Thinkpads Forum

I'm wondering if I can replace the i7-2620M in my x220 Tablet with a i7-2960XM Chip. That would essentially double the cores and keep about the same clock.

As I understand it, because it's a Sandy Bridge system, the Memory controller is on the CPU and would get upgraded with the new chip to support 32GB of RAM instead of the 8GB of RAM currently supported.
The TDP would go up by 20W but I think that's managable.

My question is:
The 2620M uses a Socket 988B rPGA according to CPU-Z.
Does the 2960XM Chip use the same socket?

It's a little confusing over at ark.intel.com because for the 2960XM Chip they say the socket is "FCPGA988".
Where as the socket for the 2620M chip is listed there as "PPGA988"
Here are the links there:
2620M:
http://ark.intel.com/products/52231/Int ... _70-GHz%29

2960XM
http://ark.intel.com/products/53478

Anyone tried this upgrade?

Anyone know why it would or wouldn't work?

Any thoughts on whether or not I could really replace the 2x4GB chips I have now with 2x8GB chips for a total of 16GB RAM when using the 2960XM Chip?

All & all thoughts and links are greatly appreciated!!

Since the CPU is soldered to the motherboard in your tablet changing to another CPU would be difficult. The 2960XM is a G1 chip.

Your ThinkPad supports 2 x 8Gb.

If you need any 8Gb x 1 memory for your ThinkPad send me a PM.

Brad

to support 32GB on current systems, quad-core sandy bridge CPUs require four DIMM sockets total.   since the X220 chassis only has provisions for two DIMM sockets, 16GB would be the max even if you somehow managed to shoehorn a QC processor into your system.

besides, the extra 20W TDP of an i7-2960XM is an absolute deal-breaker in the X220 chassis.   it simply cannot handle that much more heat dissipation at full load.   the current heatsink is already at its limit.

in short, it ain't gonna happen.

the only way you might support 32GB in an X220 is when 16GB SODIMMs are released.

If you want a ~12" laptop with quad core, you purchased the wrong laptop; try this one instead:

http://forum.notebookreview.com/7936035-post26.html

Hey AvalonXIII, that's a very interesting article, especially the gaming at 87 C!!

That's pretty much what I'd like to do here with the 220. However, the most important feature for me is a good touchscreen tablet. CPU and RAM must take 2nd place to a good touch screen. Otherwise yeah, that would have been a good option... except I'll never buy another HP, but that's another story.

I managed to get 16 GB of RAM in the x220 and my computer models at the Univ. fly compared to others. I'm finishing simultions in 50% of the time using xp64bit in a VM that runs entirely from RAM. Only problem is that the timings on the Corsair 8GB Modules are only CAS 9 (9-9...9-9-25). Went from a Ram index score of 7.4 to a 7.1 and could "notice" the slowdown. My pair of 4 GB Chips were Corsairs with CAS 7 (7-7...7-20).

I found a bunch of links to reballing a CPU, but without a machine it seems unreliable and in terms of shorts and jumps/bridges quite dangerous. But hey, maybe I'd find the right pins to short so it becomes an unlocked CPU

Anyone know of a Company in Europe (or the US) that offers professional reballing with a minimum of 90-day guarantee on their work? Any idea if I could replace the BGA type pad (socket) with a PGA type so I could use a CPU with pins?

As far as the RAM in that article (Kingston 1866 SODIMMs), do I understand it correctly that my CPU will just reduce the multiplier if the RAM Modules become faster, so that the CPU Clock doesn't change?

Does anyone have any expierence change the SPD values on RAM chips to make them run at CAS 7 instead of CAS9, or is there maybe a secret page in the Lenovo BIOS for such things??
I miss my good old Pheonix BIOS's

I managed to change the SPD values on some other RAM so that my eeePc saw a pair of Kingstons as though they were Hynix highly overclockable modules, but I'm not sure if I could change the timings using the same procedure (some sort of eeprom type flashing program).

Any tips are greatly appreciated!

torsp32 wrote:Went from a Ram index score of 7.4 to a 7.1 and could "notice" the slowdown.

are you sure this wasn't just a placebo effect?   4% is a margin of error, not a ground breaking delta.   WEI numbers can fluctuate +/- 0.2 depending on conditions.   it's not a reliable benchmark by any means.   stream is the standard for memory benching and a much more reliable way to discern the true differences between memory setups.

Any idea if I could replace the BGA type pad (socket) with a PGA type so I could use a CPU with pins?

the i7-2960XM is FCBGA988.   the X220 uses BGA1023.   the X220 planar cannot accept a 988 socket.

given your needs, you might consider purchasing a proper workstation rather than trying to use a business-class ultraportable for a role it was never intended to have.   a quad core in an X220 would be nice but there's unfortunately no way you'll make it happen.

torsp32 wrote:Hey AvalonXIII, that's a very interesting article, especially the gaming at 87 C!!

That's pretty much what I'd like to do here with the 220. However, the most important feature for me is a good touchscreen tablet. CPU and RAM must take 2nd place to a good touch screen. Otherwise yeah, that would have been a good option... except I'll never buy another HP, but that's another story.

I managed to get 16 GB of RAM in the x220 and my computer models at the Univ. fly compared to others. I'm finishing simultions in 50% of the time using xp64bit in a VM that runs entirely from RAM. Only problem is that the timings on the Corsair 8GB Modules are only CAS 9 (9-9...9-9-25). Went from a Ram index score of 7.4 to a 7.1 and could "notice" the slowdown. My pair of 4 GB Chips were Corsairs with CAS 7 (7-7...7-20).

I found a bunch of links to reballing a CPU, but without a machine it seems unreliable and in terms of shorts and jumps/bridges quite dangerous. But hey, maybe I'd find the right pins to short so it becomes an unlocked CPU

Anyone know of a Company in Europe (or the US) that offers professional reballing with a minimum of 90-day guarantee on their work? Any idea if I could replace the BGA type pad (socket) with a PGA type so I could use a CPU with pins?

As far as the RAM in that article (Kingston 1866 SODIMMs), do I understand it correctly that my CPU will just reduce the multiplier if the RAM Modules become faster, so that the CPU Clock doesn't change?

Does anyone have any expierence change the SPD values on RAM chips to make them run at CAS 7 instead of CAS9, or is there maybe a secret page in the Lenovo BIOS for such things??
I miss my good old Pheonix BIOS's

I managed to change the SPD values on some other RAM so that my eeePc saw a pair of Kingstons as though they were Hynix highly overclockable modules, but I'm not sure if I could change the timings using the same procedure (some sort of eeprom type flashing program).

Any tips are greatly appreciated!

Our needs are similar! I have both a W520 fully loaded (32gb of memory, 1000GB SSD, and 2920XM processor) and I also have a Thinkpad X220T with 16gb of memory + 500GB SSD!

Both machines are very response. When I am in the office, I prefer the W520. When I am on the road, I most definitely prefer the X220T! I am hoping sometime in the next couple of years that Lenovo will create a 13.3" tablet and put a quad-processor in it (then I will only have one laptop)!

Merry Christmas!

Shane

P.S. The memory I used in my X220T was only $139 for 16gb!

http://www.amazon.com/dp/B00698WP50/


Mod note: Link edited. Please do not post referral codes in Amazon links. Thank you.

As far as the placebo effect goes, I actually noticed the slowdown right away. It was only then later when I was checking some system specs that windows complained about not having a current performance index because I changed the ram. So I ran the test and saw it drop. Made sense. But yeah, you're right, I haven't used any proper testing yet, just subjective user experience. Might also have been some software issues. I'll benchmark it properly later. Thanks for the link.

As far as the socket configuration goes, it seems like everyone says something different.
CPU-z shows:
Socket 988B rPGA

the ark.intel entry for the 2620m has the following listed:
FCBGA1023, PPGA988

Can you post a link to your source showing that it is indeed the FCBGA1023 (yes, I realize it's quite likely what you say, but I'd like to see it in print from a reliable source).

As far as roles go, I try to seek out the best combination of hardware in the market and improve this combination to make a proper machine. I understand that the configurations are adjusted to match marketing segments and increase sales. However for enthusiasts and others with high demands and expectations, such configurations always fall short.

However the market for ultra high performance combined with extraordinary design is likely not very profitable. If it were, we wouldn't see things like locked CPU multipliers and also likely much fewer BGA configurations. In short, I'm just trying to reduce the amount of compromise I have to accept in order to have a high performance tablet. Sure, I could buy a custom Clevo based system with an unlocked extreme series CPU to use for demanding tasks, but it would likely be a 17" Screen with no touch or inking capabilities. I also don't care to lug such a beast around on the Metro. It would be great if we could get the best features all in one.

Thanks for the link. I saw that memory and considered buying it, but I didn't trust the brand/mfg. Do you know who makes the chips on those ram sticks?

Merry Christmas Everyone!!

torsp32 wrote:Thanks for the link. I saw that memory and considered buying it, but I didn't trust the brand/mfg. Do you know who makes the chips on those ram sticks?

Merry Christmas Everyone!!

I didn't even look to be honest! I figured if they didn't work, Amazon would have my back.

My memory speed is unchanged at 7.6 so I have been running with the memory for a couple of days now. I run Oracle's Virtual Box on my X220T and so I know the memory is being accessed!

Merry Christmas!

Shane

torsp32 wrote:the ark.intel entry for the 2620m has the following listed:
FCBGA1023, PPGA988

FCBGA is permanently soldered, PPGA is pinned and socketed.   you won't find a pinned/socketed CPU in the X220 and can't install a CPU with 1023 leads on a board built for 988.   since the i7-2960XM is PPGA988 your answer resides in the socket.

torsp32 wrote:Can you post a link to your source showing that it is indeed the FCBGA1023 (yes, I realize it's quite likely what you say, but I'd like to see it in print from a reliable source).

all of the X220 models use the same CPU grid array across planars.   they only differ in the CPU soldered and features enabled.

fwiw, i have an i5-2537M (FCBGA1023) in my current system, had an i5-2520M (FCBGA1023) in my last one, and one of my family members has a celeron 847 (FCBGA1023) in hers.   you won't find an X220 with any other grid array.   you're welcome to disassemble your own system and look for yourself.   that's all the documentation you'll need.   any internal documentation stating the above isn't publicly available to my knowledge.

Thanks for the reply. I'm not a big fan of taking my laptop apart just to read a chip number or check a config, especially when I have to mess with the cooling of something (saves thermal paste and a bit of cleanup). I have noticed in the past few days with multithreaded video encoding or 7zip compressing from my ramdrive, I can get the cpu up to about 86 C before the cooling can hold the temperature, so I have a feeling the upgrade to a faster chip would require a lot of re-engineering.

So... I think I'll wait on the project for a bit. Thanks everyone for the input.

Would also be interested to know if there is a laptop in existence that has basically the same specs as the x220 tablet, but includes USB 3 and a quad core with at least a clock of 2.5 GHz. Esata and/or thunderbolt (light peak) would also be nice, but I doubt that'll happen. Also mSata with Sata III would be nice... but again, maybe I'm dreaming of 2013...

either keep your X220T and enjoy it or wait until ivy bridge systems are released this year.   there's no sense in trying to build the ultimate sandy bridge setup at this point in time unless you absolutely need something today and cannot wait.

either that or buy a workstation.   that's what they're for.

when intel goes to a 22nm die shrink in 2013 you'll be upgrading again anyway.

When you see a place like provantage selling i5-2537's, who is their target customer? Is it possible to successfully r&r surface mounted cpu's outside a manufacturing facility? (not that I plan on attempting it)

vinuneuro wrote:When you see a place like provantage selling i5-2537's, who is their target customer? Is it possible to successfully r&r surface mounted cpu's outside a manufacturing facility? (not that I plan on attempting it)

Do a search in Google Videos for CPU Reballing. Lots of interesting videos out there. It can be done by hand, but I find it quite unlikely that one can really get about 1000 good contacts without the help of a maching to make things uniform all over. I thought about trying it, but I'm a novice at soldering and could envision about 8 hours of headache only to have it not boot (or cook?) due to a short. Imaging with changing temperatures from say 15 C to 85 C and having a ball disconnect, then what?

It could be a fun project if you have a lot of patience, and the videos will show you how and what basic equipment you need, but I think I'd rather pay the pros to deal with the headache and try to get at least a 90 day warranty on their work or some sort of insurance in case they damage stuff or it dies shortly after.

Someday, when I have the time and money, I'll re-engineer the cooling and the socket on a tablet and have my dream machine... Let me know if you get there before I do.