Sounds like GK208 laptops/cards will support most sm_35 features

I have to get one of this card to try the performance with scrypt hashing (mining crypto coins). The barrel shifter could really help.

I still find it really hard to tell the difference between GK 107 and GK 208. What about cards that have “901” MHz, as opposed to the 900 MHz the GK 107 runs at? And then how to tell the difference for “superclocked” cards that run at higher speeds than the default?

UPDATE: ordered myself a MSI N640-4GD3

GT630 I linked previously is a GK208 as I figured. It is also definitely PCI-E 2.0 x8, and not any faster – the previous GK107 GT 640 was PCI 3.0 x16 and I have the right flags enabled. See below:

Device 1: "GeForce GT 630"
  CUDA Driver Version / Runtime Version          5.5 / 5.0
  CUDA Capability Major/Minor version number:    3.5
  Total amount of global memory:                 2048 MBytes (2147155968 bytes)
  ( 2) Multiprocessors x (192) CUDA Cores/MP:    384 CUDA Cores
  GPU Clock rate:                                902 MHz (0.90 GHz)
  Memory Clock rate:                             900 Mhz
  Memory Bus Width:                              64-bit
  L2 Cache Size:                                 524288 bytes
  Max Texture Dimension Size (x,y,z)             1D=(65536), 2D=(65536,65536), 3D=(4096,4096,4096)
  Max Layered Texture Size (dim) x layers        1D=(16384) x 2048, 2D=(16384,16384) x 2048
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  2048
  Maximum number of threads per block:           1024
  Maximum sizes of each dimension of a block:    1024 x 1024 x 64
  Maximum sizes of each dimension of a grid:     2147483647 x 65535 x 65535
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 1 copy engine(s)
  Run time limit on kernels:                     No
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  CUDA Device Driver Mode (TCC or WDDM):         WDDM (Windows Display Driver Model)
  Device supports Unified Addressing (UVA):      Yes
  Device PCI Bus ID / PCI location ID:           2 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

Ordering that MSI N640-4GD3 was like picking up a chick in Thailand and finding out that she is really a he. It’s a GK107. Returning it. At last amazon will refund my money. ;)

Christian

hey guys, I would really like to know the die size and transistor count of gk208, and possibly a picture of the GK208 chip :) I am the TechPowerUp! lead GPU Database Editor and NVidia themselves have been giving me crappy responses on this info.

FYI, im the one who updates the Wikipedia pages, and pci devices website :), if youll notice now even the gt635 has changed to 2.0 x8

also can you upload the bios of gk208 with gpu-z and click validate so it appears in our database? :) thanks its very appreciated

I have no desire to pry the heat sink off my new GPU, and I don’t have any Windows computers around, but hopefully someone else with a GK208 will be able to help you. :)

I already uploaded the GK208 BIOS w/ GPU-Z before you posted ;) Way ahead of you. I just validated. I could take off the heatsink and take a picture I guess, but then I’d have to do a re-paste job presumably :p

But taking the heat sink won’t give you a nice die shot. Get some fuming nitric acid to proceed ;)

Good point. Flip chip packaging appears to have disadvantages as well…

And there it is. :)

yeah youd have to re-paste it but paste is cheap, need to use some alchol based cleaner to make the chip shiny and then take a die shot :) and ill upload to db, will be the first shot ever taken of a gk208.

http://www.techpowerup.com/vgabios/140773/140773.html

I suspect it’s to cut package pincount for mobile. Limiting PCIe to x8 saves 33 pins/wires. That would explain the x8 limitation. But not why there’s no PCIE 3.0 support.

Another fact: the GK107 GT650m in a Retina MacbookPro has even lower bandwidth than that… only 3GB/sec device to host and 1.5GB/sec host to device. I’m not sure why there’s an asymmetry, but even that 3GB/sec bandwidth is only PCI2 x8 rate. [Hmm, and no easy tools to show actual PCIE connecton status on Mac that I know.]

I’ll see if I get a change either tonight or sometime before the end of the week. I have paste and alcohol swabs… just lack motivation :p

well I hope you get some motivations, because it will be great to have that gpu picture and put into the database and do you have a caliper? that can be used to measure the die which is still unknown aswel, gpu-z needs accurate info

Do you have experience with this method for taking die-shots? Because I would expect that to expose only the back side of the die which still has the same measurable size but a photo would look rather boring.

have you seen the gpu database yet? heh I have gpu chip shots of every gpu except gk208 and except oensn ot released yet

Ah ok. So you are taking shots of the back side of the chip then. Still nice work!

Heh, no caliper, although I suppose if I’m doing it I might as well get one and give it use for other things later on. The only good use I’ve needed one in the past was to measure small screw sizes, though, haha. Or maybe I’ll be low-tech and use a mm-ruler.

nice, cant wait for the info :)

Sorry, no caliper, but mm-ruler it is.

Images are unedited – took them with the camera from a Samsung Galaxy S4. Enjoy!
[url]http://imageshack.us/a/img801/4787/xo4l.jpg[/url]
[url]http://imageshack.us/a/img15/5880/4w4g.jpg[/url]

I re-pasted with Arctic Silver 5 and couldn’t get the board above 62 deg C at full load w/ GPU-z rendering. Forgot to measure max temp before, however.