I would have thought it was obvious to Raistmer too, but sometimes we all overlook the simplest of things and need a second pair of eyes to help out ;-)
:)
What is not too obvious is that GTX 260 (just as GT9400 and GSO9600 I have in other hosts) is 1.x device and NV treats 1.x devices as only OpenCL 1.0 capable, no matter what driver is. I did test that via recent driver installation. So, my NV GPUs really no go for einstein's app.
EDIT: That's how it looks:
Number of OpenCL platforms: 1
OpenCL Platform Name: NVIDIA CUDA
Number of devices: 1
Max compute units: 27
Max work group size: 512
Max clock frequency: 1487Mhz
Max memory allocation: 234799104
Cache type: None
Cache line size: 0
Cache size: 0
Global memory size: 939196416
Constant buffer size: 65536
Max number of constant args: 9
Local memory type: Scratchpad
Local memory size: 16384
Queue properties:
Out-of-Order: Yes
Name: GeForce GTX 260
Vendor: NVIDIA Corporation
Driver version: 331.65
Version: OpenCL 1.0 CUDA
Extensions: cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_d3d9_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64
EDIT2: in general there are 2 OpenCL versions to consider: 1) platform version - that's what update with new drivers.
2) device version - that's what depends a) from used platform;b) from device abilities itself. Hence, when platform abilities good enough we will have device abilitiesa as limit for OpenCL version it can suipport.
'
And that's why I don't too hurry to use OpenCL 1.1.
HD4xxx will always be OpenCL 1.0 devices no matter what driver used.
Same for NV CC1.x it seems. Why NV can't use OpenCL1.1 on those devices - can't say right now, need to look tables in specification. Maybe local memory size or register pool limiting.
Oh well, as I said earlier, OpenCL on NVIDIA is a dead end, unfortunately. We'll stick to CUDA and use OpenCL only for AMD and Intel...
The majority group of vendors in the Khronos group want OpenCL to be the way forward, over CUDA and DirectX/Directcompute, with just Nvidia and Microsoft kicking and screaming that they don't want to let go of their old tech.
The expectation is that CUDA will go on for another year, year and a half, and then fold, with Nvidia moving full on to OpenCL. Open standards are the way forward.
Graphics rendering through OpenGL will triumph over DirectX, according to AMD Vice President of Global Channel Sales Roy Taylor (who says in a German interview that DirectX 12 is never going to see the light of day - ever): Ein neues DirectX hat die Branche immer wieder neu aufgefrischt, neue Grafikkarten brauchten stärkere Prozessoren und mehr RAM. Aber es wird kein DirectX 12 kommen. Das war's. Soweit wir wissen gibt es keine Pläne für DirectX 12. Falls das nicht stimmen sollte und mich jemand korrigieren möchte – wunderbar.
As for DirectCompute, only Nvidia has drivers for it. Thus far, in all my searches, I haven't seen anyone use it, only lots of examples and samples on what you can do with it. So with its development apparently at a standstill (no updates to it since it was released in 2009!), CUDA and OpenCL FTW. :-)
The majority group of vendors in the Khronos group want OpenCL to be the way forward, over CUDA and DirectX/Directcompute, with just Nvidia and Microsoft kicking and screaming that they don't want to let go of their old tech.
That's right.
I never said OpenCL itself will go down the drain. I just said that NVIDIA hasn't really cared and is not going to care much about OpenCL (see below).
Quote:
The expectation is that CUDA will go on for another year, year and a half, and then fold, with Nvidia moving full on to OpenCL.
Who's expectation is that? I know that NVIDIA doesn't plan to support anything above OpenCL 1.1.
RE: Perhaps, OpenCL 1.0 is
)
Yep, that's right. We require 1.1.
Sorry,
Oliver
Einstein@Home Project
RE: NVidia has support for
)
(both from http://streamcomputing.eu/blog/2011-12-29/opencl-hardware-support/)
It looks as if a driver update might be all that Raistmer needs - "Driver version: 263.06" is too old.
