Black Fri - Amazon - Crucial T705 - £189

[Mr_sukebe]

For UK pilots with Gen5 MVME capable motherboards, Amazon are currently selling  2TB Crucial T705s for £189.

I’ve not seen them sub £300 before , so have very much ordered one.

I know that some say there’s not apparent difference between Gen4 and 5 in DCS, ask me again next week…

[MAXsenna]

For UK pilots with Gen5 MVME capable motherboards, Amazon are currently selling  2TB Crucial T705s for £189.
I’ve not seen them sub £300 before , so have very much ordered one.
I know that some say there’s not apparent difference between Gen4 and 5 in DCS, ask me again next week…

Will you do some extensive testing pre and post the upgrade?

Sent from my SM-A536B using Tapatalk

[Mr_sukebe]

33 minutes ago, MAXsenna said:

Will you do some extensive testing prep and post the upgrade?

Sent from my SM-A536B using Tapatalk
 

Oh yeah, sure.  I've been making up a kneeboard with a suitable checklist.

[MAXsenna]

Oh yeah, sure.  I've been making up a kneeboard with a suitable checklist.

Autocorrect got me!
I meant, if it wasn't clear. Will you test to see if Gen5 is better than Gen4. Like with a stop watch or similar. (Obviously you can get the timestamps from the DCS.log).

Sent from my SM-A536B using Tapatalk

[Mr_sukebe]

Ok, will do

[kksnowbear]

11 hours ago, Mr_sukebe said:

For UK pilots with Gen5 MVME capable motherboards, Amazon are currently selling  2TB Crucial T705s for £189.

I’ve not seen them sub £300 before , so have very much ordered one.

I know that some say there’s not apparent difference between Gen4 and 5 in DCS, ask me again next week…

Yup, $239 (2TB) here in the states...EU prices always seem higher for this stuff - it is what it is, and that's just that.  However, prices change (of course).

Just a matter of time.  The price will ultimately be such that it's a genuine 'no brainer' for people who have boards that support Gen5 drives.

I have two of these drives - a 1TB model and a 2TB (I've also put them in builds for other people).  I never pay 'worst case retail', and they're worth every penny to me.

As far as difference goes:

Yes, of course, they are decidedly faster than other drives - even many of the other Gen5 drives.

No, they will not increase frame rates (at least as things are now in DCS, and I don't really get why someone might think they would, TBH).

Yes, they will shorten load times (but this is not by far the only benefit).

Yes, there will always be 'haters' who insist they don't improve performance.  They are wrong, and there's a ton of data to prove that.  The problem here is that these people apparently don't understand that "performance" is more than just sheer stupid FPS.

Historically, there were haters who said the same thing about SSDs when they first came along (they were wrong then, too).  They said the same thing about every advancement in storage performance, and they were wrong every time.  You need only look at these same people's systems today:  SSDs instead of platter-based conventional spinning hard drives.  Fast drives using NVMe protocol instead of SATA SSDs.  They all go on about it not being 'worth it' but they ultimately move to the exact same thing themselves.  If it's not a good idea, then why do they all eventually wind up doing it? The cost may be higher at first, but that's a matter of individual choice.  "Worth" is absolutely and always up to whoever is paying for it, like it or not.

In spite of all the naysayers, the industry has vigorously pursued faster storage subsystems - and not just in servers, either - because they understand the impact on performance.  While it is true the costs can be an obstacle, the prices come down because there is a market for faster storage - and there's a market for it because it improves performance.

And, there is real technology, right now, today that allows actual increase in FPS once it's supported in a game - this has been proven already, and it is measurable in a system now, regardless of game support.

While it is true DCS doesn't yet support this technology, and there is developer effort associated with supporting it, I personally believe that (just like prices coming down) it's only a matter of time.  ED did extra work to put MT in DCS because most systems support it and it improves performance.  ED put DLSS in DCS because many systems support it and it improves performance.  Obviously, ED isn't above putting work into supporting performance-increasing technologies like these.

And, even if DCS doesn't support the newer technology yet, a system that supports it *will* absolutely benefit from faster storage.  Right now, today.  Regardless of specific game support.  Anyone who actually understands how storage in a computer works also understands this.

Edited November 21, 2024 by kksnowbear

[MAXsenna]

23 minutes ago, kksnowbear said:

Yup, $239 (2TB) here in the states...EU prices always seem higher for this stuff - it is what it is, and that's just that.  However, prices change (of course).

Is that with or without tax? 

25 minutes ago, kksnowbear said:

No, they will not increase frame rates (and I don't really get why someone might think they would, TBH).

Yes, they will shorten load times (but this is not by far the only benefit).

Did any of us even mentioned that? Load times, that's what I'm interested in. And if course "load times" during play, which might be harder to time. 

As for the rest of your essay, I'm not sure why you bring in 'haters' and naysayers, as I've not seen anyone here, yet.

[kksnowbear]

Yet, indeed LOL they're here...

I brought it up because it's part of the discussion. The OP stated:

2 hours ago, Mr_sukebe said:

...some say there’s not apparent difference between Gen4 and 5 in DCS

There will almost certainly be someone to come along insisting it won't matter. As I've described, this is typically because of the misguided concept that performance can only be observed in frame rates, and/or lack of understanding how storage subsystems actually work in computers.

I prefer to make sure all sides of a topic are accurately expressed.  It happens that I also have a genuine dislike of misinformation being spread  

Edited November 21, 2024 by kksnowbear

[kksnowbear]

24 minutes ago, MAXsenna said:

And if course "load times" during play, which might be harder to time. 

This is a very important factor, and one that is typically overlooked (or ignored).

In fact, if this weren't part of the equation, then it would be impossible for the advanced storage technologies to actually increase frame rates ...

...yet they have already done so.

30 minutes ago, MAXsenna said:

Is that with or without tax? 

That's before tax IIRC.  And tax can make a big difference.  I pay 6% but some others pay much more.

[Wostg]

I thought you guys were riffing on the bit about needing to see wear tests before you'd consider touching them, per the moza ffb thread. 

We need to see loadtimes after a week at persistent 100°C! 

[kksnowbear]

2 minutes ago, Wostg said:

I thought you guys were riffing on the bit about needing to see wear tests before you'd consider touching them, per the moza ffb thread. 

We need to see loadtimes after a week at persistent 100°C! 

Nahhh...

Although to be fair I think 100c is a bit much for an SSD

[MAXsenna]

16 minutes ago, kksnowbear said:

This is a very important factor, and one that is typically overlooked (or ignored).

Exactly!

18 minutes ago, kksnowbear said:

That's before tax IIRC.  And tax can make a big difference.  I pay 6% but some others pay much more.

Prices cis side of the pond usually include taxes. 25% in my country.  

33 minutes ago, kksnowbear said:

Yet, indeed LOL they're here...

☝

I was just making banter... 

34 minutes ago, kksnowbear said:

There will almost certainly be someone to come along insisting it won't matter. As I've described, this is typically because of the misguided concept that performance can only be observed in frame rates, and/or lack of understanding how storage subsystems actually work in computers.

True, working with storage systems and 100GB files, FPS is what? 

36 minutes ago, kksnowbear said:

I prefer to make sure all sides of a topic are accurately expressed.  It happens that I also have

I'm well aware. It's part of your MO, not?  

[kksnowbear]

1 minute ago, MAXsenna said:

Prices cis side of the pond usually include taxes. 25% in my country.

I didn't know that.  Here, at least, I think they don't include taxes because sales tax varies by state and even city sonetimes.  So it's factored in at check out.  And even if you buy online, you pay local sakes tax...wasn't always that way

[MAXsenna]

I didn't know that.  Here, at least, I think they don't include taxes because sales tax varies by state and even city sonetimes.  So it's factored in at check out.  And even if you buy online, you pay local sakes tax...wasn't always that way

Yeah, that makes sense. Even on Amazon it's pre-calculated here.

Sent from my SM-A536B using Tapatalk

[kksnowbear]

5 minutes ago, MAXsenna said:

Exactly!

