Question Help understanding M.2 on Asus X99 PRO

Sep 18, 2020
8
0
10
So I'd like to finally pony up for a Gen3 NVMe M.2 for my system. Being an audio engineer, some of my software tools are super heavy in terms of libraries needing to import, and it can really slow down my workflow (imagine waiting a solid couple minutes just for something to load off a SATA 860 Evo and hope it doesn't outright crash your session). I'm looking at the SK Hynix Gold P31 for this as it has a decent balance of short and sustained read times (I don't have to worry much about write times, since the only time I'll be writing to it is when I dump my libraries to it) and comes in at a super affordable price for the 1TB variety.

What I want to know is: since the m.2 slot on my board is gen2, am I going to be tied to gen2 speeds, or should I expect to see full gen3 NVMe speed?
 
Sep 18, 2020
8
0
10
You can only get what the slowest device in the chain can produce. Here, the slow slot.

You will not get the full performance out of a Gen 3 x4 drive in that Gen 2 slot.

Thank you so much for the reply. With that in mind, how might I anticipate speeds to compare to my SATA 860 Evo? If it's not a significant improvement I'll just save my money.
 
Sep 18, 2020
8
0
10
Okay so I just got got off support with ASUS. According to them and the manual for the board, as long as I'm not trying to populate the 4th PCIe slot on my board, my m.2 slot supports speeds up to 32Gb/s, which is a 5x speed improvement over the SATA 3 connection I'm using now, and matches the drive I'm looking at.

Unless there's something I'm blatantly missing, it looks like ASUS engineers really did a solid here and I should offer condolences to my wallet for the drive I'm about to buy.
 

USAFRet

Titan
Moderator
Okay so I just got got off support with ASUS. According to them and the manual for the board, as long as I'm not trying to populate the 4th PCIe slot on my board, my m.2 slot supports speeds up to 32Gb/s, which is a 5x speed improvement over the SATA 3 connection I'm using now, and matches the drive I'm looking at.

Unless there's something I'm blatantly missing, it looks like ASUS engineers really did a solid here and I should offer condolences to my wallet for the drive I'm about to buy.
pg x

"M.2 and SATA Express onboard- The latest transfer technology with up to 32 Gb/s data transfer speeds for M.2 "

That would speak to PCIe 3.0 for that slot.
Your selected drive should be good.

However...don't expect "5x" over your SATA III 860 EVO.
Faster, yes. But probably not "5x"
 
  • Like
Reactions: misha_91

USAFRet

Titan
Moderator
You can expect ~ 1500 - 1700 MB/s for sequential transfers on PCIE 2.0 x4 connection.
860 Evo tops out at ~ 450 - 500 MB/s (on sequential transfers).

So .. it would be ~3x times faster.
And in actual use, my 660p in a PCIe 3.0 slot adapter, which benches around that same speed....is unnoticeable vs the various SATA III drives in the same system.

Benchmark, yes.
User facing difference? Maybe not.

fQSPLrD.jpg
 
  • Like
Reactions: misha_91
Sep 18, 2020
8
0
10
You can expect ~ 1500 - 1700 MB/s for sequential transfers on PCIE 2.0 x4 connection.
860 Evo tops out at ~ 450 - 500 MB/s (on sequential transfers).

So .. it would be ~3x times faster.

3x faster for loading huge sample and synth libraries is an absolute win for me. When I have a client sitting next to me and we're flipping through presets in Omnisphere to find a starting point for a sound they want, cutting those load times down to about a third will save a lot of time (and their money, since they typically pay hourly), and ultimately allow more room for their creativity to express itself, unhampered by technological nuisances.

Also this drive will eventually move to new system when I'm ready to pull the trigger and build. I just wanted to make sure the drive would be compatible and still offer a significant performance improvement over my current setup to make it worth buying now as opposed to waiting.



Side question, but still related: would I be able to see full unhindered performance if I used something like this instead, or would that still be bottlenecked to gen2?
 
Sep 18, 2020
8
0
10
Realistically, look to somewhere between 0x and 3x.

My daily use of Adobe Lightroom, writing out to each type of drive (SATA III and the above NVMe 660p) shows zero diff.

But, try it. Your specific use case might benefit.

Well now I'm genuinely curious. In the name of science, I have ordered the drive. I'll copy my libraries to it and run some "real world" loading tests, of course using my phone as a stopwatch (so not super accurate). I'll be sure to report back with my results.
 

USAFRet

Titan
Moderator
Well now I'm genuinely curious. In the name of science, I have ordered the drive. I'll copy my libraries to it and run some "real world" loading tests, of course using my phone as a stopwatch (so not super accurate). I'll be sure to report back with my results.
I have my Intel 660p in this PCIe adapter:
https://www.amazon.com/gp/product/B07GFDVXVJ

Other system specs in my sig below, and raw benchmarks seen above.

In a typical use case with Lightroom:
I took 5x RAW files from my Fuji camera.
Applied a bunch of edits and changes to them.
Exported those as .jpg, out to each type of drive.
Clearing the cache and rebooting between each iteration.

Many people would look at the 3x speed of the 660p, and assume it did this 3x faster.
No.

1 year old Samsung 860 EVO 1TB - 15 seconds
3 month old Intel 660p NVMe - 15 seconds
6 year old Samsung 840 EVO - 15 seconds

Raw drive speed can't hurt, but it is not the only factor.

As said...your use case may benefit. Only way to know is to try.
 
Sep 18, 2020
8
0
10
I have my Intel 660p in this PCIe adapter:
https://www.amazon.com/gp/product/B07GFDVXVJ

Other system specs in my sig below, and raw benchmarks seen above.

In a typical use case with Lightroom:
I took 5x RAW files from my Fuji camera.
Applied a bunch of edits and changes to them.
Exported those as .jpg, out to each type of drive.
Clearing the cache and rebooting between each iteration.

Many people would look at the 3x speed of the 660p, and assume it did this 3x faster.
No.

1 year old Samsung 860 EVO 1TB - 15 seconds
3 month old Intel 660p NVMe - 15 seconds
6 year old Samsung 840 EVO - 15 seconds

Raw drive speed can't hurt, but it is not the only factor.

As said...your use case may benefit. Only way to know is to try.

I'm definitely curious and will let you know how it goes! Side note, I love that your avatar is Klaymen from The Neverhood!
 
Sep 18, 2020
8
0
10
I'm back!

Here is the benchmark result for the SK Hynix Gold P31 1TB NVMe in my ASUS X99-PRO/USB3.1 motherboard (Core i7-5820k; 32GB (8x4GB) DDR4-2400).

shgp31.JPG.c306683e3e3c86fabf304900e40b13c7.JPG


The sequential speeds are about as expected. The random speeds are very concerning. I would like to find out why they're so different when compared to this benchmark of the same drive.