Hi,
I intended to order S2070 Teslas in Q3 and just got green light for the funding.
Anyway I noticed that the S2070 is not referenced anymore by nVidia. I was not able to find any additional info except, that the devices just disappeared.
Can someone give me an update to allow me to plan accordingly? Is it canceled? Delayed?
I really need the big MEM + bandwidth of the S2070 for my application.
Thanks,
Markus
Hi Markus
We are moving all customers to M2070-based products. Please send a note to an NVIDIA sales person
to get more info (or to me) - select your country from
http://www.nvidia.com/object/tesla_wtb.html
Sumit
Hi Sumit - is that move only relevant for S2070 or also for the S2050? i.e will you be moving
S2050 to M2050 as well???
thanks
eyal
S2050 is available and already shipping now.
There are many OEM servers now that incorporate M2050 - so its your choice between S2050 or M2050-based servers.
S2050 is available and already shipping now.
There are many OEM servers now that incorporate M2050 - so its your choice between S2050 or M2050-based servers.
Hi,
thanks for the fast reply. But I’m a bit confused. Could you please clarify how many GPUs I can control using a board with four 16x PCI slots when
a) using S2070/S2050
b) using M2070
?
As I understood I can control 2 GPUs using a single PCI 16x when using the interface cards to connect S20X0, and just one GPU when plugging in a M2070.
For single server computing systems this results in just half the number of GPUs available (4 vs. 8). This downgrades to a “Personal-Supercomputer” config with a marginal increased clock rate.
What I planned todo is using a single server and avoid all cluster management overhead that is required when using multiple machines.
Besides the fact that you are currently moving customers to M2070: Is this motivated because the S2070 is delayed or because it is completely canceled? In other words: Will the situation change when I can wait a bit longer?
Markus
Hi,
thanks for the fast reply. But I’m a bit confused. Could you please clarify how many GPUs I can control using a board with four 16x PCI slots when
a) using S2070/S2050
b) using M2070
?
As I understood I can control 2 GPUs using a single PCI 16x when using the interface cards to connect S20X0, and just one GPU when plugging in a M2070.
For single server computing systems this results in just half the number of GPUs available (4 vs. 8). This downgrades to a “Personal-Supercomputer” config with a marginal increased clock rate.
What I planned todo is using a single server and avoid all cluster management overhead that is required when using multiple machines.
Besides the fact that you are currently moving customers to M2070: Is this motivated because the S2070 is delayed or because it is completely canceled? In other words: Will the situation change when I can wait a bit longer?
Markus
Sumit,
I requested quotes from all nVIDIA distributors for Germany. They all claim they cannot get MXXX Teslas.
Furthermore EDU/R&D/University discounts are only very poorly available (only for C2050s and only for the next 2 weeks).
Only one distributor claims to be possibly able to get a discount also for other products managed, but is still checking this.
Furthermore some of the distributors didn’t even bother to send me a quote at all.
In summary:
- Availability of Fermi Teslas for vendors is very poor in general (many delays)
- MXXX products are not available at all
- No reasonable discounts for universities
I’m very unhappy with the current situation in German Tesla distribution.
Just my Feedback.
Markus
How many GPU has S2050 and how mucn does it cost? I assume C2050 has 1 gpu and costs 2666, am I right?
C Series and M Series have 1 CPU, S has 4 GPUs.
I can only give you the German pricing for C2050 and C2070 I got:
C2070: 2700 Euro
C2050 2400 Euro or 1800 Euro for EDU institutions
Thanks very much. I am too have not seen S2050 on sale. So I thought S will not be released soon.
I can tell С2050 has very competitive price.