I have a C2050 tesla card- guess it uses fermi.
Tell me if I have the notation correct. I am interested in GPCPU applications. I have been quite successful
and management may want to know whats on the horizon for cuda architecture to leverage what I already have.
is tesla an environment (operating system sort of ) name with fermi,kepler,maxwell - those three, refer to cards and a technology base onto which tesla is implementeed? Do I have that nomenclature right? I not set me straight.
I heard kepler uses GK104 chips and is coming out this month for gamers using GTX690. That’s not me. I wonder when this kepler technology will be available for general gpu computing, the people using C2050 now, for financial and commercial apps?
Is it correct to refer to a kepler based tesla superseding the fermi based tesla we have now? (and maxwell based tesla eventually) do I have my terms straight?
Is keper going to be twice as fast as fermi, and maxwell maybe 8 times as fast?
What is the timeline for kepler and maxwell, and what are the real performance increase factors over fermi? I am ready for using multiple gpu’s I think, but my current C2050 can only use the one gpu. How many gpu’ will maxwell have?
My company is windows centric, but the new PS2 bus?[PCI-Express 2.1 x 16] is this part of the future for kepler based cards? Is linux a better choice for price reasons and for getting your hands on a PCI-Express 2.1 x 16 bus?
From a person who does the programming, I need a executive summary of the cuda horizon so I can make the case that what we have will grow faster than the new multi cpu machines comming out. I understand that Nvidia has made the choice to go with AMD slow but low heat chips so the performance per watt will allow more compute horsepower but will leverage the cuda programmers ability to use the parallelism effectively.
Is kepler or maxwell using these ARM chips, or your own designs?
When is the next cuda developers conference?
Thanks
The Tesla cards are cards from Fermi ( maybe also later they will be call also Tesla) which have increased number of double precision units (about 4 times more than the gaming cards for the Tesla 20xx series) , extra memory, ECC. The core architecture is the same gamig cards of the respective generation.
For now for Kepler only the gaming cards were released so the exact performance of the next generation Tesla cards is not known yet. You can do gpgpu computing on the gaming cards as well if they meet your needs.
At this moment no card contains in it an ARM cpu. The CARMA paltform combines an ARM cpu with a card, but the ARM cpu is still separate unit from the card.
The HPC brand is Tesla. The architectures of Tesla, in order of release, have been Tesla, Fermi, and Kepler. The term Tesla is mainly used to describe the HPC brand, but is also the name of the first architecture (C1060 for instance was Tesla architecture).
NVIDIA has said it’ll come out later this year. Perhaps we’ll learn more next week at GTC.
That is correct.
Impossible to say such things generically. Depends on the algorithm. Depends on core count and clock speed of device.
If you want more than 1 GPU, you need to buy more GPUs and install each in a separate PCIe slot. You also need software to help you scale to the multiple GPUs. ArrayFire Pro is a good option for that :)
I don’t know this one. Haven’t heard anything about Kepler and PCIe 2.1.
I’m not sure I understand the question here (e.g. why did you mention AMD?). There are plenty of Multi-Threaded CPU vs CUDA benchmarks that have been released over the last 5 years. The GPU dominance continues to grow. We keep a blog that has a bunch of these benchmarks in it.
ARM is not related to Kepler or Maxwell, unless you’re referring to Denver.
Next week. You’d better hurry up and register! Come by the AccelerEyes booth and chat in more detail too!
Where can I find the where/when and how to register for the developer conferences from NVIDIA. I have been a member here for a couple of years but have had my nose to the grindstone working, so I have not taken to time to find out about and attend these things.