After using Gigabyte 990FXA-UD7 in prevoius bench sessions with AMD FX-8150, I was a bit unsure how 990FXA-UD5 would behave... UD5 should be the little brother of UD7... From the first look I noticed that there was no buttons for Power, Reset and clear cmos. Also there was no display for boot codes and there were fewer PCI-Express slots.
As I planned to run with single gfx, the number of slots didn't matter that much to me. The lack of display did matter that much either as I have seperate card for that (I didn't need it though). I was missing the buttons a bit though as I didn't put this board into a chassi. But hey, a small screwdriver can work as areset button also.
There was also one thing more that wasn't the same as my previous board. My 990FXA-UD7 was Rev1.0, this 990FXA-UD5 is Rev1.1.
For me, the LLC-feature was the most interesting to test. Would the board be more able to keep the CPU voltage more stable..
On the Rev1.0 of UD7, I had to do some hard mod in this area. Rev1.1 have some more components in the same area.
Slot layout:
Connectors on the back:
SATA connectors:
Same FX-8150 as earlier:
I got some help with cooling the system down while I was checking good temps, LLC, NB and stable clock with 8 cores.
No problems with pot filled with LN2
Higher temps than ~ minus 145 celsius degrees started to create problems, so all benches was run colder than this
For a while I thought I had BIOS issues, so I changed BIOS from F7h to F6, and then to F5. A small note here: Don't flash BIOS running 8 cores @ 7GHz, it might screw up something. (I had to reinstall Windows). Here is a screen that shows what I felt like a big issue:
When raising the voltage more than ~0.1-0.15v with AOD, both voltage and clocks got lowered. In the end I went back to BIOS F7h and booted up with 1.88v wich made it possible for me to have at least 1.98v vcore. On UD7 Rev1.0 I could not boot with higher voltage than 1.77v.
First bench was Cinebench 11.5. The highest score I have seen with FX is 11.88p so I was pretty happy to get 11.80p pretty easy.
After that I spent A LOT of time getting result in PCMark05. The test often went ok, but I got no score. I still havn't figured out why..
Benching all 8 cores with filled pot eats a lot of LN2, so I didn't have that much LN2 left for other benches. Next goal was to beat my (and Calathea's) previous record of 11.000s in SP1M. A few rounds later, I got 10.953s.
Now about 8 hours with LN2 had passed, I had about 2 litres of LN2 left... So I went for wprime as I didn't have any result in that. I had big problems running with high clocks, so I had to lower clocks to 6750 just to get a result. wprime32m: 5.235s
There was no way that I would have LN2 enough for a round of wprime1024m, so I did some quick runs in Cinebench trying to beat AMDnord's 11.88p, but I could not beat his score... Next time maybe.
Some more pics:
Overall, the Bulldozer is still fun to bench with and it would be nice spending more time with it if I could afford to have the pot filled with LN2 all the time. It really loves cold...
I also know that the UD5 is a great board. As far as I know/feel, the only difference between 990FXA-UD5 and 990FXA-UD7 is the missing features mentioned in the beginning of this post. If you really don't need these features on 990FXA-UD7, don't hesitate to get the 990FXA-UD5. It is not a bad board in any way.
Bookmarks