Romex Software Forum

Diablo 3 now suffers a lot from stuttering issue and Blizzard is promising to have a fix for this "soon". In the mean time, I tried installing the game on a high speed USB drive, the stuttering is alleviated somehow but it is still there, quite a nuance.

Then I came up with the idea of using RAM as disk cache, tried with FancyCache 0.8 beta Volume ed. I set to read-only mode, 512 KB block size and cache size of 4GB (the whole game installation size is just 8GB, so the cache is 1/2 of that). Since I installed Diablo 3 on a USB drive 16GB, so I activated the caching on that volume. The result was disappointing, no improvement whatsoever and I noticed that the reported read hit rate < 1%.

Please help.

How about if you set up the block size from 512KB to the cluster size of the file sytem on this USB drive?

Result was the same, <1% with block size 4 KB. See attached screenshots for my settings and statistic.

support wrote:How about if you set up the block size from 512KB to the cluster size of the file sytem on this USB drive?

The statistic data is reset after you close the monitor dialog, so the hit rate only refers to the period that monitor dialog opens. You may need to keep the monitor open for a long time to get the actual read hit.

I knew that, I opened the Performance Monitor dialog before running Diablo 3, it was kept open the whole time I ran around in game (~5 minutes). You may say that 5 minutes is so short for FancyCache to do any significant caching but last time I did do the same test for around an hour of gaming and the result was similar, it was consistently below 1%, just like the screenshots I captured with 5 minutes in game.

There is 1 thing that I feel not right in the Performance Monitor dialog, Diablo 3 installation size is 8GB, but even when I play game for an hour, the reported read bytes in this dialog is < 100MB. I don't think that Diablo 3 only reads 100 MB from disk, is it possible that FancyCache does not intercept all read disk calls from Diablo 3?

Testing with CrytalDiskMark shows result very close to posted on your web site: 5000 MB/s for sequential read performance or something like that, while without caching is just 50 MB/s.

support wrote:The statistic data is reset after you close the monitor dialog, so the hit rate only refers to the period that monitor dialog opens. You may need to keep the monitor open for a long time to get the actual read hit.

taihd wrote:I don't think that Diablo 3 only reads 100 MB from disk, is it possible that FancyCache does not intercept all read disk calls from Diablo 3?

FC shall intercept all read calls. I don't think there is a problem on interceptions. Is it possile that D3 only reads about 100MB from that volume, or reads other data from other volumes?

I did activate caching for all volumes in my system to check that and saw that data is read only from the volume I installed Diablo 3, other volumes showed 0 bytes of read bytes.

support wrote:

taihd wrote:I don't think that Diablo 3 only reads 100 MB from disk, is it possible that FancyCache does not intercept all read disk calls from Diablo 3?

FC shall intercept all read calls. I don't think there is a problem on interceptions. Is it possile that D3 only reads about 100MB from that volume, or reads other data from other volumes?

Well, how about if you use third-party tools like "Process Monitor" (http://technet.microsoft.com/en-us/sysi ... s/bb896645) to verify the read data amount?

Thanks.

OK, I opened Performance Monitor of FancyCache and Process Monitor before running Diablo 3, then ran around, killed monsters in game for 5 minutes, here is the statistics, big discrepancy wrt read bytes reported by FancyCache (2MB) and Process Monitor (~1.1GB). See attached screenshots.

support wrote:Well, how about if you use third-party tools like "Process Monitor" (http://technet.microsoft.com/en-us/sysi ... s/bb896645) to verify the read data amount?

Thanks.

Er... I am sorry that I made a mistake. Process Monitor is a file-level monitor that intercepts the data at the file system level, earlier than FancyCache. I think the discrepancy of read bytes actually indicates that most of read data was provided by Windows cache itself.