Worthwhile(?) gtx-660 upgrade to AMD965 system

nss000

Distinguished
Apr 18, 2008
673
0
19,010
Gents:

I am considering a HW vidcard upgrade to a legacy AMD965 / 790fx-gd70 / UBUNTU system. This system does everything , but game. Of-course the main question is one of value and improved performance, though currancy & reliability may be issues.

Reading reviews, the high-value vidcard champ seems now to be the NV_660 (I'll exclude ATI) . It would be replacing a loyal, but aging 5-yo BFG-9800gtx+ which can manage only 2-fr/sec in UNIGINE-HEAVEN on a 1920x1200 screen.

A new EVGA_650x2G gives me 10-fr/sec on my Xeon 2560x1440 workstation. Upgrading vidcards may I expect analogous improvement in the older AMD system?



 
Solution
Oh the possibilities. In my wayward youth I believed spherical Bessel functions were a good way to model cell motility. Cranked on it with a refrigerator-sized SGI mini. Retired and wiser now ... I seem to have stumbled into a two-photon light-scattering attack on Leventhals Paradox ... if I can ever get poly-aminos solidly stuck to colloid metals on-the-cheap and keep the pointers holding polymer configurations something besides globally scrambled eggs!

So yes I can find an excuse. Of-course when reds are running in the St Johns fageddabouddit.



Your point is well-taken. Here's a small hitch; some graphics-load is placed on that hardware as I do run CAD, drawing and circuit-analysis (SPICE) programs.

Had planned to shift-over that graphics work to the Xeon system, but discovered I liked the old Hanns 1920x1200 monitor better than the new 2450 x 1440 Auria !!



 
Yes, a "pro" vidcard is always an option for my XEON/gd55 workstation assuming a BIOS flash. The EVGA_660 could be moved over to my legacy AMD965 (lucky duck I should ever find a used AMD980 !!) since the EVGA-650 now works just fine in that system.




 
Before even considering workstation cards you need to check if your workload will benefit from them. Some "professional" workloads work as well or better on consumer graphics cards and save you coin in the process. If you don't have any use for strong double-precision floating-point performance then you may be wasting money.
 
Oh the possibilities. In my wayward youth I believed spherical Bessel functions were a good way to model cell motility. Cranked on it with a refrigerator-sized SGI mini. Retired and wiser now ... I seem to have stumbled into a two-photon light-scattering attack on Leventhals Paradox ... if I can ever get poly-aminos solidly stuck to colloid metals on-the-cheap and keep the pointers holding polymer configurations something besides globally scrambled eggs!

So yes I can find an excuse. Of-course when reds are running in the St Johns fageddabouddit.





 
Solution