Will the 960 pro improve compile speeds?

Dev Zone
1

I’ve been working with Unreal for awhile now, and building changes can take a fairly long time even, with a 5960X. On Windows everything builds a bit faster than Linux. Most of the build time comes from Unreal fixing the output binaries based on changes to the intermediate files. (As far as I’m aware there is no option for storing intermediate files in ram. Even if I could I don’t believe I have enough ram for that)

I already own an 850 pro. I’m just curious if the faster speeds of the 960 pro could significantly impact my 1 to 2 minute build times. I’m sure it will speed up building the engine during installation, but that is not nearly as much of a worry for me.

2

Is it worth it for you to spend so much money in slightly faster compile times?
How does CPU clockspeed affect the compile time.

3

I'm pretty sure CPU clock speed has a greater effect. What would really decrease compile times is implementing support for ccache. I'm going to be getting that ssd anyways, since I need additional room for my Windows partition. I just wanted to check if improvements in disk access would have a noticeable effect on compile times.

4

I would think, A CPU with higher L1 cache and L2 cache would be a better solution to compile times...

5

Yeah, upgrading to Broadwell E would be a better solution, but that costs even more than the 960 pro. Plus, I get other general benefits with the 960 Pro, like improving install times and a whole other set of improvements. Plus I need more room on fast media for a Windows install, and this way I can use my 256 GB Windows drive for Swap space. When I get the Pro I'm going to upgrade to 128 GB of ram, so I can use Yaourt, rather than having to save to the SSD before installing.

I'm just trying to figure out if having a faster disk would actually provide a significant benefit with Unreal having to access intermediate files stored on the disk. The best solution is Epic, or a contributor who knows the codebase adding support for ccache, but that's going to be a ways off from what I can tell.

6

I would just keep task manager up during a compile and see what's being stressed out the most. Are only some of the cores being used? That would lead me to believe that clock speed is needed. Is your current SSD being slammed? Then you know you need 960 xP

7

Ouch...

If it's on a drive, then yes it would help considering how large the files are. Remember you have a buffer on SSD's once that buffer is full transfer speeds tank. Samsung's are really good performers though.

This guy is a trusted reviewer, and I consider him a walking NAND chip. Check it out

https://www.pcper.com/reviews/Storage/Samsung-960-PRO-2TB-M2-NVMe-SSD-Full-Review-Even-Faster

8

Do a test with a RAM disk since that will give you nearly unlimited disk speed

1 Like
9

Since I'll have a left over disk when I upgrade I assume keep swap at twice my ram size, so I don't trigger a freeze in a few corner cases.

I can definitely test by putting the intermediate files, and binaries in a ram drive.