"Load times" during play will improve with faster storage.  How much so will vary and, as you said, is much harder to nail down.  I personally feel it improves overall responsiveness, and helps reduce "hitches" from texture loading during play.

But it will make a difference, either way.  And, assuming incorporation of improvements in storage technologies, that difference absolutely can translate to better frame rates.  This has already been demonstrated.

13 minutes ago, MAXsenna said:

It's part of your MO, not?

Indeed it is

14 minutes ago, MAXsenna said:

I was just making banter... 

I gotcha

[kksnowbear]

15 minutes ago, MAXsenna said:

True, working with storage systems and 100GB files, FPS is what? 

Absolutely.  But, if I may:  The argument will then become DCS won't improve because we're not talking about 100GB files on a server...

...and that's misleading.

Performance in storage is a matter of "contention"; that is, the requestors waiting on data they want, because the system can only provide so much per unit of time.  The more you ask a storage subsystem to do, the longer it takes that system to do it - and this is irrespective of speed, type of storage, etc.  And no matter how you do it, more contention equals less performance.

It should be obvious, then, that the faster a storage subsystem is, the faster it can satisfy demand, thus there is less contention.

This is why faster storage will always perform at least somewhat better: Lower contention.  Sheer data throughput is only part of measuring performance in storage subsystems.

[MAXsenna]

@kksnowbear  

I wonder if it would somewhat be possible to measure with tracks. All the time stamps are in the DCS.log So if one runs tracks on the same version of DCS, and then run the same track on different iteration of storage on the same computer.  

[kksnowbear]

26 minutes ago, MAXsenna said:

@kksnowbear  

I wonder if it would somewhat be possible to measure with tracks. All the time stamps are in the DCS.log So if one runs tracks on the same version of DCS, and then run the same track on different iteration of storage on the same computer.  

Great question.  I dunno TBH.

I would say at minimum it depends on the resolution of the timestamps.  Also, how accurate the timestamps are to begin with...we have to remember timestamps are just another software functionality, and I personally have seen examples where they were *way* off (due simply to poor programming, in my experiences).  We used to say (of software) that "If it's capable of having the problem, then it's capable of causing the problem"

I'm just not familiar enough with DCS tracks etc to know how reliable and helpful it might be at that level of detail.

I will say this, though: improvement is improvement.  It gets harder to distinguish if it occurs in smaller increments...but if we want an example of the difference, all we need do is run the same tests on the system, once with the fastest Gen5 drive and one with a conventional hard disk.

That ought to answer any questions about the difference in performance, right there

Any other difference will be in terms of it being "worth" the cost...and, as above, that's an individual choice.

I personally cannot see spending the kind of money it takes to build a top end DCS machine, then choking it just to save <2-3% on storage.

Edited November 21, 2024 by kksnowbear

[kksnowbear]

PS I should say that, in my professional career, I have spent *days* analyzing all sorts of trace files, from machine language to ASCII text. I even designed equipment whose sole purpose was to monitor control systems and generate traces that could be used for troubleshooting. (Still do some of this now, just not as much...supposed to be "retired" lol)

So I know it can be very useful- but again, it depends entirely on how good the software is.  You have to be really sure your data is good before you start drawing conclusions based on it  

Edited November 21, 2024 by kksnowbear

[Mr_sukebe]

Just booted up DCS using the Gen5.

On the Samsung 990 Pro (Gen 4), it took 45 seconds.  Now, 42 seconds.  So a blazing fast improvement...NOT!

The above boot up, used the same Saved Games folder, and was conducted as per the previous test, i.e. just after a restart of my PC, and with the same background software running, e.g. Oculus, Simshaker etc.

Haven't tried it in game

Being fair, can't say that I'm bothered.  My son and I both have PCs and this frees up another 2TB unit that I can pass on to him when he's ready.

[kksnowbear]

1 hour ago, Mr_sukebe said:

Just booted up DCS using the Gen5.

On the Samsung 990 Pro (Gen 4), it took 45 seconds.  Now, 42 seconds.  So a blazing fast improvement...NOT!