Sure, I thought that was
)
Sure, I thought that was obvious ;-)
Einstein@Home Project
RE: Sure, I thought that
)
I would have thought it was obvious to Raistmer too, but sometimes we all overlook the simplest of things and need a second pair of eyes to help out ;-)
:) What is not too obvious is
)
:)
What is not too obvious is that GTX 260 (just as GT9400 and GSO9600 I have in other hosts) is 1.x device and NV treats 1.x devices as only OpenCL 1.0 capable, no matter what driver is. I did test that via recent driver installation. So, my NV GPUs really no go for einstein's app.
EDIT: That's how it looks:
Number of OpenCL platforms: 1
OpenCL Platform Name: NVIDIA CUDA
Number of devices: 1
Max compute units: 27
Max work group size: 512
Max clock frequency: 1487Mhz
Max memory allocation: 234799104
Cache type: None
Cache line size: 0
Cache size: 0
Global memory size: 939196416
Constant buffer size: 65536
Max number of constant args: 9
Local memory type: Scratchpad
Local memory size: 16384
Queue properties:
Out-of-Order: Yes
Name: GeForce GTX 260
Vendor: NVIDIA Corporation
Driver version: 331.65
Version: OpenCL 1.0 CUDA
Extensions: cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_d3d9_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64
EDIT2: in general there are 2 OpenCL versions to consider: 1) platform version - that's what update with new drivers.
2) device version - that's what depends a) from used platform;b) from device abilities itself. Hence, when platform abilities good enough we will have device abilitiesa as limit for OpenCL version it can suipport.
'
And that's why I don't too
)
And that's why I don't too hurry to use OpenCL 1.1.
HD4xxx will always be OpenCL 1.0 devices no matter what driver used.
Same for NV CC1.x it seems. Why NV can't use OpenCL1.1 on those devices - can't say right now, need to look tables in specification. Maybe local memory size or register pool limiting.
Oh well, as I said earlier,
)
Oh well, as I said earlier, OpenCL on NVIDIA is a dead end, unfortunately. We'll stick to CUDA and use OpenCL only for AMD and Intel...
Oliver
Einstein@Home Project
I do have a newer NVidia card
)
I do have a newer NVidia card online and available for testing, if that helps: host 1226365.
Driver is a little long in the tooth currently, but that's easy to fix. OS is WinXP - that would be harder to change.
RE: Oh well, as I said
)
The majority group of vendors in the Khronos group want OpenCL to be the way forward, over CUDA and DirectX/Directcompute, with just Nvidia and Microsoft kicking and screaming that they don't want to let go of their old tech.
The expectation is that CUDA will go on for another year, year and a half, and then fold, with Nvidia moving full on to OpenCL. Open standards are the way forward.
Graphics rendering through OpenGL will triumph over DirectX, according to AMD Vice President of Global Channel Sales Roy Taylor (who says in a German interview that DirectX 12 is never going to see the light of day - ever): Ein neues DirectX hat die Branche immer wieder neu aufgefrischt, neue Grafikkarten brauchten stärkere Prozessoren und mehr RAM. Aber es wird kein DirectX 12 kommen. Das war's. Soweit wir wissen gibt es keine Pläne für DirectX 12. Falls das nicht stimmen sollte und mich jemand korrigieren möchte – wunderbar.
As for DirectCompute, only Nvidia has drivers for it. Thus far, in all my searches, I haven't seen anyone use it, only lots of examples and samples on what you can do with it. So with its development apparently at a standstill (no updates to it since it was released in 2009!), CUDA and OpenCL FTW. :-)
RE: The majority group of
)
That's right.
I never said OpenCL itself will go down the drain. I just said that NVIDIA hasn't really cared and is not going to care much about OpenCL (see below).
Who's expectation is that? I know that NVIDIA doesn't plan to support anything above OpenCL 1.1.
I totally agree with you on this one!
Oliver
Einstein@Home Project