Never ending wait for NVMe

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

BFG10K

Lifer
Aug 14, 2000
22,709
3,000
126
So by that same logic would it be fair to say that for the most part there won't be a noticeable difference between sata 2 and PCIe drives?
From a performance standpoint? Yes, absolutely.

Also are you aware of the difference between an AHCI PCIe drive and a NVMe PCIe drive?
The 750 is NVMe. You don't need to know the difference to read graphs showing very little difference in the real-world.

I'm still waiting for examples of these "instantaneous" game load times and boot times that have been claimed on PCIe, but allegedly elude SATA drives.

And again, my tests used a RAM disk. A RAM disk is a magnitude faster than any of your NVMe/PCIe drives.

Disk I/O stopped being the primary bottleneck in typical desktop situations when we dropped from 12ms access times (typical 7200 RPM spinners) to <1ms. First generation SATA2 SSDs can do that. Heck, el-cheap USB sticks can do that.

That’s why there’s virtually no difference in boot times across SSDs going back to 2008. Heck, even a 7ms Raptor competes quite well with boot times - only 4 seconds slower than the Intel SSD in my tests above.

After that, I/O makes no difference for general desktop tasks. Also many rigs with these drives tend to have huge amounts of RAM, which means the file cache further removes I/O speed from the equation.
 

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
From a performance standpoint? Yes, absolutely.


The 750 is NVMe. You don't need to know the difference to read graphs showing very little difference in the real-world.

I'm still waiting for examples of these "instantaneous" game load times and boot times that have been claimed on PCIe, but allegedly elude SATA drives.

And again, my tests used a RAM disk. A RAM disk is a magnitude faster than any of your NVMe/PCIe drives.

Disk I/O stopped being the primary bottleneck in typical desktop situations when we dropped from 12ms access times (typical 7200 RPM spinners) to <1ms. First generation SATA2 SSDs can do that. Heck, el-cheap USB sticks can do that.

That&#8217;s why there&#8217;s virtually no difference in boot times across SSDs going back to 2008. Heck, even a 7ms Raptor competes quite well with boot times - only 4 seconds slower than the Intel SSD in my tests above.

After that, I/O makes no difference for general desktop tasks. Also many rigs with these drives tend to have huge amounts of RAM, which means the file cache further removes I/O speed from the equation.


Obviously the disk isn't the only factor, like you mentioned before games have to be processed by the CPU and GPU and loaded into memory so yeah it's not going to be instantaneous, that is too big a claim to make.

I'm intrigued, your boot time tests were they done after a fresh install or after say a year of windows usage, programs installing and things generally slowing down to the point that even with a fast CPU, a windows 7 install will take 4 mins to load on a HDD?

Even with SSD's there is a slowdown, it's rarely noticed because it gradually creeps up on you with time. It's akin to living with a pet an not noticing it grow, but periodically visit someone else's pet and you notice the difference. I certainly noticed the difference when I freshly installed windows recently on the same SSD. A lot of the slow down is going to be due to the comparatively slow way that small data like 4k read and writes are handled, as well as reading multiple things at once.

AHCI is there to cope with both HDD's and SSD's, it has to do around 30 commands for each read and write. NVMe on the other hand is designed purely for SSD's and the amount of commands it needs for the same process is around 6 iirc. This is only one of the many enhancements. Another is the way it handles multiple requests.


HDD's can manage between 0.2 and 0.3MB/s for single queue 4k reads. Most SSD's can do 100 times that which is where he biggest performance increase comes from. The reduced latency with NVMe will mean a 50% increase in worst case scenario performance compared to the best AHCI SSD.

In my case it will be double the performance, 2.5 times what I have been used to with the slight constraints that Sata2 meant on my old rig (20B/s vs 25MB/s on Sata 3).

So for me it really makes sense to make the most of the M.2. connection in the z170 chipset now that NVMe dries are affordable and not just limited to expensive PCIe cards. It is the next big leap in storage technology.
 
Last edited:

larryccf

Senior member
May 23, 2015
221
1
0
I have hard numbers to back my claims when testing RAM disks - far faster than any of these devices.

