ... 836MG.html

have the same standard clock rate and are documented as the same Penryn Core architecture. Yet the T9900 benchmarks considerably better, considering all the similarities, but only in Geekbench, Geekbench 64 and Passmark benchmarks. The others are very close with the X9100 having a very slight advantage in most cases. Oddly, in Geekbench 32 the X9100 shows a slight advantage. Particularly in Geekbench 64 the T9900 has over a 20% advantage, normally I'd say that it was a testing error but straight Geekbench also shows a decent advantage. And the Passmark test shows it as 5% better. Anyone know what's making the difference? I suppose that these are collected from users randomly running the test and the systems and memory timings are not fixed and controlled. Also, some early X9100 systems had DDR2 RAM whereas most later systems would have had DDR3: http://cpuboss.com/cpus/Intel-Core2-X91 ... -Duo-T9900

I also noticed that the T9800 (2.93 GHz): http://www.cpu-world.com/CPUs/Core_2/In ... 776MG.html
and T9900 (3.067GHz): http://www.cpu-world.com/CPUs/Core_2/In ... 836MG.html

had almost the same clock rate, hardly enough difference for Intel to bring out a new processor, so I wondered if they were making process and minor architectural improvements. Perhaps they just ticked the multiplier up from 11 to 11.5 along with the other changes: http://cpuboss.com/cpus/Intel-Core2-Duo ... -Duo-T9800

Anyone one know the details or an article on how the Core design evolved?

The T9900 and X9100 are virtually identical except that the T9100 has an unlocked core so you can run it at any speed you like, however that means there is more likelihood some gamer has tried (perhaps many times) to run it beyond it's capable speeds. It's true you can run chips beyond their clock speed, but go too far and you start to fry transistors or at least become unstable. Those who overclock for sport tend to always go too far then back up and hope there was no perm damage. Benchmark tests are also no very reliable, many seem to measure only single threaded no-ops or non-ops, which don't represent real world function of the chip. For example, when comparing two chips, the T7700 and T8100 many sights say the T7700 wins hands down, but tests using real world conditions the speed was only a nominal difference and when under heavy load or harsh environmental conditions the cooler running T8100 wins hands down the second the T7700 throttles down due to high temps and the T8100 can continue to run at full speed. Speaking of overclocking, all these chips can be run at about 110% of their clock speed even if the core is locked by enabling dual IDA mode. This usually requires a bios mod and a program like Throttlestop who's developers happened to use the T8100 in development because it's believed to be a purposefully underclocked version of the T8300 designed to run at 2.4ghz, so it's 2.1 clock speed seems to run at 110% (2.31ghz) using dual IDA without showing any stress.

The bottom line, if you can find a laptop that was owned by some corporate exec with an X9100 and snag the cpu, go for it, or if you get a T9900 that was never abused you should be able to overclock that a bit too, but these are both hot running chips even at factory clock speed, so proceed with caution, and if you plan to run them at normal clock speed and both chips are healthy, then there should be no difference, but keep in mind that the T9900 cpu chips on auction sites are also risky as they tend to be abused just as any cpu that is the fastest in it's series. The T9800 was more common and generally costs a lot less. You can also consider the X9000 which doesn't require any board mods on a T61, but you still need to watch your temps.

It's also interesting that iMAC also used a similar core2duo chips running at 3.07ghz, but essentially was identical to the T9900. There were a few different sspec codes but the most desireable of them I've tested and seems to behave the same as a T9900 and I got a really great deal on a handfull of them and built some R500 frankenpads which I deemed the worlds fastest Rseries and sold them as Race500 Frankenpad. I sold a few and kept one for myself (which I may sell if anyone is interested), and it was perfect for anyone who needed a lot of cpu power in a relatively inexpensive laptop. I had access to a large number of R500 laptops, some virtually mint condition, but not so much with the cpu chips so had to stop. I wouldn't recommend using one in a modified T61 board unless you have an engineering understanding of any differences in these chips. The are rated for a higher power requirements but as many others have stated, the chips were probably designed with max power requirements supplied by apple computers when the chip was designed for them and they seem to work great in thinkpads designed for 1066fsb, assuming you get the proper sspec code which I cannot recall at the moment, nor the chip number, but if you're seriously interested in knowing I can look it up.

Generally Intel doesn't release info on their chip development but sometimes it leaks or is reconstructed by speculation and experimentation, like the theory I mentioned about the T8100 and it makes sense that when the Penryn chip was developed they had two dies, one for power (t9xx series) with an unheard of 6mb of cache, and one for efficiency and cool running (T8xxx series) and they developed the T8300 that ran nicely at 2.4ghz, but they wanted an entry level chip so they could charge more for the 8300 so they declocked it to 2.1 and reduced the price, and if you aren't trying to squeeze every bit of speed from your computer and love a cool running computer then you might agree that it's one of the best chips ever made and hugely underrated. I liked them so much I bought a couple cases of them brand new NOS in sealed lenovo boxes and still have a few left, but regarding your question, I'm sure they were busy tweaking the design during production and testing the chips at various speeds and designs that run faster were given new part numbers and set to run at higher speeds. With the 1066fsb series they changed the cool/efficient line to "P" series and the performance line retained the T prefix and X for the models with unlocked cores. Quad core was also added with a Q in front of the prefix. The P9700 is a confusing chip because it appears to be built on the performance die so my best guess is they found a way to merge the P series silicone and instruction set with the larger performance die with the larger cache. This is just speculation, but I like the idea, just not sure if there are any drawbacks, and if not, why would they not build all their chips that way? I haven't tried a P9700 yet, so can't offer any first hand opinions.

Sorry if much of this reply is based on speculation, but there isn't any hard facts on a lot of this but it's an interesting topic IMO, so thanks for posting.

Hi TuuS, yes I also find this very interesting having worked in digital and CPU design.

I happen to have an X9100 and a P9700 on the way, both 1066 FSB and I'm wondering
if I can plug them into a T61 Penryn, unmodified motherboard, Middleton BIOS and
expect them to work at the lower 800 FSB? I seem to remember reading that it would
not but has anyone tried it?
Why doesn't it work, BIOS microcode or ...?
Where can I get a BIOS as close as possible to Middleton that will support them? I don't
need quad core support yet.

I want to try this as a first step to what will follow with a PLL pin mod for 1066 FSB.

Next quesion of course is for a link to a clear guide for flashing SPD on the RAM.

Thanks everyone!

Pete B wrote: Why doesn't it work, BIOS microcode or ...?

Correct.

Where can I get a BIOS as close as possible to Middleton that will support them? I don't
need quad core support yet.

MDL forums come to mind.

Found this:
http://forum.notebookreview.com/threads ... 7/page-121
where he points out that if the chipset sees FSB>800 on the BSEL
pins then the system will lock to the lowest clock rate. Not sure if this is for
800 FSB processors, would this happen with a 1066 FSB processor in the Santa
Rosa chipset? I've never looked at the schematics so I don't have a clear
understanding. But what is done in that link is that he points out that the
processor/chipset BSEL pins are connected to the PLL through 1K resistors
so the processor chipset side can be held to 800 FSB and the PLL forced to
1066 by tying one side of the resistor to ground and the other high. This
makes sense to me for 800 FSB CPUs.
A 1066 FSB CPU will tell the chipset to run at 1066 which gets back to the first
problem, but if the BSEL pin is forced (and that is safe for the processor) then
the 1066 FSB CPU BSEL is ignored and the chipset again thinks it is at 800
with the PLL forced to 1066 in which case it might work.
I have to research this more to better understand if this is correct.

Pete B wrote:I have to research this more to better understand if this is correct.

Just read this here, thoroughly: viewtopic.php?f=29&t=110620

rumbero wrote:

Pete B wrote:I have to research this more to better understand if this is correct.

Just read this here, thoroughly: viewtopic.php?f=29&t=110620

I have looked at that a few times, I'll have to read it again and actually try it and
take some notes.
Also, they used an NVIDIA motherboard and I'd like to try Intel but I do understand
that then the Intel graphics is overclocked unless there is a multiplier for it and I
can figure out how to down clock it.

Thanks for the tip.

I have a t61p (close enough) and the P9700 wouldn't run without the FSB 1066 mod. You need to force FSB 1066 (I did the wire mod) for a 1066 FSB CPU to run.

And the modded BIOS I take it?
Which one did you use?