[SOLVED] Why does my 850 EVO SATA saves Blender files just as fast as my 980 Pro?

[novus111]

I've just bought a new 980 Pro 1TB, running on a X570-E Gaming motherboard, so it should have full PCIe Gen 4 support. Yet this "apex" of technology saves my Blender scene not even one second faster than my old 850 EVO SATA (250Gb) drive. How come? According to Magician benchmark my 850 EVO has 530 MB/s Read and 486 MB/s Write, while my 980 Pro reads at 7047MB/s and writes at 5155MB/s, that's over 10 times faster! How come there doesn't seem to be a difference when saving files?

 

[drivinfast247]

Maybe the PC can't process the save any faster?

 

[novus111]

The scene is at 580 Mb. Processor is a Ryzen 9 3900X. So considering the large gap in performance indicated by Magician there should be at least SOME differences in saving speed between the drives, no?

 

[USAFRet]

I've just bought a new 980 Pro 1TB, running on a X570-E Gaming motherboard, so it should have full PCIe Gen 4 support. Yet this "apex" of technology saves my Blender scene not even one second faster than my old 850 EVO SATA (250Gb) drive. How come? According to Magician benchmark my 850 EVO has 530 MB/s Read and 486 MB/s Write, while my 980 Pro reads at 7047MB/s and writes at 5155MB/s, that's over 10 times faster! How come there doesn't seem to be a difference when saving files?

A file of 580MB is not large enough for the performance of the 980 to manifest.
Many other things are in play...CPU, RAM, etc.

In my own use and testing, with an 850 EVO and Intel 660p NVMe, I see a ZERO difference in saving data from Adobe Lightroom.
The 660p Sequential benchmarks at 3X that of the 850.
In actual tested performance? Zero difference.

 

[novus111]

That's hard for me to understand. Is that supposed to mean that the two drives work like two cars, only at a large distance will one notice a difference in acceleration? The saving speed of the files is 20s for both, that's not long enough for the 980 to pull ahead at least 1sec? It's freakin 10 times faster xD That's a huge leap compared to any hardware upgrade people make in general. People buy new CPU's for a 25% performance increase...

 

[USAFRet]

That's hard for me to understand. Is that supposed to mean that the two drives work like two cars, only at a large distance will one notice a difference in acceleration? The saving speed of the files is 20s for both, that's not long enough for the 980 to pull ahead at least 1sec? It's freakin 10 times faster xD That's a huge leap compared to any hardware upgrade people make in general. People buy new CPU's for a 25% performance increase...

As said, it is also all the OTHER components.

If you were copying a 5GB file from 1 980 to another 980, you would absolutely see a huge difference over 2x 850 EVO drives.
In your particular thing, Windows, CPU, RAM, and Blender all come into play.

At 20 seconds to do this....that is almost certainly the rest of the hardware taking the vast majority of that time.


Unrelated to your Blender, but differences in game load times:
The PCIe 4.0 benchmarks at twice the speed of a 3.0, and may times faster than SATA III, right? Not so much...

View: https://www.youtube.com/watch?v=4YoRKQy-UO4

 

[novus111]

Thanks, yes I saw that video before. Not too interested in gaming, I use my build for 3D work and Photoshop.

I just made another test, this time opening a 4.6GB psb file. My 850 EVO took 1min 51sec. While my oh so fast (I can't believe I bought this <Mod Edit>) 980 Pro was able to open it a whopping 9 seconds faster. Incredible... If you'll excuse me I'll go mourn for my 200 bucks in fetal position.

 

[USAFRet]

Thanks, yes I saw that video before. Not too interested in gaming, I use my build for 3D work and Photoshop.

I just made another test, this time opening a 4.6GB psb file. My 850 EVO took 1min 51sec. While my oh so fast (I can't believe I bought this <Mod Edit>) 980 Pro was able to open it a whopping 9 seconds faster. Incredible... If you'll excuse me I'll go mourn for my 200 bucks in fetal position.

I'm not a huge gamer either. But that video is a good visualization of the diminishing returns we're chasing.

 

[USAFRet]

The main benefit of solid state is the near zero access time. This applies to ALL SSD types, including SATA III.

There was a HUGE jump from spinning drives to solid state.
Between SSD types? Not so much.

 

[USAFRet]

And CPU/RAM upgrades DO make a big difference.

Some years ago, changing from an i5-3570k/16GB RAM, to i7-4790k/32GB RAM.

Testing in Adobe Lightroom
Starting with 200 RAW images from my Fuji camera, all taken at the same time.
Importing 100 images in each test.
Exact same drive, Samsung SATA III SSD. Either an 840 or 850 EVO.