Can you please tell me which of your games load &#8220;instantly&#8221; on PCIe but didn&#8217;t on the SATA SSD? Here are my tests:



Looking at the &#8220;RAM&#8221; column, you&#8217;re looking at a best case scenario of ~13 seconds because I/O is no longer the bottleneck. 13 seconds is not &#8220;instant&#8221;. GPU device drivers start to have far more effect on game load times, which has nothing to do with I/O.

I tested around 50 games and those shown were the only ones that had any meaningful difference; the rest got barely any improvement, even from a RAM disk. If you think your PCIe drive is going to load things faster than a RAM disk, you&#8217;re dreaming.

Likewise, we&#8217;re told boot times are allegedly faster. What&#8217;s particularly ironic is in reality these drives can lengthen boot times because POST takes longer, so this expensive device actually makes you wait longer to get into your system:





I&#8217;d be particularly pissed off if I'd just purchased an expensive 750 and was seeing those results. In fact there was another thread a little while ago with someone doing just that.

Would you care to point out where those &#8220;instant boot times&#8221; are on the PCIe drives? I can&#8217;t see them anywhere above, even with the second test which multi-tasks four startup applications at once.

With the other two drives, they&#8217;re 3 seconds faster than the next fastest SATA drive. So you paid 80c/GB to shave off 3 seconds? Cool story, bro.

Numerous other benchmarks show the PCIe drives being no faster than even an X25-M in the real-world: http://techreport.com/review/28050/intel-750-series-solid-state-drive-reviewed/5

The claim that these drives in the real world are just like going from HDD to SSD is absolutely laughable. The situation is akin to when overclockers claim "it feels faster" when twiddling with RAM settings that don&#8217;t make a lick of difference in the real world. Same thing back in the day when people were trying to push expensive quad-cores over dual-cores (&#8220;it feels smoother!&#8221

Again, real-world results don&#8217;t back those pretty synthetics. Sure, corner cases are primarily bound by I/O (e.g. copying files all day, certain content creation) , but the vast majority of the real-world won&#8217;t show any meaningful difference.

More evidence:
https://www.techpowerup.com/reviews/Intel/750_SSD_1.2_TB/11.html
https://www.techpowerup.com/reviews/Intel/750_SSD_1.2_TB/12.html
https://www.techpowerup.com/reviews/Intel/750_SSD_1.2_TB/13.html




well for starters, you're asking me about statements i haven't made. 1) i have NOT made any stmts about loading games, but about loading heavy programs like adobe photoshop, and i stated the improved loading times.

2) i haven't made any stmts about boot times being faster as i haven't bothered measuring boot times.

but rendering video files, or converting a video file from one format to another, using Handbrake, a 34 GB video that takes 65-75 minutes using a sata SSD (samsung 840 evo) (rendering from one sata SSD to a 2nd, both new 840 EVOs, drops to 25-35 minutes on an xp941. That drop in time to render i noticed without even looking for - when i did, i then did a comparison. That drop in times reconciles with the increase in write speed of 860 MB/s vs the 510 MB/s write speed of my 840 EVO.

That translates, for me, into rendering 3-4 BD movies in an evening vs 2 previously.

Those improved rendering times are not a figment of my imagination but i am not computer literate to reconcile them with your stmts - all i can do is report them. And for the record, cpu usage shows at 99-100% whether rendering from sata SSD or PCIe M.2 SSD

For the record my system is:
I7-4790 cpu
G.Skill DDR3 1600, 16 GB
Asus Nvidia 750 GTX Ti
Asus Z97M-Plus mobo
 
Last edited:

rumpleforeskin

Senior member
Nov 3, 2008
380
13
81
That drop in times reconciles with the increase in write speed of 860 MB/s vs the 510 MB/s

There must be more to it than a faster SSD write to half your encode time, maybe some changed encoding settings?

Unless your not recoding video then drive speeds will be the only limit? but then just transferring from BD would be limited by the BD read speed (fastest I can get is about 70 MB/s)

Even your slower SSD at 510 MB/s should spit out a 34gig file in a couple of mins assuming the SSD was actually the slowest point in the process.

