Might be interesting for some, who are looking for building ultimate machine, a server barebone from TYAN
(if I'm right, i believe VDACTR8 from Render Stream uses the same - kind a beast(8x580s), isn't it? =)
server Barebones for 8GPU rig
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
-
gabrielefx
- Posts: 1701
- Joined: Wed Sep 28, 2011 2:00 pm
last year I was buying this unit but I decided to split my purchase in two separate units.
This machine generates a lot of noise and hot.
You need a rack and about 2,6Kw of power on one single line.
Unfortunately Kepler gpus will not substitute our GTXs. to get more power from our workstations.
Nvidia said that Kepler is 3x time faster than GTX...this is not true neither for the new Tesla.
This machine generates a lot of noise and hot.
You need a rack and about 2,6Kw of power on one single line.
Unfortunately Kepler gpus will not substitute our GTXs. to get more power from our workstations.
Nvidia said that Kepler is 3x time faster than GTX...this is not true neither for the new Tesla.
quad Titan Kepler 6GB + quad Titan X Pascal 12GB + quad GTX1080 8GB + dual GTX1080Ti 11GB
The noise with this is an issue. The only viable solution in my eyes to use this monster via thin client =) in a small studio that would make sence (deploing the render server somewhere in the basement)
As far as i remember about the factor if '3x' they have stated that's more efficient, not 3times faster =) but one or the other is right. An it is true with their products as “Tesla K10 at single-precision delivering a total of 4.58 teraflops” (while 580 and it's counterpart in Quadro an Tesla range delivers ~1,5tflops, if i remember good enough =)
The bad is that k10 is not aimed at workstation, only on server market, but later this year we ar going to see k20 based on gk110, it's already stated that we actually going to see a product for workstations ( hope it will perform even better =)
As far as i remember about the factor if '3x' they have stated that's more efficient, not 3times faster =) but one or the other is right. An it is true with their products as “Tesla K10 at single-precision delivering a total of 4.58 teraflops” (while 580 and it's counterpart in Quadro an Tesla range delivers ~1,5tflops, if i remember good enough =)
The bad is that k10 is not aimed at workstation, only on server market, but later this year we ar going to see k20 based on gk110, it's already stated that we actually going to see a product for workstations ( hope it will perform even better =)