UPDATED in February 2017
Some optimisation tricks work really well on one architecture, and are useless on others. And even with better drivers, the older architectures need some help. In other words, it helps to know what architecture the GPU has. Therefore you get some help from your friends at StreamHPC.
Below you’ll find a list of the architecture names of all OpenCL-capable GPU models of Intel, NVIDA and AMD. It does not contain the professional lines for now – first we are focusing on getting the general models right.
Understand it took a lot of time to gather the below information, and normally we share such information only with our clients.
The table of GPU architectures
You see it is quite a mix and it’s hard to find general rules for each articture.
|Vendor||Architecture||GPUs||OpenCL Version support|
|AMD||VLIW5||HD5xxx, 63xx - 68xx, 69xxM, 73xx, 83xx - 84xx||1.2|
|AMD||VLIW4||Radeon HD 69xx, 7xx0D, 7xx0G||1.2|
|AMD||GCN 1 (Southern Islands)||Radeon HD 74xx - 79xx, 85xx - 89xx||1.2|
|AMD||GCN 2 (Sea Islands)||Radeon R7 2xx, R9 2xx.||2.0|
|AMD||GCN 3 (Volcanic Islands)||R9 285, R9 380[X], Fury [X], Nano, Pro Duo||2.0|
|AMD||GCN 4 (Arctic Islands)||RX 460 - 480||2.0|
|NVIDIA||G80||GeForce 8800GTX/Ultra, 9400GT, 9600GT, 9800GT||1.1|
|NVIDIA||G80-2||GeForce 8400GS/GT, 8600GT/GTS, 8800GT/GTS, 9600 GSO, 9800GTX/GX2, GTS 250, GT 120/30/40||1.1|
|NVIDIA||Tesla||Geforce (GT) 3xx, 405.||1.1|
|NVIDIA||Tesla2||Geforce GTX 260 - 295||1.1|
|NVIDIA||Fermi||Geforce GTX 465 - 480, GTX 560 - 590||1.1|
|NVIDIA||Fermi2||Geforce GT 420 - 440, GTS 450, GTX 460, GT 510 - 545, GTX 550 - 560, GT 605 - 630, GT 640, GT 645||1.1|
|NVIDIA||Kepler||GTX 645 - 690, GTX 760 - 770||1.2|
|NVIDIA||Kepler2||GeForce GTX TITAN, GTX 780||1.2|
|NVIDIA||Kepler3||Geforce GT 635, 640||1.2|
|NVIDIA||Maxwell1 (GM10x)||GTX 745, GTX 750 [Ti], GT 945A||1.2|
|NVIDIA||Maxwell2 (GM20x)||GTX 9X0 [Ti], GTX Titan X||1.2|
|NVIDIA||Pascal (GP10x)||GTX 10X0 [Ti], Titan X||1.2|
|Intel||Gen8-1||HD5300 - 6000, Iris 6100 - P6300,||2.0|
|Intel||Gen9||HD510 - P530, Iris 540 - P580||2.0|
|Intel||Gen10 (7th Gen. Intel Core)||HD610 - 630, Iris Plus 640 - 650||2.1|
All but NVIDIA G80 have mostly been validated by checking more than one website – if you see any mistake, let me know.
There are some GeForce GPUs with different architectures, while having the same name: GeForce GTX 560, GeForce GT 630, GeForce GT 640. On CUDA you can check compute-capability to find out which version you have, but on OpenCL you cannot easily do that.
AMD tends to put the latest architectures on the high-end GPUs and uses the previous architecture for the budget and mid-range GPUs. NVIDIA on the contrary introduces their latest architectures in the mid-range GPUs. IMHO should both NVIDIA and AMD rethink their naming conventions, to make things more clear.
Intel’s GPUs not always have a very clear name, but do reference to genX for some GPUs.
The OpenCL Version Support is the maximum of driver and hardware. For example NVidia’s later GPUs theoretically support OpenCL 2.0, but does not have any drivers available.