So confusing
 

larryccf

Senior member
May 23, 2015
221
1
0
There must be more to it than a faster SSD write to half your encode time, maybe some changed encoding settings?

Unless your not recoding video then drive speeds will be the only limit? but then just transferring from BD would be limited by the BD read speed (fastest I can get is about 70 MB/s)

Even your slower SSD at 510 MB/s should spit out a 34gig file in a couple of mins assuming the SSD was actually the slowest point in the process.

So confusing

No settings were intentionally or knowingly changed. In fact i used what i call the "fred flinstone" approach to installing the OS on the xp941 - i cloned it from the sata SSD, then used the windows dvd to "repair" the boot files that were missing. I haven't noticed any settings changed but not being that computer literate, i suppose that possibility remains possible.

I'm not rendering direct from the BD - i copy from the BD disk using MakeMKV, then use Handshake to convert to mpg4-H264. The MKV file is saved to a sata SSD, then rendered from there to the xp941
 
Last edited:

bradly1101

Diamond Member
May 5, 2013
4,689
294
126
www.bradlygsmith.org
No settings were intentionally or knowingly changed. In fact i used what i call the "fred flinstone" approach to installing the OS on the xp941 - i cloned it from the sata SSD, then used the windows dvd to "repair" the boot files that were missing. I haven't noticed any settings changed but not being that computer literate, i suppose that possibility remains possible.

I'm not rendering direct from the BD - i copy from the BD disk using MakeMKV, then use Handshake to convert to mpg4-H264. The MKV file is saved to a sata SSD, then rendered from there to the xp941


I think you're seeing the benefits of sustained read/write speeds. I know the consensus around here is that the true speed of the drive for "normal" tasks is the single queue depth small (4k) read/write. Although that technically may be true, there are some like you who run things differently than the "most telling metric." What's the queue depth when I'm working on a RAW file with the Windows PageFile, the program's cache, and the program itself installed on the NVMe drive?

I guess all SSD's give us blistering performance, but if one type has an advantage in a task like yours (which reflects benchmarks) then that's great, and others may learn from your experience.

Arguments like this are strange to me. Most, if not all, of us buy the best option in our budget even though it's only marginally (or much more in your scenario) better than the next available piece of hardware. I guess those of us with these drives should feel ripped-off?
 

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
The AHCI model. Isn't the whole point of this thread that the NVMe model isn't available for sale? I've had for less than a week.

I have seen the 128GB and 256GB NVMe models pop up at Scan over the last week but they disappear quickly.
 

beginner99

Diamond Member
Jun 2, 2009
5,310
1,749
136
Not sure why anyone is waiting for these. Unless all you do is copy files all day, you won't see a lick of difference in the real-world compared to SATA SSDs.

If you already have an SSD, most of your operations aren't even bottlenecked by I/O speed anymore.

Wanted to say exactly this. I recently updated my ssd from intel G2 80 Gb to a MX-100. Yeah, MX-100 isn't the fastest drive out there but still a lot faster according to benches than the G2. But in real-life, there isn't any difference at all.

So you really need some really heavy IO to benefit from NVMe.
 

larryccf

Senior member
May 23, 2015
221
1
0
initial reports called for a "mid-july" release that never happened. That moved to a mid-august release, but no reports detailed anything specific re the 512 gb variant. A British website,
( http://www.scan.co.uk/search.aspx?q=samsung+sm951+nvme ) indicated they were taking orders on the nvme SM951 that seemed to evolve back to taking "pre-orders"

then the other day, johnnylucky, over on toms, made a cryptic stmt that those waiting for the nvme sm951 were in for a pleasant surprise in about a month.

I'd like to think that pleasant surprise means samsung is going to release the sm951 nvme with boot files as a retail unit, not an OEM, with after sale support direct from samsung
 
Last edited:

baydude

Senior member
Sep 13, 2011
814
80
91
I think that pleasant surprise is that when the SM951 NVME is finally released to retail, it has become obsolete and a better technology is scheduled to replace it soon
 

Redstorm

Senior member
Dec 9, 2004
293
0
76
It's still the biggest bottleneck on the system. Most modern SSD's can do about 30MB/s with single queue depth 4k random reads. These NVMe drives will double that performance. It will be more of a performance leap than say going from Sata II to Sata III.

it's also the reduced latency we're excited about, but then i'm someone who can spend hours testing all the minor latency controls for Ram (after Cas and Ras, TRFC has the biggest impact)

Agree, 4k random workload at low queue depths is where these shine, (client work load). Unless a competing product arrives soon, Revo400, Phison ps5007 to compete with the SM951 in the m.2 form factor, ill be grabbing a 512G model as soon as i can find one.
(Forget the Intel 750) too slow to enumerate on boot, and would be frustrating. Designed for a server workload and not a client workload.

Wanted to say exactly this. I recently updated my ssd from intel G2 80 Gb to a MX-100. Yeah, MX-100 isn't the fastest drive out there but still a lot faster according to benches than the G2. But in real-life, there isn't any difference at all.

So you really need some really heavy IO to benefit from NVMe.

Why buy obsolete tech when you can get on the NVMe band wagon.


The PM953 is an enterprise product and has worse performance than the SM951.

We originally posted above that the Samsung PM953 would likely be the successor to the Samsung SM951, but we were recently notified that that information is incorrect. The Samsung PM953 is designed to be for datacenter/enterprise applications while the SM951 will be for client applications.

That being said, we’ve heard rumors of an upcoming Samsung 3D V-NAND based PCIe client SSD that should be announced fairly soon. Samsung will be holding their annual Global SSD Summit at the end of next month, so with any luck we’ll get more information during the event. Be sure to check back for more!

Whether announced fairly soon means a paper launch or Available now, based on the availability of NVMe devices at the moment, the smart money would be on another paper launch followed by a frustrating wait for product to appear. I hope im wrong and they drop something awesome with 3D NAND and available immediately.
 

Redstorm

Senior member
Dec 9, 2004
293
0
76
Why spend a lot more money for something you get little to no benefit from?

Each to their own, its not a case of i need, its a case of i want. Same reason why i built an X99 system, i don't need it, i want it. Its a hobby to be enjoyed, and NVMe is the next toy on the list.

The trolls are lurking in here now, lobbing their troll bait "Ahh you wont see any performance difference, fools" and waiting for the unsuspecting to try and prove the troll wrong, which he is.
 

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
I don't think it's trolls, it just the psychology of not wanting to believe what they have invested in has been significantly superseded so their arguments will be slightly biased.

It's too early to really be able to tell and unless you see systems with both technologies side by side maybe most people won't notice the difference.

Personally I've been excited about NVMe for a couple of years now and having support for it was a factor in my decision for buying my new system, so my bias will be towards the hope it does make a difference.
 
Last edited:

iiiankiii

Senior member
Apr 4, 2008
759
47
91
Each to their own, its not a case of i need, its a case of i want. Same reason why i built an X99 system, i don't need it, i want it. Its a hobby to be enjoyed, and NVMe is the next toy on the list.

The trolls are lurking in here now, lobbing their troll bait "Ahh you wont see any performance difference, fools" and waiting for the unsuspecting to try and prove the troll wrong, which he is.

That's not a fair statement at all. Besides, copying large files, I hardly notice any difference between a gen 1 SSD (old school vertex 2) and the newer, faster SSD (Sandisk extreme). It just isn't too noticeable. I don't think people are trolling.

Plus, awhile back, I have tested RAM disk against SATA SSD. I noticed no difference in game load time at all. The time was nearly identical. Considering that RAM disk is way, way, way faster than NVMe drives (like 2 or 3 times faster), I don't expect NVMe to load games, and programs noticeably faster than SATA.
 

steve wilson

Senior member
Sep 18, 2004
839
0
76
That's not a fair statement at all. Besides, copying large files, I hardly notice any difference between a gen 1 SSD (old school vertex 2) and the newer, faster SSD (Sandisk extreme). It just isn't too noticeable. I don't think people are trolling.

