Aspettando Nvidia Pascal

  • Autore discussione Autore discussione Utente cancellato 159815
  • Data d'inizio Data d'inizio
Pubblicità
Stato
Discussione chiusa ad ulteriori risposte.
è talmente grande il die che offusca il sole...

- - - Updated - - -

Ha un tdp anche leggermente inferiore a parità di grandezza(235w vs 250w).
 
[h=2]The Pascal GP100 Architecture: Faster in Every Way[/h]With every new GPU architecture, NVIDIA introduces major improvements to performance and power efficiency. The heart of the computation in Tesla GPUs is the SM, or streaming multiprocessor. The streaming multiprocessor creates, manages, schedules and executes instructions from many threads in parallel.
Like previous Tesla GPUs, GP100 is composed of an array of Graphics Processing Clusters (GPCs), Streaming Multiprocessors (SMs), and memory controllers. GP100 achieves its colossal throughput by providing six GPCs, up to 60 SMs, and eight 512-bit memory controllers (4096 bits total). The Pascal architecture’s computational prowess is more than just brute force: it increases performance not only by adding more SMs than previous GPUs, but by making each SM more efficient. Each SM has 64 CUDA cores and four texture units, for a total of 3840 CUDA cores and 240 texture units

gp100_block_diagram-1-624x368.png


https://devblogs.nvidia.com/parallelforall/inside-pascal/
 
1480 :)...quanto arriverà in overclock?
1800:lol:

- - - Updated - - -

Certo che 610mm2 è la gpu più grande mai prodotta.... batte l'enorme gm200 (601mm2)
mamma mia sono già innamorato, è immensa

- - - Updated - - -

comunque ragazzi 3584 cc fp32

quindi i 4096 cc di gp104 erano completamente fasulli

[TABLE="width: 1"]
[TR]
[TD]Tesla Products[/TD]
[TD="width: 117"]Tesla K40[/TD]
[TD="width: 123"]Tesla M40[/TD]
[TD="width: 116"]Tesla P100[/TD]
[/TR]
[TR]
[TD="width: 122"]GPU[/TD]
[TD="width: 117"]GK110 (Kepler)[/TD]
[TD="width: 123"]GM200 (Maxwell)[/TD]
[TD="width: 116"]GP100 (Pascal)[/TD]
[/TR]
[TR]
[TD="width: 122"]SMs[/TD]
[TD="width: 117"]15[/TD]
[TD="width: 123"]24[/TD]
[TD="width: 116"]56[/TD]
[/TR]
[TR]
[TD="width: 122"]TPCs[/TD]
[TD="width: 117"]15[/TD]
[TD="width: 123"]24[/TD]
[TD="width: 116"]28[/TD]
[/TR]
[TR]
[TD="width: 122"]FP32 CUDA Cores / SM[/TD]
[TD="width: 117"]192[/TD]
[TD="width: 123"]128[/TD]
[TD="width: 116"]64[/TD]
[/TR]
[TR]
[TD="width: 122"]FP32 CUDA Cores / GPU[/TD]
[TD="width: 117"]2880[/TD]
[TD="width: 123"]3072[/TD]
[TD="width: 116"]3584[/TD]
[/TR]
[TR]
[TD="width: 122"]FP64 CUDA Cores / SM[/TD]
[TD="width: 117"]64[/TD]
[TD="width: 123"]4[/TD]
[TD="width: 116"]32[/TD]
[/TR]
[TR]
[TD="width: 122"]FP64 CUDA Cores / GPU[/TD]
[TD="width: 117"]960[/TD]
[TD="width: 123"]96[/TD]
[TD="width: 116"]1792[/TD]
[/TR]
[TR]
[TD="width: 122"]Base Clock[/TD]
[TD="width: 117"]745 MHz[/TD]
[TD="width: 123"]948 MHz[/TD]
[TD="width: 116"]1328 MHz[/TD]
[/TR]
[TR]
[TD="width: 122"]GPU Boost Clock[/TD]
[TD="width: 117"]810/875 MHz[/TD]
[TD="width: 123"] 1114 MHz[/TD]
[TD="width: 116"]1480 MHz[/TD]
[/TR]
[TR]
[TD="width: 122"]FP64 GFLOPs[/TD]
[TD="width: 117"]1680[/TD]
[TD="width: 123"]213[/TD]
[TD="width: 116"]5304[1][/TD]
[/TR]
[TR]
[TD="width: 122"]Texture Units[/TD]
[TD="width: 117"]240[/TD]
[TD="width: 123"]192[/TD]
[TD="width: 116"]224[/TD]
[/TR]
[TR]
[TD="width: 122"]Memory Interface[/TD]
[TD="width: 117"]384-bit GDDR5[/TD]
[TD="width: 123"]384-bit GDDR5[/TD]
[TD="width: 116"]4096-bit HBM2[/TD]
[/TR]
[TR]
[TD="width: 122"]Memory Size[/TD]
[TD="width: 117"]Up to 12 GB[/TD]
[TD="width: 123"]Up to 24 GB[/TD]
[TD="width: 116"]16 GB[/TD]
[/TR]
[TR]
[TD="width: 122"]L2 Cache Size[/TD]
[TD="width: 117"]1536 KB[/TD]
[TD="width: 123"]3072 KB[/TD]
[TD="width: 116"]4096 KB[/TD]
[/TR]
[TR]
[TD="width: 122"]Register File Size / SM[/TD]
[TD="width: 117"]256 KB[/TD]
[TD="width: 123"]256 KB[/TD]
[TD="width: 116"]256 KB[/TD]
[/TR]
[TR]
[TD="width: 122"]Register File Size / GPU[/TD]
[TD="width: 117"]3840 KB[/TD]
[TD="width: 123"]6144 KB[/TD]
[TD="width: 116"]14336 KB[/TD]
[/TR]
[TR]
[TD="width: 122"]TDP[/TD]
[TD="width: 117"]235 Watts[/TD]
[TD="width: 123"]250 Watts[/TD]
[TD="width: 116"]300 Watts[/TD]
[/TR]
[TR]
[TD="width: 122"]Transistors[/TD]
[TD="width: 117"]7.1 billion[/TD]
[TD="width: 123"]8 billion[/TD]
[TD="width: 116"]15.3 billion[/TD]
[/TR]
[TR]
[TD="width: 122"]GPU Die Size[/TD]
[TD="width: 117"]551 mm²[/TD]
[TD="width: 123"]601 mm²[/TD]
[TD="width: 116"]610 mm²[/TD]
[/TR]
[TR]
[TD="width: 122"]Manufacturing Process[/TD]
[TD="width: 117"]28-nm[/TD]
[TD="width: 123"]28-nm[/TD]
[TD="width: 116"]16-nm[/TD]
[/TR]
[/TABLE]
 
Stato
Discussione chiusa ad ulteriori risposte.
Pubblicità
Pubblicità
Indietro
Top