r****y
发帖数: 1437 | 1 attend a seminar by a NVIDIA research scientist,
seems GPU has a lot of potential for scientific computing. |
j**u
发帖数: 6059 | 2 heard it a long time ago. is there anybody here doing this?
【在 r****y 的大作中提到】
: attend a seminar by a NVIDIA research scientist,
: seems GPU has a lot of potential for scientific computing.
|
r****y
发帖数: 1437 | 3 The gain seems very impressive, just the investment to
export all codes to a platform only supported by one chip maker
seems not that worth at this moment.
【在 j**u 的大作中提到】
: heard it a long time ago. is there anybody here doing this?
|
l*****a
发帖数: 119 | 4 I know some guys in the CS dept of my university have been working on this
for several years. They bought at least up to 5 PS2 for research. admire
them. :-( |
l***g
发帖数: 1035 | 5 cuda... they have an interface for matlab.. but the programming part is huge
task and it's too problem specific.
【在 j**u 的大作中提到】
: heard it a long time ago. is there anybody here doing this?
|
l***q
发帖数: 208 | 6 That's for cell, not CUDA
【在 l*****a 的大作中提到】
: I know some guys in the CS dept of my university have been working on this
: for several years. They bought at least up to 5 PS2 for research. admire
: them. :-(
|
L*******g
发帖数: 913 | 7 我在学校机房里看见过用gpu计算的cluster。别的cluster都不要显卡，它上每个单机
都插两块显卡，大概是SLI。每个显卡都在那里闪着荧光，整个cluster看起来很华丽，
当然发热也很厉害。 |
x*****u
发帖数: 3419 | 8 Yes.
This is the way to go..
【在 r****y 的大作中提到】
: The gain seems very impressive, just the investment to
: export all codes to a platform only supported by one chip maker
: seems not that worth at this moment.
|
x*****u
发帖数: 3419 | 9
【在 r****y 的大作中提到】
: The gain seems very impressive, just the investment to
: export all codes to a platform only supported by one chip maker
: seems not that worth at this moment.
|
