New Posts  All Forums:

Posts by tictoc

I believe you need an 11.8mm bit.
Just a little update on system temps.   The temps reported above were with only one of the 7970s running at 100% load.  With both GPUs and the 4790k running all out, temps are a little more in line with what I was expecting.  I am looking into adding another radiator (120mm or 140mm) to the loop, but I still need to figure out exactly where, and how, I can fit it into the case.   Ambient temp - 19ºC D5 - Setting "5" 4800 rpms Radiator Fans are intake/push @ 1500...
Nice work.    If you are overclocking the 280 watch the VRM temps.  MilkyWay can really heat up the VRMs, if they don't have adequate cooling.
I had some renderings to do, but I am back up and running on Leiden now.   We should be able to make it to 41 for sure.
The F@H benchmark can benchmark both single and double precision, but all of the current F@H tasks use single precision. There is a Gromacs dp core, but it is not currently being implemented in any of the CPU or GPU tasks.
I was going to wait, until I relocated my res and installed the backplates on my GPUs, to post my latest build in this thread, but it is going to be a while before I have time to redo the loop.           With ambient temps at 16-17ºC, radiator fans at 1500 RPM, exhaust fan at 1000 RPM, both GPUs and 4790k at 100% load; temps are: 56ºC on the GPUs, 66ºC on the 4790k.    The a10-7700k doesn't really add any noticeable load to the wc loop.  Running IBT on the...
 Nice.  Prize has been added to the OP.
580 to 7970 I would say was a very good upgrade. 7970 is faster in every way. I have yet to see any Ghz edition 7970s OC better than the original reference cards. I think tsm106 was running reference Diamond 7970s for his 4-way x-fire that topped the bench charts for quite awhile. 770s are 680s with faster memory. Titan is worth it if you need dp compute or the additional VRAM. Titans continue to hold their value due to their use in the professional sector. A Titan...
I'm on my phone, but here is a link to a thread in the NVIDIA developer forums. https://devtalk.nvidia.com/default/topic/745504/comparing-cpu-and-gpu-theoretical-gflops/ There are a couple of links in that thread that should answer your question. If not, I will post up a more detailed response when I get back in front of my computer.
If double precision is what you are looking for, there really is nothing on the consumer side that beats the 7970/280x.  With used 7970s regularly going for less than $150 on ebay the price/performance ratio, for crunching MilkyWay, is pretty fantastic.   OpenCL support has been steadily improving and with the exception of GPUGrid (only CUDA tasks) and PrimeGrid (much faster CUDA app) the AMD OpenCL apps are very efficient.
New Posts  All Forums: