Faster Storage Matters - $h!t Manufacturers Say

[greenmax]

I dont like the black/white theme. White is an interesting color, visible dirt marks too easily.

[veritronx]

A few things.. the thumbnail shows an entry level kingston sata ssd, but in the video we see a WD Blue m.2 sata ssd installed which is noticably faster and competes with the 860 evo for fastest sata consumer ssd.

 

The gen3 ssd looks to be a cheap 2x pci-e drive with a slow controller and possibly even QLC flash, which would be a fair comparison to the cheap kingston drive shown in the thumbnail but would be slower in actual use than the WD Blue sata drive.

 

Gen4 ssd's currently are tuned for bandwidth and not latency, to the point where AMD told wendle not to use one when benching threadripper 3000 because it will slow it down compared to good gen3 drive (like a corsair MP510 or samsung 970 EVO)

 

So really the problem was with the drives used, but that wasn't discussed in the video and was possibly even misrepresented by the video's thumbnail because if people buy the sata drive shown in the thumbnail performance will be worse than every drive shown in the video, sometimes not even reaching more than sataII speeds.

[greenmax]

3 minutes ago, veritronx said:

thumbnail shows

They have a props room to grab shit out of for their clickbait thumbnail pictures. However I bet they just slip in pictures off the web that are free for use.

[veritronx]

3 minutes ago, greenmax said:

They have a props room to grab shit out of for their clickbait thumbnail pictures.

Doesn't matter, it's the first thing people see and later if they're shopping and they see a cheap option that looks like what linus was holding they would probably pick that one.

 

The problem with clickbait thumbnails is they're also memorable to an extent.

[greenmax]

Just now, veritronx said:

Doesn't matter, it's the first thing people see and later if they're shopping and they see a cheap option that looks like what linus was holding they would probably pick that one.

I completely get your point, but clickbait is clickbait, and pumping videos out as fast as they can is another angle as well. They just do whats easy for them.

 

 

[veritronx]

With great power / influence comes great responsibility.. this video is shocking to me because it's misleading and they don't seem to have noticed or even care, but this video will be referenced for years to come and could have a meaningful impact on sales in the industry.

 

If they had also included another high end gen3 system this video would have been a huge positive note because it would shown that both cheap nvme drives and gen4 drives are a step backwards compared to top end sata and gen3, but as that isn't even mentioned the conclusion will instead be that no nvme is worth it and people shouldn't even consider them.

[Mr. Shenanigans]

So now i am more confused than before.  I am shopping currently for more storage.  I want it fast and i dont want my drive to slow down my i9 9900k or my 2080ti. 

 

I was on a understanding that if i buy this: https://www.amazon.com/Samsung-970-EVO-1TB-MZ-V7E1T0BW/dp/B07BN217QG/ref=sr_1_1?crid=3IYW9JM19N792&keywords=m.2+ssd&qid=1582084258&refinements=p_89%3ASamsung&rnid=2528832011&sprefix=m.2%2Caps%2C430&sr=8-1

 

I would be getting the biggest baddest thing out there(the evo pro is supposedly better tho the price for the 2-4-tb of storage or more isnt great on the wallet)  

 

If i bought 2 of them and raided them that i would be somewhat faster as well.  Not 2x but maybe 1.5 times faster as long as i used the supplied m.s slots on my msi mag z390 tomahawk board.

 

but now this video tells me to get just a standard wb blue ssd like this :  https://www.amazon.com/Blue-NAND-500GB-SSD-WDS500G2B0A/dp/B073SBZ8YH/ref=sr_1_1?crid=2L46W4351R4Y8&keywords=wd+blue+ssd&qid=1582084499&sprefix=wd+%2Caps%2C216&sr=8-1   

(or something similar)

 

and my over all experience will be faster????????

 

the confusion is real

[Dylanc1500]

Should have thrown an Optane drive in the comparison as well.

[kokosnh]

The random reads, average, and 99th percent delays are the moust important, depending on the use, it has a lot of variables. Good SATA SSD can beat bad NVME SSD in some not bandwitch related task easily. Gen 4 can have problems with delays. And the windows being windows. 
then the whole thing with cache and ram cleaning for testing. 

 

The bandwidth alone isn't the speed of an SSD.

[RejZoR]

I guess I'm fine with my few years old Samsung 850 Pro 2TB hehe. What's weird now is copying or moving from SSD to my cold storage 8TB HDD at ~200MB/s. I'm not used to such low speeds lol...

 

M.2 form factor with NVMe in various speeds is only useful if you need absurd sequential speeds or you want drive without extra power or external cables. Otherwise SATA are still great options and quite a lot cheaper too. It's all about fast random access and even 550MB/s sequential is plenty for a lot of tasks.

[Letgomyleghoe]

19 hours ago, Razor512 said:

Hopefully Samsung will avoid price gouging for the 980 pro. The 3.5GB/s reads of the 970 pro is really inadequate for many general tasks.

wUt

 

3.5GB/s is better than what people need for general tasks

[CadetSparklez]

thought the standard for itx boards with dual channel was to have 1 dimm slot on each side.. thanks for clarifying.

 

what else could be why the pcs seemed slower? multiple people said that on the 3rd system it got hung up at 75% in doom I think it was.. seems extremely odd as both types of ssds use the same technology so the faster ones should be better in all scenarios.. unless the ones they were using had a crippled 4k random read performance or something. But that doesnt apply to initially loading up a game like with doom

[RejZoR]

10 hours ago, CadetSparklez said:

thought the standard for itx boards with dual channel was to have 1 dimm slot on each side.. thanks for clarifying.

