Folding Forum

I don't find a better word to describe it - windows task manager says overal CPU usage doesn't exceed 50%, memory even less.
Yet when FAH is running (two Core x17, on Radeon HD 66xx and 69xx, Win 7 64Bit, driver up-to-date 13.152) everything is getting sluggish - reacting delayed. I'm not talking about games (I usually pause FAH for that), but regular windows - it's quite noticeable in the Browser (Firefox right now) even.
Typing (this posting or a URL) is noticeable different from before, suggestions aren't available immediately, everything reacts somewhat slower.
Minimizing this window doesn't go smoothly either.

I don't think I ever had that problem with core x16.

Could it be the task manager isn't really telling the truth? From the graphs I gather that single cores are defintely used up to 80% at times - but only for a split-second I guess.

The words that describe your situation is "screen lag". This can be caused when you are folding on the primary GPU. The simplest solution is to configure your Primary GPU to fold only when Idle. This ensures that your secondary GPU will continue to fold in the "background". The reason for this lag is that currently, there isn't any GPU scheduling system so the GPU operates in a FIFO (First In First Out) manner unlike the CPU which has a mature scheduling system. Alternatively, you can choose to disable GPU Acceleration in applications that you commonly use. This forces that CPU to render the screen instead of the GPU so lag might be reduced/eliminated.

FahCore_17 uses the AMD GPU in a much more efficient manner when compared to FahCore_16. Moreover, it can be due to optimized drivers. Not all donors experience screen lag, it just depends on the combination of WU and GPU.

Please note that if you want instructions on how to configure your Primary GPU to fold when idle, please tell us the F@H version installed on your system. If you don't know, please post the initial section of the log which contains the system information and F@H configuration.

To eliminate hardware acceleration in Firefox go to Tools>Options>Advanced. If there is a check next to "Use hardware acceleration when available" remove it and restart Firefox. For flash video right click on a video while it's playing, click on Settings and remove the check next to "Enable hardware acceleration".

Thanks for the replies - "screen lag", yes that seems to be right. I remember having similar issues when I accidently plugged the monitor to the weak card.

I'll look into it all once I'm back home, maybe Core x17 isn't as bad as I'm imagining it to be
("bad" as in "causes me specifically a lot of problems"; I know it utilizes the card much better, about 99% usage instead of 60-70)


[edit]Disabling hardware acceleration in Firefox definitely helped.

Have you guys done something to the Core x17 code???????
I have 2 Gigabyte WF-3 7950's OCed and F@H says that I can pull 210 ~ 225K PPD. This is the assuming that I'm folding each WU @ < 6 hrs x 4 = 24 hrs or one (1) full day. My BIG Complaint is that before Dec 2013 :
I had this posted under buy AMD or Nvidia GPU for PPD's. So I just copied and pasted for this thread.

When using the Sapphire Toxic HD-6950 ( Factory overclocked to HD-6970 by BIOS switch) with the new Core_17, it still uses a Full Core. Estimated ETA is 12 hrs @ 945 core and 1350 memory, Estimated PPD is < 40K. This also applies to the Sapphire HD-6970 @ 945 Core and 1455 Memory.

When using the Gigabyte WF-3 HD-7950 ( Overclocked to 1200 Core and 1385 Memory) with the new Core_17, It uses all cores to Download and Prep the 79XX series. This lasts for about 1-minute or until the core and card are properly prepped. Then the GPU takes over and the CPU usage drops to less than 2% for the rest of the folding process.

How much Superior is the 79xx over the 69xx when using the new Core_17 :
Using 2x HD-7950's stock @ 1000 Core and 1250 Memory ( Version 1 with no Voltage Lock)

EDIT !!!!!! This is using WU 8900 Core_0x17 Zeta Version 2 Beta, AMD 13.8 Version 2 Beta Drivers and MSI Afterburner Version 3.0.0 Beta 11.
The BUG In The Software, is that the PPD is calculated for how much work is done in 24 HOURS.
My HS Edubacation, stated that there are only (4) FOUR (6) SIX Hour groups in 24 Hours. My Times Below do not add up to this, if I start them @ Exacly 0000 - 2359. I lose 27K - 30K per card per day if counted like this.

