umtalelab.com
NAVIGATION
  • Writing
  • Lab
  • Archive
  • About
TOPICS
  • Blogs
  • News
  • Reviews
  • Tests
CONTACT
  • [email protected]
  • Telegram
  • YouTube
  • Patreon
© 2019–2026 UMTALE LABGLORY TO UKRAINEPrivacy Policy
Writing ▾
Lab
About
Contact
// UMTALE.LAB/MENUESC — close · ↵ — go
// UMTALE.LABMENU
MAIN// MENU
Lab→About→Contact→
UMTALE LAB · BENCH ONLINE
UMTALE.LAB/LABORATORY/DB/// GPU
UMTALE.LAB/DB/gpu
DATABASE · HARDWARE

Graphics chips GPU

Graphics chips (the silicon, not the cards): compute blocks, memory, TDP, FP/INT format support for AI. Filter by architecture, process node and ranges.

IN VIEW
2
SHOWN
2 of 2
FILTERS
no filters
⌘K
FILTERS

Manufacturer

Architecture

Process node

Foundry

RANGES

Die sizemm²314 – 471
Compute units (SM/CU)20 – 30
Execution units (FP32)2560 – 3840
Max TDPW180 – 250
Base clockMHz1607 – 1607
Boost clockMHz1733 – 1733
Max VRAMGB16 – 24
Max memory busbit256 – 384
FP16 (half)TFLOPS0.14 – 0.2
FP32 (single)TFLOPS8.87 – 12.63
FP64 (double)TFLOPS0.28 – 0.4
INT8 (shaders)TOPS35.5 – 43.5

FEATURES

  • AV1
  • Async DMA / TMA
  • With sparsity
CompareManufacturerArchitecture
NVIDIA GP102GP1027 variantsNVIDIAPascal303,840—2502412.6302/08/2016
NVIDIA GP104GP10415 variantsNVIDIAPascal202,5601,733180168.87327/05/2016
// DATABASEReference data. No buy prices, stock, or ads.UMTALE.LAB/DB · 002