 

what else could be why the pcs seemed slower? multiple people said that on the 3rd system it got hung up at 75% in doom I think it was.. seems extremely odd as both types of ssds use the same technology so the faster ones should be better in all scenarios.. unless the ones they were using had a crippled 4k random read performance or something. But that doesnt apply to initially loading up a game like with doom

The standard was to read a diagram in the board manual where they describe in detail which slots can be populated to achieve certain channel specifications... Or am I the only one reading those things since I've been jumping through so different configurations every time. Had dual channel, triple channel (with weird capacities of 18GB RAM 3x2GB+3x4GB slotted as symmetrical triple channel) and now I'm on quad channel. I literally had to read the manual coz I wasn't sitting on dual channel for decades like most ppl.

[jones177]

I have two gaming computers that use the same motherboard, CPU, ram and both have 2080 tis. One uses a 970 EVO and the other uses two 1tb Crucial SATA SSDs.  

 

Only my modded games that use 4 and 8k textures and have more than 5 times the NPCs in them benefit from my 2tb 970 EVO right now. 

The difference is stutter in open world in game load-ins. I have my modded games with the same saves on both computers so it is easy to tell the difference. 

 

This sort of thing is not new to me. In 2008 my modded Oblivion game was slow and stuttered. My character could not take a step or turn without the 7200 rpm HDD being accessed and that meant stutter. In desperation I tried a 10,000 rpm WD Raptor that cost more than double the 7200 rpm drive and it was half the size. The stutter was gone.

I also put them in my work computers and a PhotoShop file that took 5 seconds to load with a 7200 rpm drive only took 2 seconds with the Rapor.  Windows load times were about the same.

 

The WD Raptors lasted up until 2016 when my heavily modded Skyrim started to get load-in pauses/stutter. Since I added at least one mod almost every weekend since 2011 I was not surprised that I broke the game.  A 1tb SATA got the game smooth again.      

 

In a couple of years SATA SSDs will be too slow for open world games since they will have more content and larger textures. It really depends on the speed of the SSDs on the new consoles when it will happen.   

[4thVariety]

Loading time is just a general term between a player indicating his intent and the game starting. Here is what happens when you load a game, such as Division 2.

 

You click on "Play Game".

This will load the copy protection scheme and perform a check.

Since the game is competitive, the cheat protection will quickly scan your memory for anything funky.

Now the real exe is being started and you are connected to the login server, which checks the ID received from your Uplay client.

You receive a list of servers to which to connect.

Upon getting a connection the engine itself is loaded to memory to connect to the server.

Your savegame is downloaded now.

Depending on your character look, the texture required to display are loaded.

Welcome to the character selection screen. You may have seen some videos to cover up what was happening up to this point.

 

Now, if you click "Benchmark", the game will fully locally load the level to do the benchmark and display it. This loading time is the only time you can actually benchmark your loading time in this game. It consists of a few file transfres from SSD to memory and quite a bit of unpacking by the CPU followed by a bit shader crunching by the GPU.

 

If you load into the game itself, you will notice that it take A LOT longer. This is because there is another round of network connectivity happening in which you get a list of servers transmitted to your PC and then connect to one of them for the instance that you are a part of. Essentially, it is the first part of this list all over again. Especially online games have reached the point at which loading times are more like 'waiting on the server times'. It will be funny to see what Sony will be doing once they outfitted the PS5 with super fast SSDs, while companies such as Ubisoft retain poor server performance leading to 'slow' loading. 'Loading' the Tower in Destiny already is its own meme.

 

 

You can track behavior such as this on a basic level with your Windows 10 Resource Manage that is part of the Task Manager. If you are so inclined you can throw a firewall at it and have some fun blocking individual IPs to see how the game is squirming when it does not reach the servers it wants to reach. The server the game contacts initially is always the same, so if you block that, you can see the next server in rotation. You can also see why Windows 10 users really need the functionality to sandbox their games in the future. The amount of game developers taking the liberty to scan your memory for whatever they want (claiming it was to protect you from cheaters) is really concerning.

[kingmeowser]

Most games don't really care if you have a SATA SSD or an NVMe SSD. In fact, a 7200RPM HDD is totally fine for most games because they only load once per level / area. However, the games where you absolutely will notice a massive difference are heavily modded skyrim and fallout 4. I had to put these games on my NVMe RAID array because even on SATA SSD the load times get really aggravating when you have 100+ mods going (over a minute going outdoors / indoors sometimes). I bet these two games would benefit hugely from being installed on a PCIe4 drive.

[Thanatopsis]

What bugs me about the video was that the editing test was given no time at all. and in addition stacking clips is not shown in the video editing test for drive performance even though its the most likely to show if there is any actual performance difference between systems.  

SSD's and Ethernet have one basic characteristic similarity for testing there are two speed characteristics to both.  there is concurrent use and there is bandwidth bandwith is how much data can go down the pipe at any one time where concurrent use is how many different data-streams can use that bandwidth.  If you only test one or the other then you get a bad impression of what the system can do as a system that can only handle 1 or two requests at a time but can blast the data down the pipe can develop a backlog of send requests that make the system look slower than if you had a lower bandwith but higher ability to handle multiple send and receive requests.