Plus, awhile back, I have tested RAM disk against SATA SSD. I noticed no difference in game load time at all. The time was nearly identical. Considering that RAM disk is way, way, way faster than NVMe drives (like 2 or 3 times faster), I don't expect NVMe to load games, and programs noticeably faster than SATA.

I can't believe a RAM disk is not faster than a SATA drive.
 

shabby

Diamond Member
Oct 9, 1999
5,782
45
91
That's not a fair statement at all. Besides, copying large files, I hardly notice any difference between a gen 1 SSD (old school vertex 2) and the newer, faster SSD (Sandisk extreme). It just isn't too noticeable. I don't think people are trolling.

Plus, awhile back, I have tested RAM disk against SATA SSD. I noticed no difference in game load time at all. The time was nearly identical. Considering that RAM disk is way, way, way faster than NVMe drives (like 2 or 3 times faster), I don't expect NVMe to load games, and programs noticeably faster than SATA.

I think these reading or writing tasks are too mundane to notice the difference.
In my situation i read/write and download at the same time and notice my downloads actually slowing down to a crawl(from 330mbps down to 50-100mbps) because my intel 330 is too busy with the other read/write task.
Im sure i'd see a benefit in this situation with these new pcie ssd's.
 

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
That's not a fair statement at all. Besides, copying large files, I hardly notice any difference between a gen 1 SSD (old school vertex 2) and the newer, faster SSD (Sandisk extreme). It just isn't too noticeable. I don't think people are trolling.

Plus, awhile back, I have tested RAM disk against SATA SSD. I noticed no difference in game load time at all. The time was nearly identical. Considering that RAM disk is way, way, way faster than NVMe drives (like 2 or 3 times faster), I don't expect NVMe to load games, and programs noticeably faster than SATA.

What are you copying to and from?

Interested to know what games you tested.
 

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
I think these reading or writing tasks are too mundane to notice the difference.
In my situation i read/write and download at the same time and notice my downloads actually slowing down to a crawl(from 330mbps down to 50-100mbps) because my intel 330 is too busy with the other read/write task.
Im sure i'd see a benefit in this situation with these new pcie ssd's.

In the mean time you could move your downloads folder to another drive.

Go to C:Users/<username>

right click/properties on downloads, go to location and move it to another disk.
 

iiiankiii

Senior member
Apr 4, 2008
759
47
91
I can't believe a RAM disk is not faster than a SATA drive.

I didn't say that. I said in game loading scenerioes, I hardly noticed the difference. Maybe a second or 2. Nothing where you went, "wow, that was a lot faster." Not like the difference I notice from the switch from HDD to SSD.

Of course a RAM disk a lot faster than SATA SSD and NVMe. It just isn't noticeable enough with what I do with the PC. You would think the RAM disk would be blazing fast in real world situation but it wasn't much faster than an SATA SSD. Dont' believe me? Go test it.

The only situation where having high end SSD really seems to matter is video editing, and/or anything that is moving large files around.

What are you copying to and from?

Interested to know what games you tested.

For copying large files, I have 2 SSD to copy to and from. As for the games tested on a RAM disk, I haven't tested it in a while. The games I tested were older games like Skyrim, Starcraft 2, Counter Strike.

I might try again on newer games to see if the bottleneck shifted back to the SSD on game loads where I noticed a difference between a RAM disk and SSD.
 

john3850

Golden Member
Oct 19, 2002
1,436
21
81
Two years ago I set up a ram disk called Fancy Cache the name has changed since but it is still around.
I set it up with a few ssd durning normal game play loads were the same.
When I loaded saved games the loading time was instant but only on large saved games.
 

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
For copying large files, I have 2 SSD to copy to and from. As for the games tested on a RAM disk, I haven't tested it in a while. The games I tested were older games like Skyrim, Starcraft 2, Counter Strike.

I might try again on newer games to see if the bottleneck shifted back to the SSD on game loads where I noticed a difference between a RAM disk and SSD.

Were they identical pairs of SSD's? If you are copying from one model to another you are going to be limited by the respective read and write speed.

Of those 3 games from what I remember Skyrim and counterstrike loaded pretty quickly on an HDD.

Could you try Dragon Age Inquisition?
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |