Thinkpads Forum

Posted: **Fri Aug 24, 2007 4:24 am**

danny_isr wrote:just went back to my T43 , well yeah there is a temp difference
feels like burning hot compare to the 61

Which one is hotter? 61 or 43?

Posted: **Fri Aug 24, 2007 7:36 am**

of course the T43 was the hotter.
the T61p idles here at 36C while the T43 at around 60C (under volted and power play at minimum)

how can i see what's the GPU and HD temp. NHC doesnt seems to work fully under this machine or VISTA. i get only CPU temp and cannot under volt either ...

I don't see the same leakage (it really depends on the angle you took that picture too), where can i check the Monitor manufacture ?

Posted: **Fri Aug 24, 2007 10:06 am**

which machine you got ?

I got the 14" t61p. But both the 15" and the 14 inch run at the same speed! The 15" just has an additional memory chip!

Posted: **Fri Aug 24, 2007 12:02 pm**

sullivan18 wrote:It would be worth it for one of you with the 14" t61p w/ the 570m to install the GeForce drivers from laptopvideo2go.com and run a gaming benchmark. You have many people at notebookreview.com sitting on the edge of their seats with this discussion.

I installed the newest drivers (16250) from that page and I gained about 30 points in 3Dmark 06 and I cant overclock the gpu anymore (don't know why...)

Posted: **Fri Aug 24, 2007 12:25 pm**

anyone knows where i can order 4:3 T61p in UK?

Posted: **Fri Aug 24, 2007 12:34 pm**

danny_isr wrote:of course the T43 was the hotter.
the T61p idles here at 36C while the T43 at around 60C (under volted and power play at minimum)

how can i see what's the GPU and HD temp. NHC doesnt seems to work fully under this machine or VISTA. i get only CPU temp and cannot under volt either ...

I don't see the same leakage (it really depends on the angle you took that picture too), where can i check the Monitor manufacture ?

I found the best way to check monitor manufacurer is to download PC Wizard 2007, then search google for the model number of the LCD they give you.

Also, if you have a cam, can you put a black screen saver on and take a pic? I didn't even really see this bleeding until I took the pic, the picture makes it really stand out.

Posted: **Fri Aug 24, 2007 12:44 pm**

sullivan18 wrote:It would be worth it for one of you with the 14" t61p w/ the 570m to install the GeForce drivers from laptopvideo2go.com and run a gaming benchmark. You have many people at notebookreview.com sitting on the edge of their seats with this discussion.

The 128MB FX570M is already benchmarked.
Check this thread:
http://forum.thinkpads.com/viewtopic.php?t=48123
It gets 2300 in 3DMark06, which is 1000 points less than the 256MB FX570M, bringing its performance very close to the NVS 140M.

Posted: **Fri Aug 24, 2007 4:53 pm**

nofuture wrote:anyone knows where i can order 4:3 T61p in UK?

We don't seem to have that option here in the UK. And it seems to me that we always get these high specs much later than folks in the US. Perhaps any tech export restrictions?

PPB

Posted: **Sat Aug 25, 2007 3:41 pm**

XIII wrote:
sullivan18 wrote:It would be worth it for one of you with the 14" t61p w/ the 570m to install the GeForce drivers from laptopvideo2go.com and run a gaming benchmark. You have many people at notebookreview.com sitting on the edge of their seats with this discussion.
The 128MB FX570M is already benchmarked.
Check this thread:
http://forum.thinkpads.com/viewtopic.php?t=48123
It gets 2300 in 3DMark06, which is 1000 points less than the 256MB FX570M, bringing its performance very close to the NVS 140M.

If you'll read the post you quoted, you'll see that the thread you linked meets neither of the two requests in the post. It does not use gaming drivers from laptopvideo2go, and it does not have gaming benchmarks, only the synthetic 3dmark06, which as any gamer will tell you is not an indication of real life performance, especially in this case. 3dMarkk06 is highly optimized for cards with 256mb of dedicated VRAM, so any 128mb cards will always take a hit. And you can't compare 3dmark06 scores from this lappy to the 128mb MBP (the only 128-bit 128mb G84 chip that I know of) because the MBP runs 3dmark06 by default at 1280x800 and this lappy runs it at 1280x1024, so this card is being pushed over 20% harder in the test than the MBP's GPU is.

