900p impressions

[mackerel]

I got the 900p 280GB M.2 version. It is actually a 2.5" drive which has a U.2 connector, and also in the box is a cable that goes from U.2 to M.2. Note the 2.5" drive is thicker than common modern drives. Anyway, I also bought a M.2 to PCIe adapter card as I thought I might need it in case I decided to copy between M.2 drives in systems without multiple connectors.

I wanted to test two things:
1, does where you connect it affect performance? M.2 slots, chipset PCIe, CPU PCIe?
2, does the CPU speed affect results? This follows on from an Intel sponsored white paper by Shrout Research, showing that CPU power saving can impact performance.

Key parts of test system:
Asrock Z370 Pro4 bios 1.3
2x8GB Samsung B-die ram at 3600C16
i3-8350k at 4.0 or 5.0 GHz.
R7 260X
Win10-64 FCU

Tests run were CrystalDiskMark 6.0.0 x64 and AS SSD 2.0.6485.19676. Each was run 3 times, and the best of the 3 used for comparison. Default settings were used.



Here's an example CrystalDiskMark result. Of interest to me is the 4k QD1 read rate, shown here around 281MB/s! This compares to my best flash SSD, an Evo 960 500GB, which gets around 55MB/s in that test.

I have a table of results, and while I think of how to present the results, I'll just write some observations on where the difference are.

Test configurations:
1, 900p connected to mobo M.2 connector at bottom of board
2, 900p connected via PCIe adapter to chipset PCIe slot
3, 900p connected via PCIe adapter to CPU PCIe slot (GPU moved temporarily to chipset slot)
4, same as 3, but with CPU OC 25% to 5.0 GHz (also used for above CDM result)

Comparing 1 and 2, there is no significant difference (<5%), which could be expected as they're both going through the chipset.
Comparing 2 and 3 is more interesting. Sequential reads consistently dropped over 5% on both. There is also some similar effect for high QD/transfers, but to lesser extent. Randoms did increase in performance, low single digit % for reads, up to double figures for writes.
Comparing 3 and 4 didn't show a change in sequential speeds, but it did bump up random reads and writes. CDM Q32 increased 14/17% respectively for reads/writes. Q1 around 5% for both, AS SSD 7% for both. Optane is so fast your CPU might be a bottleneck!

Back to my original questions, for the first the answer is yes, but I'm not clear why. Randoms are more latency sensitive, and by not going through the chipset it likely gives a slight edge to performance. I'm not sure why sequentials would actually drop. For the 2nd question, that's also another yes. Faster CPU = faster transfer rates. Combine the two, comparing the onboard M.2 connector, to a CPU connected PCIe slot, biggest gains are T1 randoms in CDM, with 12-20% increase. AS SSD similarly 9-13%. So... not insignificant!

 

[Robert17]

This parallels some reviews I'd read; the Q1 is insane. But there is where most activity occurs, at least according to data center records. I'm imagining a convoy of tracker/trailers of these things heading to Amazon if not already arrived at their regional datacenters.

 

[EarthDog]

These were already coming in on preconfigured racks.

 

[mackerel]

Just installed it into my main system. A single run of CDM gave random Q1 reads around 280MB/s with it connected to a CPU PCIe slot, so similar to what I was seeing on the Coffee Lake system. In initial testing there doesn't seem to be a noticeable difference to the GPU which is now only running at 8x. Done a single run of game loading benchmark so far. My old SATA SSD, where the game actually was used off, took just over 18 seconds. With the 900p it is down to 12 seconds. For comparison, past testing on a ramdisk was around 9 seconds, and a 960 Evo 15 seconds. You might think that's a lotta cash spent to gain a few seconds, and you're probably right. The thing is, as an MMO there is a fair bit of loading going on, so it is a long term thing. Save a few seconds here, a few seconds there, it'll add up in the longer term.

I'm about to have a short break from the screen before I give it a good workout.

 

[HankB]

Nothing wrong with getting a little more speed, maybe even the fastest one on the block (for now. )




When I was a young man the cool thing to do was to soup up cars. Nowadays some of that aggressiveness has moved to PCs and it seems to me that it is a little less expensive and a whole lot safer.

 

[SPL Tech]

I'd like to see how this compares to the Evo 850 SATA for real world usage. Windows loading times, game loading times, ect. Synthetic benchmarks are meaningless for virtually everyone and they dont demonstrate anything interesting useful for real world use. I want to know will upgrading make my computer faster, yes or no.

 

[mackerel]

My drive only contains one game install, the one I wanted to speed up. Without a stopwatch, I can't say I feel a noticeable difference. Based on earlier testing for the game, I did find an influence on the rest of the system for load times also, suspecting both CPU and GPU contributing to some degree. With a 1080Ti I'm not going to get any faster on that, but next I wonder if my stock 6700k is getting a bit old to keep up with this... longer term I still want to move to the 7800X as my gaming system but it'll probably be obsolete before that happens.