To be accurate, that's actually around 7% improvement.

However, it bears considering that much of what is being done while "booting" (starting) the game has nothing to do with drive speed (or even the drive at all).  For example, a faster drive obviously cannot increase the speed at which your CPU (and RAM, for that matter) actually processes everything being done when the game is started.

So the 7% overall difference in time might just as easily be a substantial improvement.  If the entire amount of time required for starting the game was 100% drive activity, then the improvement is 7%.  If, however (strictly as an example) the amount of time is composed of 80% CPU time and 20% drive time, then a faster drive will have zero impact on the 80% CPU activity - it will only affect the 20% that the drive is responsible for.

So, using your 45>42 second example, and assuming 80/20 "split" as I described, then:

Of the 45 seconds total, 36 seconds (80%) are not affected by any drive changes (and cannot be, since they are not caused by the drive to begin with).  The remaining 9 seconds (20%) are subject to improvements in drive performance.  Thus, if that 9 seconds is reduced by 3 seconds, that's an actual improvement of 'in-game' drive performance of over 33%.

IOW you're now spending the same 36 seconds waiting on your CPU and RAM (as would be expected, since they didn't change at all), and 6 seconds waiting on the drive.

Of course, this is just an example - although I'd speculate it might not be terribly far from what's really happening.  In any event, I'd wager the 'startup' of the game is much more CPU time than drive time - still, just a guess.  Could be more than I'm guessing.

Also, there are also other factors:  What slot on the board is being used by the new drive (it matters).  Speaking of which, what board are you using?  CPU?  Was the test repeated to generate a reasonably average result?

Have you checked to make sure the drive is actually using four PCIe 5.0 lanes?

Just some things worth considering.  Hopefully it makes sense.

Edited November 22, 2024 by kksnowbear

[Mr_sukebe]

Got to admit, just kind of threw it in, quick format, copied DCS, repaired DCS etc.

It's definitely in the correct slot.  I have an MSI 670e tomahawk.  

I did see that there's an app to check whether I'm using the correct settings, I'll go look for that now.

As for the improvements, frankly, I'm not too bothered either way.  It wasn't that expensive, we'll always need the space and at some point, there's a high chance that DCS will take advantage of it.

[kksnowbear]

10 minutes ago, Mr_sukebe said:

Got to admit, just kind of threw it in, quick format, copied DCS, repaired DCS etc.

It's definitely in the correct slot.  I have an MSI 670e tomahawk.  

I did see that there's an app to check whether I'm using the correct settings, I'll go look for that now.

As for the improvements, frankly, I'm not too bothered either way.  It wasn't that expensive, we'll always need the space and at some point, there's a high chance that DCS will take advantage of it.

Well, just to be clear, it's my perspective that there is a need to be accurate - not that I'm faulting you, at all, and I understand you're not terribly concerned about it, which is entirely your prerogative, of course

The unfortunate reality, though, is that some others will read this, and will immediately pounce on "See, there's no real improvement" (which isn't really accurate).  So please understand that my observations are more about that than they are any reflection on you/your situation.

Still others might be wondering about the nature of the actual improvement, and I believe it's important to set expectations appropriate to fact.  Many might not actually realize the 'breakdown' applies as I've described - it's my experience that storage subsystems are greatly misunderstood, among even experienced gamers.

Crucial offers the "Storage Executive" for drive management.  Helps with other tasks, in addition to checking the slot lanes/PCIe revision.

I was surprised that board only has one PCIe 5.0 slot on it (between the CPU and GPU), but I suppose that pretty much makes it the "right" slot

[Mr_sukebe]

LOL, I'm glad you prompted me, thanks.

Sure enough, I needed to make a change within my BIOS, as it was running at Gen 4 speed.

Updated and now have a test from Crystalmark with a read speed of 14485 and a write speed of 12589.  

Re-ran the boot test.  45 seconds.  Probably worth assuming that there's a degree of error in maybe the config or similar.  As you've rightly noted, each of the above was a single run, not multiple to create a statistical sample.