PC: Battlefield V exceptionally slow map load times
TL;DR for everyone: The solution may be to ensure you update your firmware for your SSDs, in addition to using the latest driver.
All excellent advice:
Driver: I don't know if I got a newer one than I already had installed, but I downloaded the latest as of today, installed it, and foolishly didn't check version numbers prior to running the installer, so I'm not sure if I already had this or not. Still, if others haven't installed the driver - do that for sure.
Secondly, the firmware: The Samsung app in question is (currently) named Samsung Magician. With it, I updated the firmware of all drives and also confirmed that TRIM was enabled on all drives. I also I ran the Performance Benchmark tests on all drives that were NTFS (it was going to have a lot of trouble with an Ext4 linux drive, so I ignored it).
Results were as follows (note that the units for these dimensionless values are not present, so it's unclear what it's claiming these values imply. One can probably assume bigger -> better, but everyone knows what they say about assuming):
960 EVO (m.2):
- Sequential: 3,035 Read, 1,814 Write, Random IOPS: 279,052 Read, 254,882 Write
840 EVO (SATA III)
- Sequential: 550 Read, 531 Write, Random IOPS: 66,603 Read, 64,453 Write
850 EVO (SATA III)
- Sequential: 549 Read, 529 Write, Random IOPS: 81,298 Read, 75,439 Write
I also verified that BFV was indeed installed on the 960 EVO.
In addition to these, there's also a neat little "System Compatibility" function in the Samsung Magician utility that will tell you if it has knows of any issues - mine came up clean, but that doesn't necessarily guarantee that is the truth. However, if it does say something, that seems like it could be a fruitful troubleshooting lead.
I haven't been able to test the results rigorously yet. However, I did try a new server once this morning and it seemed to load much, much quicker. I still want to take a few gaming sessions before I mark this as solved, but initial results are promising!
While I was testing this, I did a few things differently, and this may be helpful for others in diagnosing slow map load times:- Open Resource Monitor. (Start, type Resource Monitor in omnibar)
- Start BFV and jump into a game
- After you have made it into a game, alt-tab and stop the monitor by going to Monitor-> Stop Monitoring
- Stop playing (don't be a * who idles and makes your team be down a warm body!) and go check the Disk section of the monitor (not the graph, there's a little expandable section in the left hand side)
How do you analyze this data you just captured?
- Sort by image name and look for bfv.exe. BFV is clearly multiprocess, with the executable doing multiple things on your behalf as you play. Some of these will have very low Read (B/sec), but they also likely did very small things so don't be overly alarmed. You can ignore those. What you're looking for instead are the ones with the largest numbers. If those large numbers are actually fairly small, you probably do have some form of an IO problem. My highest was 398,958 bytes/sec (or ~399kB/s), with a second place of 270,885 (or ~271kB/s). I should note that these values didn't show up at all until the game was actively engaged in loading a map after I had found a server - I feel fairly confident these are the processes that actually load maps into memory and GFX memory.
There might be an inclination to look at the Disk tab and look at the bfv.exe image there. This aggregate is actually not that useful for you - all that time you sit waiting in menus prior to loading the map is going to really work against you; the bfv.exe isn't touching disk because everything is already loaded and you're just lowering your average. That said, if you can minimize the time spent in menus and get into a game really quickly, it becomes a much more accurate view of what is happening re: file IO and the bfv.exe process.
I wish I would have run this prior to (possibly) updating the m.2 driver and (definitely) updating the m.2 firmware, so I could compare before and after. However, if anyone else has an m.2 (or any form of SSD, really) and their loads seem to be unreasonably slow, it would be good to measure both before and after so you can verify quantitatively how the updates went.
