5 replies [Last post]
LociOiling's picture
User offline. Last seen 2 hours 46 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders

Several people have mentioned problems playing larger puzzles, especially after the recent scoring update.

As an experiment, I fired up my old laptop, which started life as Windows XP, and ran puzzle 1505: Aflatoxin Challenge: Round 7. I used only recipes, and kept it going for pretty much the entire two weeks the puzzle was active.

I wasn't using the laptop for anything while 1505 was active.

During this time, I didn't see any crashes or problems until the very end. I restarted the client after every recipe. (After the very last recipe, the client stopped responding, and I had to force it to close. No idea why that happened, there's was nothing in log.txt to explain it, and no debug.txt was generated.)

I ended up with a score of 13,602, good enough for 61st place solo.

To me, this indicates that on Windows 10 at least, Foldit is now relatively stable, and, in theory, even an older system may be able to handle at least one puzzle at a time.

Back when this laptop was on Windows XP, Foldit was not playable. It tended to crash with cryptic Windows error: "The application has requested the runtime to terminate in an unusual way.". My sympathy if you're seeing that.

More details in the next post.

LociOiling's picture
User offline. Last seen 2 hours 46 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
old laptop specs

My old laptop is Lenovo T61P, purchased refurbished.

I used Speccy, a free utility available from https://www.ccleaner.com/, to extract the details:

Operating System: Windows 10 Pro 32-bit
CPU: Intel Mobile Core 2 Duo T7500 @ 2.20GHz, Merom 65nm Technology
RAM: 3.00GB Dual-Channel DDR2 @ 332MHz (5-5-5-15)
Motherboard: LENOVO 6457AK7 (None)
Graphics: LCD 1680x1050 (1680x1050@60Hz), 256MB NVIDIA Quadro FX 570M (Lenovo)
Storage: 232GB Samsung SSD 850 EVO 250GB (SSD)

The BIOS on this system is dated 8 April 2010, but I think it actually dates back to 2007.

Long story, but to get to Windows 10, I bought a copy of Windows 7 Pro on eBay, and then did a clean install over Windows XP, wiping out the original setup. The disk drive failed shortly after the upgrade, so I got the solid state disk (SSD) as a replacement. All this was back when Windows 10 was a free upgrade for older systems, going back to Windows 7 at least. Officially, the upgrade program ended up in 2016, but according to some sources, Microsoft kept allowing this type of upgrade through 2017.

This laptop is so old that it can't take full advantage of the speed of the SSD. I figured the SSD was a worthwhile investment, since it will probably be useful after this system bites the dust.

LociOiling's picture
User offline. Last seen 2 hours 46 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
cores and threads (more technical stuff)

Most computers these days have multiple "logical processors", which are also called "cores". A laptop might have 4 logical processors, a desktop might have 8. Newer gaming systems might have 12 logical processors.

My old laptop has two logical processors, reflected in the "Core 2 Duo" name.

Newer Intel processors talk about cores and "threads". So a desktop might have 4 cores with 2 threads each, giving 8 logical processors.

These hardware threads shouldn't be confused with software threads.

Each Foldit client has about 15 software threads running at one time. Each software thread can use a maximum of one logical processor at a time.

Before the big update, one thread did most of the work when you were running a recipe with the Foldit window minimized. So running one client on an system with eight logical processors left almost 7/8 of the available processing power unused.

After the March update, Foldit clients now use 1.5 or more logical processors in the same scenario. There's now a second thread doing some of the work. If you want to keep each Foldit running as quickly as possible, this means running fewer clients.

The trend toward using more than one logical processor per client was actually evident on larger puzzles (such as aflatoxin) before the update, but it's hard to ignore now.

The bottom line on this part is that as long as your total CPU use is below 100%, your Foldits are folding as fast as possible. If you see CPU running at 100% all the time, chances are that shutting down one or more Foldits will let the others run a little faster.

In Windows, the CPU percentage includes all logical processors, so it can't go over 100%. Other systems may report each logical processor as a separate CPU, so you could see 800% CPU on a system with 8 processors.

The Task Manager in Windows can report CPU, but I also use the Process Explorer utility from Microsoft, https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer

LociOiling's picture
User offline. Last seen 2 hours 46 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
"slower" update

We were told that the March update to Foldit is slower than the previous version in some ways.

So far that seems true, although there's not an easy way to compare old and new. In the new version, wiggle often seems to finish its work more quickly, and start to "spin". Perhaps recipes can adjust their wiggle strategies by specifying fewer wiggle cycles to gain back a little performance.

Other operations, especially mutate, seem to be taking much longer in the new release. It's not clear if anything can be done about that.

The other complicating factor is that Foldit now routinely uses significantly more than one logical processor when running recipes. This means that one Foldit can perform more computations in a given amount of time, but you can run fewer Foldits at one time. (At least if you want each Foldit running at max speed.)

Even running fewer Foldits, the new release seems much slower than the old. Maybe some new recipe strategies can help compensate. Foldit is doing more work, but the advantage may be better results.

Joined: 09/24/2012
Groups: Go Science
Thanks

When we'll have some more precise technical info, one can try, for further recipes, to take this into account.
I wonder if there would be a system to return score info to the recipe in order to adapt wiggle strategy according to client speed.
May be using the points per second or something like that ? (current wiggle strategies only use the points per iteration ppi).

LociOiling's picture
User offline. Last seen 2 hours 46 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
heat! (quick fix)

Related to the topic of performance, I noticed the fan on my newer laptop started making noise. The area around the vents also seemed very warm.

Since this "newer" laptop is nearly six years old, and has been running Foldit almost full-time since 2013, it may be time to replace the fan.

I started Speccy to check the CPU temperature. It showed the reason for all the warm air -- the CPU was near 90° C with two Foldit clients running. A little warm for comfort.

Current CPUs slow down if they get too hot, preventing damage, but hurting performance. (Thanks to Wbertro for pointing this out.)

As a first step, I got out the vacuum cleaner and cleaned out the vents. I have computer cleaning attachments, including a 1/2" / 12 mm crevice tool and little brush.

The vacuum worked better than expected. The CPU temp immediately dropped, and is now around 70° C with two clients running. Still a little on the warm side (158° F), but more within a normal range.

I did the vacuuming with the computer running. Other sources have suggested using canned air. I'll try that the next time I shut down.

It didn't look like vacuuming was needed -- no visible dust build-up on the vents. But a desktop will inevitably have lots of dust inside the case, particularly around the CPU cooler, so really not a surprise.

This wasn't the first time I'd had the vacuum out, but previously I've focused more on the keyboard and the case. Just a few seconds on the vents made a big difference.

I'll probably still replace the fan (or "thermal module" as Lenovo calls it), but a bit of hoovering bought me some time. The noise has decreased, but it's still there.

Next step: the desktop is running a little warm. Time for some spring cleaning.

Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Microsoft, Adobe, RosettaCommons