Gigabyte WF-3 HD-7950 @ 1200 Core and 1385 Memory : ETA = 6 Hrs and 10 Min - TPF = 3 Min and 47 Sec - PPD = 112~K
Gigabyte WF-3 HD-7950 @ 1145 Core and 1325 Memory : ETA = 6 Hrs and 30 Min - TPF = 4 Min and 5 Sec - PPD = 101~K
This gives an Estimated Total - PPD = 213~K

Now with the ATI driver (either 13.4 - 13.11) my 7950's now have to use an entire core during the time a Core x17 WU runs.
What have you guys broke????? I was always envious of the NVidia guys talking about being able to fold with their cpu's and Vid card at the same time. They would always BRAG that their video card was using less than 2% cpu usage while folding a GPU WU. Then I got the new ZETA Core x17, I was very happy.
Now I'm back to where I was before you guys released the Core x17.

With core 17 I need one full core for one NV card ...

Also here viewtopic.php?nomobile=1&f=81&t=25440

The last core 17 change was v49 to v52 a few months back. GPU drivers change more often!

bollix47 wrote:To eliminate hardware acceleration in Firefox go to Tools>Options>Advanced.

Manually deactivation of HW acceleration is not a serious option.
The issue affects too many applications. In Open Office I can watch the tool bars being redrawn several times with each change. In Firefox it takes half a second to redraw. Is there just one queue? Does the 2D GUI use the same resources as the folding? Does the issue affect all GPU cores?
The only thing I can come up with, is using the CPU integrated GPU for everyday stuff. I do not know how that works though, I assume I have to connect the primary monitor to the on-board connector and change some settings, but what about the second monitor?

In the early days of PCs, the CPU did integer operations and you could add on a math coprocessor (e.g.-8087 to your 8080, etc.) if you needed to do floating point operations. There was only one floating point queue and if multiple applications needed it (very rare, in those days) they had to share FIFO. Today's operating systems treat the GPU in much the same way. A block of work is submitted and the GPU does whatever is necessary to process it, returning results when it's done.

FAH generates work requests. The Desktop generates work requests. YouTube generates work requests. The GPU decides which portions of its hardware are busy and if there are enough resources available, it will process more than one at a time -- or wait until the necessary resources are available. With every generation of GPU, there are improvements in how the hardware and the drivers deal with resource conflicts, but in the systems available today, those conflicts are not managed by the operating system or by any application running in the CPU.

Ok, I get that and I suppose it will get better.
But this behavior has not always been there. I have been with the GPU cores since my ATI Radeon X1900. In those 10 years there might have been issues with video playback but not with the GUI to this degree.
Maybe the work batches are too large to be reconcilable with a good "user experience" ?

moritzgedig wrote:Maybe the work batches are too large to be reconcilable with a good "user experience" ?

There has been a major change in the type of WU's being processed since the first GPU folding cores. Originally they were folding small proteins, and the cores used the implicit solvent model to calculate the steps. Newer folding cores allow the use of larger proteins and explicit solvent modelling. That does use more resources. In light of that, the default settings for the folding client do set GPU folding to only occur when the system is "idle". That would not interfere with video playback or other apps using the GPU.

I suspect both are true.

If the atom-count of the protein is small, the amount of work enqueued for processing is simpler so even on a slow GPU, a screen refresh can be injected into the queue more quickly.

Then, too, there's a tradeoff in the amount of work that should be enqueued at one time on the slowest GPU with the fastest data transfer rate and what might work well on a fast GPU with the slowest data transfer rate. If you optimize the WU for one, it becomes inefficient on the other so some compromise has to be made.

TCP/IP can adjust the block-size differently for a satellite connection and for a local fiber link. Many years of development work went into making that work well under widely varying conditions. At one time, the application programmer had to figure out how to include all that logic in a program that drove the modem. GPU hardware and driver developers still have a lot of work to do. I'll bet that's still the responsibility of the FahCore developer and they're have to continue to optimize for a wider and wider range of options until the drivers are smart enough to partition work batches and understand a processing priority setting which doesn't yet exist.