So, to anyone who's received this laptop, please could you install Geforce drivers from laptopvideo2go, and report on real world gaming performance. Any modern 3d game would do, even something like a CS:S stress test, or some FRAPS data from FEAR, Oblivion, Bioshock, or any other modern 3d game. Also, could we get some data on the 3dmark05 scores, because as synthetic benchmarks go, that one would be the most helpful (being optimized for 128mb of dedicated VRAM). Thank you to anyone who can help those of us out who are looking for a good 14" gaming laptop.

Posted: **Sat Aug 25, 2007 5:47 pm**

Well I have the newest driver from laptopvideo2go, installed and Half Live 2 runs absolutely smooth on everything on High and 1280 x 1024. I'm to lazy for fraps n stuff...

3DMark 05 gets about 5700 points on my t61p (2.2Ghz 2 Gb Ram)...
With over clocking you could probably push it up to 7000 points...

Hope that helps...

Posted: **Sat Aug 25, 2007 6:46 pm**

danny_isr wrote:it does make sense to me, if it's 128bit it needs half the amount of
clocks. to have the same performance of the 64bit with twice the clock rate..

So resolution wise you have more, but performance wise they should be similiar....

at least that's what i can think of.

Code: Select all

$0100000000 Graphics core : G84 revision A2 (32sp) 
$0100000001 Hardwired ID : 040c (ROM strapped to 040c) 
$0100000002 Memory bus : 128-bit 
$0100000003 Memory type : DDR3 (RAM configuration 00) 
$0100000004 Memory amount : 262144KB 
$0100000100 Core clock domain 0 : 297.000MHz 
$0100000101 Core clock domain 1 : 594.000MHz 
$0100000006 Memory clock : 300.856MHz (601.712MHz effective) 
$0100000007 Reference clock : 27.000MHz

This leads to 9,375 GB/s

Code: Select all

$0100000000 Graphics core : G84 revision A2 (32sp) 
$0100000001 Hardwired ID : 040c (ROM strapped to 040c) 
$0100000002 Memory bus : 64-bit 
$0100000003 Memory type : DDR3 (RAM configuration 00) 
$0100000004 Memory amount : 131072KB 
$0100000100 Core clock domain 0 : 513.000MHz 
$0100000101 Core clock domain 1 : 1026.000MHz 
$0100000006 Memory clock : 702.000MHz (1404.000MHz effective) 
$0100000007 Reference clock : 27.000MHz

Here we have 10,93 GB/s

Even DDR2-667 in Dual Channel offer the same bandwidth. All mobile GPU suffer form this lake of bandwidth (in compare with the desktop part).

3D Mark 2005 data base :
http://www.tt-hardware.com/img/news5/news260807_1.gif

Posted: **Sat Aug 25, 2007 7:05 pm**

All mobile GPU suffer form this lake of bandwidth (in compare with the desktop part).

In comparison with desktop parts, maybe so, but something like the GDDR3 8600M-GS is not particularly bandwith starved (22Gb/s for 16 stream processors). However this 570m is incredibly bandwith starved (10 GB/s for 32 stream processors). However it still appears to perform similarly to a 256mb DDR2 8600M-GT for lower resolutions, and considering the battery life an weight on this notebook, it's still one of the most portable gaming notebooks available.

Posted: **Sat Aug 25, 2007 7:06 pm**

i have the Doom3 Demo installed. but i dont know how to get the frame rate. if anyone knows how to check the frame rate that would help...

Posted: **Sat Aug 25, 2007 7:19 pm**

danny_isr wrote:i have the Doom3 Demo installed. but i dont know how to get the frame rate. if anyone knows how to check the frame rate that would help...

I believe there's a built-in benchmark, try this:

First, open the console with Ctrl – Alt - ~
Once you're in the console, simply type timedemo demo1 to run the benchmark. There can be significant disk I/O during the first run, so you should run it twice every time you test, ignoring the first result. You'll often see visible stuttering during the first test run, but in most cases, that's caused by data paging in from disk. On the second run, you should see the frame rate smooth out quite a bit. At the end of the demo, a screen will appear showing the number of frames rendered, the time it took, and the average frames per second.

I would try that benchmark once at all high settings and native resolution, and once at the scanned optimal settings. If that doesn't work, you can always just download FRAPS and use it to record average FPS.

Posted: **Sat Aug 25, 2007 7:40 pm**

Odin243 wrote:
All mobile GPU suffer form this lake of bandwidth (in compare with the desktop part).
In comparison with desktop parts, maybe so, but something like the GDDR3 8600M-GS is not particularly bandwith starved (22Gb/s for 16 stream processors). However this 570m is incredibly bandwith starved (10 GB/s for 32 stream processors). However it still appears to perform similarly to a 256mb DDR2 8600M-GT for lower resolutions, and considering the battery life an weight on this notebook, it's still one of the most portable gaming notebooks available.

I agree. It's not possible to have GeForce 8800 with a laptop.

If you compare with previous T60p, memory bandwidth already hit about 10 GB/s with FireGL V5200. I noticed that graphic memory clock increase most time far less then the GPU power itself. It's beacause mobile GPU are based on mid and entry level desktop GPU. All GeForce Serie 8 have good to excellent performances. NVIDIA's DirectX 10 architecture is very accomplished and efficient.

Posted: **Thu Sep 13, 2007 1:47 pm**

Pascal_TTH wrote: This leeds to 9,375 GB/s
Here we have 9,375 GB/s
Even DDR2-667 in Dual Channel offer the same bandwidth. All mobile GPU suffer form this lake of bandwidth (in compare with the desktop part).
3D Mark 2005 data base :
http://www.tt-hardware.com/img/news5/news260807_1.gif

It seems that the official nvidia document claims that the 570M has up to 22GB/s memory bandwidth with 128bit memory bus:
http://www.nvidia.com/page/quadrofx_go.html
Therefore I think the dump for the 256MB version is not correct. I would guess that the memory for both the 128MB and 256MB version are running at the same frequency, i.e. the 128MB ones has ~9GB/s bandwidth while the 256MB ones has ~19GB/s (close to the nvidia spec), which will explain the performance difference we are seeing right now. (I can not think of any other reason that will explain the in game performance difference if both card has the same chip and same memory bandwidth. The difference maybe large in high resolution, but it seems the performance hit also exists under 1024x768, which can't be explained if the only difference is just memory size).
If this is true, then I would blame Lenovo. The Nvidia document clearly states that the 570M has 128bit memory bus. Is it really that difficult to reduce the vram to 128MB but not change the memory bus?

Posted: **Wed Sep 19, 2007 4:06 pm**

In overall performance of the card(Both 128mb and 256mb). What is bottlenecking the most would you say or are they well balanced? GPU or Memory? If one lacking, would overclocking it yield good worthwhile results(improve efficency of the card I suppose). I'm not too familar with these series of cards.

- Tweak

EDIT: This was informative I thought,

http://www.anandtech.com/video/showdoc.aspx?i=2984&p=2

Posted: **Wed Sep 19, 2007 6:55 pm**

dickeywang wrote:
Pascal_TTH wrote:

*Up to* is the best way to do marketing by showing big numbers ! NVIDIA did not tell to manufacturer you have to run a those specs. Each manufacturer can adjust core and memory clocks to have a TDP that the laptop can handle.

It's very common when you design a thin laptop to reduce memory clock speed, memory bus width or core clock because it's not possible to install a big fan assembly. BTW, you also increase battery life !

