GT 520

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "GeForce GT 520"
  CUDA Driver Version / Runtime Version          5.5 / 5.0
  CUDA Capability Major/Minor version number:    2.1
  Total amount of global memory:                 1024 MBytes (1073741824 bytes)
  ( 1) Multiprocessors x ( 48) CUDA Cores/MP:    48 CUDA Cores
  GPU Clock rate:                                1620 MHz (1.62 GHz)
  Memory Clock rate:                             533 Mhz
  Memory Bus Width:                              64-bit
  L2 Cache Size:                                 65536 bytes
  Max Texture Dimension Size (x,y,z)             1D=(65536), 2D=(65536,65535), 3D=(2048,2048,2048)
  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: 32768
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  1536
  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:     65535 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:                     Yes
  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:           1 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 5.5, CUDA Runtime Version = 5.0, NumDevs = 1, Device0 = GeForce GT 520

GT 640(GK208)

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "GeForce GT 640"
  CUDA Driver Version / Runtime Version          5.5 / 5.0
  CUDA Capability Major/Minor version number:    3.5
  Total amount of global memory:                 1024 MBytes (1073741824 bytes)
  ( 2) Multiprocessors x (192) CUDA Cores/MP:    384 CUDA Cores
  GPU Clock rate:                                1046 MHz (1.05 GHz)
  Memory Clock rate:                             2505 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:                     Yes
  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:           1 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 5.5, CUDA Runtime Version = 5.0, NumDevs = 1, Device0 = GeForce GT 640

コメント

差分
Compute Capability: 2.1 → 3.5
CUDA Cores: 48 → 384
GPU Clock rate: 1620 MHz → 1046 MHz
Memory Clock rate: 533 Mhz → 2505 Mhz
L2 Cache Size: 65536 bytes → 524288 bytes
Max Texture Dimension Size(3D): 2048 x 2048 x 2048 → 4096 x 4096 x 4096
Total number of registers available per block: 32768 → 65536
Maximum number of threads per multiprocessor: 1536 → 2048
Maximum sizes of each dimension of a grid: 65535 x 65535 x 65535 → 2147483647 x 65535 x 65535

GPU Clock rateが下がってるが、CUDAコア数がめっちゃ増えてるのでだいぶ性能が高そう。
カタログスペック上でも、 GT 520 が 155.5 GFLOPS に対して GT 640(GK208) は 803.3 GFLOPS となっており圧倒的。
しかしこのクラスのGPUで1TFLOPS目前ってなんか感慨深いものがある。

最大テクスチャサイズが増えてるけど、4096 x 4096 x 4096 x float4のサイズ(16バイト)で計算すると1Tバイトになるので実質そこまでは使えないだろう。

L2キャッシュとレジスタ数が増えたのはすごくいいと思う。
グリッドの最大分割数のx成分が増えてるのもいい。

GT 520

[CUDA Bandwidth Test] - Starting...
Running on...

 Device 0: GeForce GT 520
 Quick Mode

 Host to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     6294.4

 Device to Host Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     6319.3

 Device to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     6559.3

GT 640(GK208)

[CUDA Bandwidth Test] - Starting...
Running on...

 Device 0: GeForce GT 640
 Quick Mode

 Host to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     3145.1

 Device to Host Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     3246.2

 Device to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)        Bandwidth(MB/s)
   33554432                     31320.5

コメント

Host to DeviceとDevice to Hostの帯域幅が半減してしまってるが、これは GT 520 が PCIe 2.0 x 16 対応なのに対し、 GT 640(GK208) は PCIe 2.0 x 8 対応止まりだからである。
このクラスなら性能的なデメリットより消費電力的なメリットの方が大きいと判断しての仕様かなと思う。
カーネルが超軽くて、デバイス-ホスト間のデータ転送がボトルネックになるようなアプリケーションだともしかしたらGT 520に逆転される場合もあるかもしれない。

Device to Deviceに関しては、GT 520が少し遅すぎな気はするが、まぁ納得の値になってる。
カタログスペック上ではビデオメモリの帯域幅は、 GT 520 が 14.4 GB/s、 GT 640(GK208) が 40.1 GB/s となっている。

GT 520

GPU Device 0: "GeForce GT 520" with compute capability 2.1


Sorting 1048576 32-bit unsigned int keys and values

radixSort, Throughput = 34.3259 MElements/s, Time = 0.03055 s, Size = 1048576 elements
Test passed

GT 640(GK208)

GPU Device 0: "GeForce GT 640" with compute capability 3.5


Sorting 1048576 32-bit unsigned int keys and values

radixSort, Throughput = 98.2300 MElements/s, Time = 0.01067 s, Size = 1048576 elements
Test passed

コメント

GT 640(GK208) のほうが3倍くらい速い。
Thrustのソートアルゴリズムについてはよく理解してないのでこれ以上はノーコメ。

GT 520

65535 particles
15-30fps

GT 640(GK208)

65535 particles
65-94fps

コメント

煙の球がふわふわ動くやつ。
fpsの数値は、画面上に表示されてるものを目視で計測。
煙の球が視点に近いと重く、遠いと軽くなる。
煙のシミュレーションもさることながら、影（半透明影？）の描画もこのクラスのGPUにとっては重い部類ではないかと思う。

GT 640(GK208) のほうが3−4倍速い感じ。

GT 520

GPU Device 0: "GeForce GT 520" with compute capability 2.1

> GPU does not support HyperQ
  CUDA kernel runs will have limited concurrency
> Detected Compute SM 2.1 hardware with 1 multi-processors
Expected time for serial execution of 32 sets of kernels = 0.640s
Expected time for fully concurrent execution of 32 sets of kernels = 0.020s
Measured time for sample = 0.351s

GT 640(GK208)

GPU Device 0: "GeForce GT 640" with compute capability 3.5

> Detected Compute SM 3.5 hardware with 2 multi-processors
Expected time for serial execution of 32 sets of kernels = 0.640s
Expected time for fully concurrent execution of 32 sets of kernels = 0.020s
Measured time for sample = 0.050s

コメント

HyperQのテスト。
HyperQについてはここの説明が分かりやすい。

ソース見てないけど、serial executionとfully concurrent executionに差がないことから、カーネル自体はGPUの性能差で速度差が出ないようなものになってるのではなかろうか。
で、HyperQの有り無しで7倍程度の速度がが出てるみたい。

HyperQが無効でも、serial executionの2倍近い速度が出てるわけで、複数ストリームの同時実行がいかに大事かが分かる。