(can't remember the exact numbers, but this is in the realm)
i5-3570k - 120 sec
i7-4790k - 80 sec

That was a BIG difference.

 

[novus111]

The main benefit of solid state is the near zero access time. This applies to ALL SSD types, including SATA III.

There was a HUGE jump from spinning drives to solid state.
Between SSD types? Not so much.

It just doesn't make sense man. How come on paper aka benchmark, there's a HUGE difference which seems to have barely any "real world" application? Like how is that possible? Is it really that bottlenecked by my other hardware? As I said I got a fast CPU and I got 32gigs of 3600Mhz corsair rgb pro sticks, not exactly the slow kind either.

 

[USAFRet]

It just doesn't make sense man. How come on paper aka benchmark, there's a HUGE difference which seems to have barely any "real world" application? Like how is that possible? Is it really that bottlenecked by my other hardware? As I said I got a fast CPU and I got 32gigs of 3600Mhz corsair rgb pro sticks, not exactly the slow kind either.

It's not being "bottlenecked", but rather all the other components DO come into play.

Just like cars.
Car A has a top speed of 160mph, Car B has a top speed of 120mph.
Both have a 0 to 60 time in the same range, within 0.5 sec.
Additionally, the only roads you drive on have stop lights and speed limits.

If the only ad text you see talks about the top speed, you'd automatically assume Car A is better and faster.
Real world begs to differ.

Big numbers sell product.

 

[hotaru.hino]

It just doesn't make sense man. How come on paper aka benchmark, there's a HUGE difference which seems to have barely any "real world" application? Like how is that possible? Is it really that bottlenecked by my other hardware? As I said I got a fast CPU and I got 32gigs of 3600Mhz corsair rgb pro sticks, not exactly the slow kind either.

Here's a non-gaming video talking about if getting an NVMe SSD is worthwhile or not.

The gist of it is, your processor can only do so much. If it was taxed to begin with to the point where it's not saturating a drive, adding a faster drive isn't going to help.

All a benchmark of the SSD is doing is testing what the SSD in isolation (more or less) can do. A render also requires the CPU or whatever doing the render to do its part too.

 

[novus111]

Sigh, lesson learned. One can only hope that the purchase will be worth it in 5-6 years when the CPU's, RAM and Mobos caught up.

 

[JMComputers]

Ok, Im sorry this one has just really bugged me for a long time, first off rapid mode only works on the 250-500GB models, so if it was fake that wouldnt be the case, second the new updates, along with MSI who yes has there own additional version of a mini booster through MSI X Boost I think it was. Anyways all this aside I'm not here to troll I'm here becuase while pople cried it was a gltich I test times, I tested software I tested period and got same results faster times and nothing that I got said it wasnt working and very well.
Now the newest Versions (Last 2 versions of Samsung Magician that came out lets you select the size, times it runs.

So you can have :
a 2GB test range, that runs 9 times for your general options
Sequential Read / Sequentia Write
Chunk size up to 2MB, Queve Depth upto 32, and upto 16 thread (Assuming these settings are alot to do with your amount of L3 Cache and CPU (ON my System since its a Ryzen 2700X)

Random Read / Random Writes:
Can be upto 64KiB Chunk Size @ 64 cue depth and on upto 16 thread again.

Now my problem is with all these new setting I Have my 10 year old Samsung Evo 850 500GB (THE DRIVE THOR USED APPARENTLY)
And I have my new Samsung 970 EVO M.2 and on the Same PC with the same parameters for their tests, the 970 Barely takes the Random Lead while the Sequential is out of the park in the Evo 850's side


EVO 850:
https://drive.google.com/file/d/10c5cICZtXaRFU0q_gndCE1LsWZ-sNZDr/view?usp=sharing
EVO 970:
https://drive.google.com/file/d/1OzxrxZ6RGVkQOi0oV5Bn5N0tzHgSUkni/view?usp=sharing

Someone explain again to me why or how rapid mode isn't working and please be someone who has tested it off and on and see there is a difference in load times from everything, are we saying the more Cache the CPU has the better it works?>

 

[USAFRet]

Ok, Im sorry this one has just really bugged me for a long time, first off rapid mode only works on the 250-500GB models,

RAPID mode uses part of your RAM as a cache for the drive.
That uber benchmark number you see for the 850 EVO is mostly reading your RAM.
Seriously.

RAPID is not applicable to the 970 EVO, because it is already too fast.

I tried the RAPID on my 840s and 850s.
Saw exactly zero user facing performance benefit.
So I turned it off.

Ignoring any benchmark numbers from Magician, do YOU see any real performance benefit?


Also, please don't dredge up year old threads.