In the other side, if you design a 17" laptop, you can go over NVIDIA recommendation and use higher frequency for both GPU and memory. NVIDIA did not care about it. You have a lot of room in an 17" to mount tall fan assembly. If you have done the good job, your HSF handle the extra power.

I got a lot of Thinkpad, IBM/Lenovo is NOT the manuacturer whitch use the highest frequencies. Most time, they even runs the lowest.

About bus, it's easy. When you design the mainboard, you set for exemple 4 *slots* for memory chip with 32 bits bus each. You place 4 64 MB chip to have 256 MB with 128 bits bus. You install 2 64 MB chips to reach 128 MB with 64 bits bus. To achieve the same bandwitdh, you clock the 2 64 MB chips twice faster.

To find the true, we need a pict in high quality of a T61p memory chip or just the chip ref. Thus, it's easy to find the specs including freq, voltage, bus width and so on. But I think 15,4 inch T61 have 128 bits bus with low freq mem while T61 14 have 64 bits bus with 2 times the memory speed.

In my past post, I make a wrong quote. I correct it now. But something seems to be wrong in the Core clock domain 0 and Core clock domain 1 nearly two times bigger for second one.

Posted: **Thu Sep 20, 2007 12:14 pm**

Your correct, they both have the same bandwidth. Still im curious about overclocking these cards. Obviously the 256MB memory overclocks will get more bandwidth per clock then the 128mb. I wonder if the bandwidth potentionals are the same, meaning that when you overclock the 256 version Xmhz max that the 128version will be Xmhz*2 max. Again the bottlenecking GPU/Memory still isn't clear yet.

Posted: **Thu Sep 20, 2007 1:03 pm**

tweakfreak1988 wrote:Your correct, they both have the same bandwidth. Still im curious about overclocking these cards. Obviously the 256MB memory overclocks will get more bandwidth per clock then the 128mb. I wonder if the bandwidth potentionals are the same, meaning that when you overclock the 256 version Xmhz max that the 128version will be Xmhz*2 max. Again the bottlenecking GPU/Memory still isn't clear yet.

It's hard to know. Most time, you have more chance to push low freq memory then already high clocked *two times*. I search the net from time to time to got the sepcs of those chip but I never found them.

I really would like to know about this because I'am looking for a T61 but still not sure if I will go for 15,4" 16/10 1680x1050 with 256 MB FX 570 or for 14,1" 4/3 1400x1050 with 128 MB FX 570. I'am really fad up with AMD driver and their buggy ballshit Catalyst Control Centrol as slow as not ergonomics.

Posted: **Thu Sep 20, 2007 2:48 pm**

I was thinking the same thing concerning the memory overclock. I ended up purchasing the 14.1" T61P(have yet to recieve it), portablity and battery life is a high prioity for me since I am always "on the go."

Another thing I was thinking about is a program that will overclock to my specified clocks when certian apps are launched and will underclock when at desktop or non-specified apps. I used a few for other programs with other cards, but im not sure what works with the 570M.

Posted: **Thu Sep 20, 2007 3:17 pm**

Don't worry about underclock, the Quadro/GeForce are much more power saver then the Radeon/FireGL. I use RivaTuner to monitor the clock. You don't have to bore about it. Clock goes up when you need power and goes down during gentle applications. The cloks speeds change dynamically like SpeedStep !

Posted: **Thu Sep 20, 2007 4:25 pm**

Rivatuner is only used to adjust clock speeds when playing games(i.e. an O/C profile).

I wonder what the power consumptions and battery run times look like compared between the 570m, 140m and the x3100 with identicial system configurations. I haven't found too much yet on Google, most likely because these are fairly newer chips.

Posted: **Thu Sep 20, 2007 4:59 pm**

RivaTuner can do far more then you think !

Here is how freq can change during running some 3D :

According to some guys from Toshiba, from X3100 to Quadro NVS 140m, battery life is about 3/4 to 1 hour shorter... Can be until more then one hour less with Quadro FX 570m.

Thinkpads Forum

ok i got it finally T61p 14.1 4:3