960 evo works fine in benchmarks, but not in Adobe Premiere Pro.

Toggle sidebar Toggle sidebar
N

naturalbehavior

Prominent
Apr 28, 2017
2
0
510
It's quite weird!

My M.2 drive shows up fine in windows, and when testing with Crystaldiskmark I'm getting the read/write numbers that I should be getting from that disk. Around 3200mb/s in read and around 1500mb/s in write. But when I run the PPBM (which is a benchmark for Premiere Pro) the speeds are only around 4-500mb/s.

Something must be wrong... I'm using the latest drivers and firmware for the drive, been testing both with and without Samsung magician installed (magician is giving me the same numbers when I test the speeds).

So, how come I'm getting the read/write speeds that I'm suppose to have when benchmarking with Crystaldisk/magician, but not when I'm exporting in Premiere?

My setup is as follows:

AMD RYZEN 7 1700 3.0GHz 8-Core Processor
Asus PRIME X370-PRO ATX AM4 Motherboard
Corsair Vengeance LPX 16GB (4 x 8GB) DDR4-3000 Memory
Samsung 850 EVO-Series 250GB 2.5" Solid State Drive
Samsung 960 EVO 250GB M.2-2280 Solid State Drive
Western Digital Red 3TB 3.5" 5400RPM Internal Hard Drive
Gigabyte GeForce GTX 1060 G1 ROCK 6G
EVGA SuperNOVA G2 650W 80+ Gold

OS/Boot is on the 850 evo. Projects/Scratch/Media DB on the 960 evo.
Memories currently running at default, which was 2133MHz.

Any ideas about what's going on here? Is there some kind of bottleneck in my system that I'm not aware of?
 
Solution
R
3200MB/s and 1500MB/s is for sequential operations only. This is pretty much the maximum the drive can ideally ever achieve in your system.

CrystalDiskMark should also provide you with random 4K IOPs or IO/s performance. This is representative of the ideal case for how fast the drive can service commands in your system. The 4K here means each command carries 4KB of data. If you convert the IO/s to MB/s then you'll see it quite a bit slower than your sequential numbers.

Actual general use workloads are more a combination of those. A general model of average PC usage is about 25% sequential writes and 75% random writes. Additionally, there's other commands like Trim and Flush that could be happening in the background.

So, going back to...
Sort by date Sort by votes
R

rkzhao

Respectable
Mar 8, 2016
183
1
1,860
3200MB/s and 1500MB/s is for sequential operations only. This is pretty much the maximum the drive can ideally ever achieve in your system.

CrystalDiskMark should also provide you with random 4K IOPs or IO/s performance. This is representative of the ideal case for how fast the drive can service commands in your system. The 4K here means each command carries 4KB of data. If you convert the IO/s to MB/s then you'll see it quite a bit slower than your sequential numbers.

Actual general use workloads are more a combination of those. A general model of average PC usage is about 25% sequential writes and 75% random writes. Additionally, there's other commands like Trim and Flush that could be happening in the background.

So, going back to what qazzi said, PPBM gives you a number that is more representative of what you can expect Premier to be able to achieve rather than just purely ideal sequential performance.
 
Upvote 0 Downvote
Solution
N

naturalbehavior

Prominent
Apr 28, 2017
2
0
510
I'm quite new to storage hardware and still not fully informed of how they work, so that answer was superb!

So I can't expect higher results with my actual hardware setup? The bottleneck is not the drive, but the rest of my setup?
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom