I remember relic.
I'm still doing the old f@h on an AMD 4p and HD 7970s. Recently PPD had been majorly cut but this hardware is due for replacement. If I stick to folding at home the next hardware will be GTX 980 Ti GPUs or something 14nm.
Nothing much has changed for me, I haven't updated my FAH setup in at least a year. I'm running the same 4p and triple HD 7970 system as last winter and still getting good PPD.
I think the announcement could turn out even better than before, bigadv machines becoming obsolete was a bad idea. Too bad VJ didn't wait for his new PR hire to word it a bit nicer than a termination date for bigadv. Really it could be more of a merging of bigadv and SMP if they set QRB right...
What if the fix is to get rid of SMP and move all WUs to a point scaling that matched bigadv? Fast machines would still be rewarded throughout an increased QRB and the current bigadv WU would only be assigned to machines that had enough RAM and were returning lighter WUs fast enough. That way...
The threshold is not arbitrary, it helps Stanford control how much research is being done by projects that use bigadv WU vs standard SMP WU. Some WU benefit from the fastest machines but Stanford doesn't want a large fraction of their network power being donated to these projects because the...
That sounds very positive. The main remaining issue is that because the "# of cores" metric is so weakly tied to performance there is still no indication at all about how fast systems will need to complete WUs to make deadlines and stay in the ppd inflated "top 5%" category. When Standford works...
F@H drama is a pain and you are not helping. Stanford's views seem to perpetually differ from those of the enthusiast donors which is unfortunate but the layers of forum moderators and representatives in between seem to only stir up disagreement. Everyone can read what Stanford says and either...
I like the first long term roadmap and communication points. If I'm buying a new piece of hardware for the project the three factors I look at are upfront cost, power usage, and the "usefulness to the project". The first two are easy to track and don't suddenly change on me, it would also be...
One trick is to use VNC such as tightVNC to access the system, Windows Remote Desktop messes with GPU drivers while VNC has no effect. Teamviewer is great for remote access over the internet, I don't know if it works well for folding.
Another trick is to right click on the desktop to change...
Seems the FPGA chip alone is around $7000 if you want to solder it on yourself, I think the dev board could be cheaper than that though since the goal is to get people trying out the tools.
Altera's new tools write the Verilog for you.
Paste in your OpenCL functions with some modifications here and there and at the end Altera produces the Verilog needed to run the functions on their FPGA. Basically it automatically create a custom chip architecture for you with advanced arithmetic...
So using the average between Tear's and Biffa's numbers the Intel and AMD systems have about the same cost per point over 2.5 years with $0.11/kWh electricity.
i4p Xeon
Front Price: $3500
Performance: 980 kppd
Power: 650 watts ($1700)
Points/$: 172k
OC 4p
Front Price: $1600
Performance: 700...