x61 pinmod questions

[vavet]

I want to try the pinmod on my spare x61 with t7300 (2.0 Ghz). I did some reading including the famous German guide
(http://thinkwiki.de/Mehr_Power_f%C3%BCr ... nkl_Tablet), but as I don't speak German and had to use chrome translation
some aspects stay unclear to me.

So hear what I learned (correct me if I am wrong please)

1. middleton bios enables dual IDA which I can activate with RMClock
2. I need to cut the pin 7 of the clock to get 266Mhz instead of 200Mhz which tuns 2.0Ghz to 2.66Ghz without dual IDA and 2.93 with Dual IDA
3. I need to use a tool RWeverything to rewrite EEPROM chip.

Now, the questions I still have.
1. I am currently using pc2-5300 ram. As I understand it corresponds to 533MHz, right? will it work or I have to buy different memory like pc2-6400 or pc2-8500?
(my memory is crucial 2x4Gb) As I understand the rewriting of EEPROM is to make the ram to be used at 712MHz instead of 880 after the pinmod, right?

2. regarding heat. If I do the pinmod but not the dual IDA will it help to keep the heat closer to the normal? My impression after some reading that dual IDA is the main cause for heat increase
Is this true?

My aim is not the highest possible performance, but a balance between reliability, heat and performance. Any hints?

Just want to say upfront that:
- I did searched and read the threads on this
- I know that it is easier to buy a new and faster machine
- I am ready to risk this machine for the sake of experiment

Thanks in advance

[axur-delmeria]

I haven't OC'd an X61 yet, but at one point I had a spare X61 Tablet board with L7500 CPU and explored the possibility of overclocking it. My only OC experience is with an R52/T42 Frankenpad.

I am currently using pc2-5300 ram. As I understand it corresponds to 533MHz, right?

No, it's 667MHz. The the last set of digits correspond to the approximate transfer rate of the RAM in megabyte/s.

It's calculated like this: (memory clock rate) * (memory bus width in bytes)

DIMMs are 64 bits wide (since the SDRAM days), so that's 8 bytes.
So, 667MHz x 8 bytes = 5336 MB/s

As I understand the rewriting of EEPROM is to make the ram to be used at 712MHz instead of 880 after the pinmod, right?

Simply put, yes. You need DDR2-800 RAM, then rewrite its EEPROM to do two things:
1. remove the 667 and 800 speeds
2. change the 533MHz timing profile (CAS Latency, RAS-to-CAS, Precharge, etc.) to match that of the 800MHz.

If DDR2-1066 SODIMMs exist, the procedure is similar, except both the 533 and 667 timing profiles are to be changed.

My aim is not the highest possible performance, but a balance between reliability, heat and performance. Any hints?

As long as your laptop is able to finish loading Windows without crashing, you can manage the CPU speed with software like RMClock or IBM_ECW.

IBM_ECW can also change the speed of the GMA X3100 graphics. by default it's set to the second lowest speed. Setting it to the lowest speed may improve the IGP's stability.

[vavet]

So ram upgrade is recommended, but if SPD is rewritten then there is a chance that stock 667 will work as well (depending on the particular producer of ram)
Am I right?

[axur-delmeria]

but if SPD is rewritten then there is a chance that stock 667 will work as well

Chance? Yes there is a chance. Just loosen the timings (base it from a DDR2-800 profile)

It really depends on how much the RAM chips like running at 712MHz.

Don't hold your breath though.

[vavet]

this might be a silly question.
Does the SPD rewrite modify the ram bar itself ? If so, Then obviously I need to do that for all ram bars that I plan to use separately right?

[axur-delmeria]

Does the SPD rewrite modify the ram bar itself ?

Yes. Each stick of RAM has a little EEPROM chip that contains data about its capacity, brand/model, serial number, date of manufacture, timing profiles, etc. See http://en.wikipedia.org/wiki/Serial_presence_detect

Changing this data carries significant risk, especially since a mistake can render the entire stick unusable, unless you have an EEPROM programmer and can connect the necessary pins, like this picture of a home-made SPD reader/writer:



The picture comes from this thread: http://forum.thinkpads.com/viewtopic.ph ... w=previous.

Then obviously I need to do that for all ram bars that I plan to use separately right?

Yes.

[wileE]

Then obviously I need to do that for all ram bars that I plan to use separately right?

No. You only need to rewrite on one of the RAM. Put that one in the primary RAM slot and the other one will use the same timings.

What you do need is RAM that lets you rewrite the SPD. Not all do that.

Some RAM work without the rewrite. I have 3 pairs of Kingston Value Ram which work just fine.
But test the RAM you plan to use with memtest, at least overnight.

[axur-delmeria]

No. You only need to rewrite on one of the RAM. Put that one in the primary RAM slot and the other one will use the same timings

I see how this would work. Since the modified RAM has the worst/slowest timing profile, the memory controller chooses it.

[vavet]

I rewrote a 2gb samsung bar, seems ok, no error messages. Did the pinmod. Now I have 2.13Ghz on an L7500 without dual ida and 2.4Ghz with dual ida, cool!

No. You only need to rewrite on one of the RAM. Put that one in the primary RAM slot and the other one will use the same timings.

What you do need is RAM that lets you rewrite the SPD. Not all do that.

Some RAM work without the rewrite. I have 3 pairs of Kingston Value Ram which work just fine.

So if I rewrite one of my 4gb corsair bars. I can use them both in normal and in modified manner depending on the positions?

But test the RAM you plan to use with memtest, at least overnight.

What soft would you reccomend for ram test?

[axur-delmeria]

memtest86+ http://www.memtest.org/ and/or memtest86 http://www.memtest86.com/

If you're wondering why their names are so similar, memtest86 was the original open-source program, but development stagnated at V3.0, and was forked to memtest86+.

Development of the original eventually resumed. V4.x was released with new features, and was ultimately sold to PassMark, who rewrote the program for V5.0 and the software licence became proprietary.

You can also try the Ultimate Boot CD: http://www.ultimatebootcd.com/ which contains an assortment of diagnostic and repair software, both memtest86 programs included.

[wileE]

So if I rewrite one of my 4gb corsair bars. I can use them both in normal and in modified manner depending on the positions?

No. The RAM with the rewrite will not miraculously change back if you put it into the secondary slot.

I mentioned memtest because if your machine with the pinmod crashes, it can happen due to errors which never show with the normal timings.
And memtest is included in the boot menu of any decent live linux.

[vavet]

Did the test overnight - 0 errors!

I searched a bit on L7700, T8300 and T9300 pinmod experiences. The results are controversial somewhat.
What was your experiences?

[flyingfishfinger]

I recently did it on a T8300, but I didn't to a load test. I got to 2.92Ghz without overvolting the CPU (I plan to make a provision for that next time I take the heatsink off, since the voltage regulator is partially obscured by it. That'll let me get to 3.2Ghz+). I have 8GB of Elpida RAM, I reflashed the timings.

R

[vavet]

I recently did it on a T8300, but I didn't to a load test. I got to 2.92Ghz without overvolting the CPU (I plan to make a provision for that next time I take the heatsink off, since the voltage regulator is partially obscured by it. That'll let me get to 3.2Ghz+). I have 8GB of Elpida RAM, I reflashed the timings.

R

thanks for the info, few questions:

1. could you please give more information on heatsink and voltage regulator issue? I didn't know about this.
2. what about heat, did you notice any change in heat?
3. did you change the timings to the same parameters described in the german writeup?