 

[Woomack]

If you enable full caching on drive ( sometimes it's disabled, sometimes it needs good driver ) and OC CPU cache+memory then you may notice faster access time and read bandwidth. On some SSD i see up to 10-15% improvement. I guess it still won't give you any significant difference but can play with settings

 

[mackerel]

Not sure which cache setting you're referring to. My main system already runs dual rank 3200 so I'm happy with the bandwidth of that, but I'm not currently OCing the CPU after I backed off for summer. I should revisit it some time... not sure I'll bother, since my plan was to use a separate system dedicated to gaming. Still undecided if I'll use the 7800X, or if it takes me much longer, the 8700k might come actually be in stock at sane price and I'll get that.

 

[Woomack]

Drive/controller cache in hardware manager. Also when you install drivers then driver sometimes disables cache. I guess you have that enabled but can check if there are any new drivers.
I guess you can get higher results using faster CPU but 6700K isn't slow so hard to say without tests. Integrated storage is using CPU to perform all operations so faster is always better but not always much faster.

Btw. I wonder what with Micron as they were designing xpoint memory with Intel and they have no NVMe SSD on the mass market. There were drives in plans for middle of 2017 and I even had sample reserved but it wasn't released at all.

Still it's amazing improvement in low queue 4K comparing to last gen of SSD. I wanted to order one but when I checked prices then I guess I will wait for something else.

 

[mackerel]

Write cache is off, but since I'm interested in reads I'm not sure that'll help. It'll be using standard system drivers... just another PCIe storage device. Reason for wondering if even CPU might help is that on the 8350k I saw a difference, about 260MB/s at 4.0 GHz, and 280MB/s at 5.0 GHz. Now, I'm getting around 280MB/s already on the 6700k at 4.0 GHz... too many variables, I don't know what else might be going on.

One thing I want to do is try short term is moving it to chipset PCIe slot. This might reduce performance, but I have to decide what to prioritise here. I've been observing the GPU bus usage during gaming, and I think I've had a bit of a fps drop. Bus usage is going up to around 11% or so which seems excessive to me. If it is simply bandwidth limited, restoring it to 16x should about halve that. I suspect it might also show up in GPU benchmarks but I haven't done that yet. Longer term, I think I should move it to the Skylake-X system along with my 1080Ti as there's more lanes to go around there (looks like it can do 16+8 off CPU). If I do that, I'll give up on getting the 8700k as I've spend enough for this year! I have some vacation days I have to use up by end of the year so I'll possibly take a long weekend some time and do it.

Oh, as often the case I've paid an early adopter tax on the 900p. When I ordered it, it was £400, but some models have already dropped to £375 now supply is more common in UK.

 

[SPL Tech]

Yes well all the numbers are great but if it doesent actually make the computer faster it's meaningless. That's like saying my car gets 100 MPG, just never when I actually drive it. What's the point? If the tech wont actually do anything for you, then it's worthless tech. Unless you buy SSDs to run benchmarks all day, it serves no purpose. We want a product that will actually make the computer faster, not just produce some cool benchmarks that dont translate to anything meaningful.

 

[EarthDog]

I can esaily notice a difference between a standard SATA SSD and NVMe M.2.... It's not like HDD to SSD of course, but, hell, my wife even noticed and had no clue I swapped drives out.

 

[habbajabba]

Real world differences to me would be no moving parts, portable (as in a tablet or other), with top-end desktop power and speed. It's simply not for sale near me. Gaming is pointless in terms of real word usage.
The 900P is a rather nice drive I must say. My 840 evo still works fine for my needs. I'll probably at some point just turn it into a server anyways, and because it won't be a commercial enterprise, my evo would work fine for that too.

 

[Woomack]

Personally I see no difference in games or most other programs between SATA SSD and NVMe. I have couple of computers, different SSD and I was moving drives between them many times. I simply see no difference so I'm using NVMe only for benchmarks and reviews. It's probably because of nearly the same random bandwidth of all of these drives. 900p has much faster random 4k than most other NVMe SSD so maybe I would see the difference but I won't pay so much for SSD. Current price is ridiculous.

 

[mackerel]

I'd go as far as to say it is expensive (relative to NVMe flash), but not sure I'd use ridiculous. It is a new class of device. Pay more, get more. If it is worth it to you is a different matter. I still haven't done any of the extra things I wanted to do yet... I'm too lazy to migrate my gaming to another system so I'm still on my old 6700k, although half keeping an eye open for any 7700k's on sale as drop in replacement. I note 1800X on sale are now cheaper than my 1700 was when I got it! Shame it is useless for high fps gaming.

For now, I still can't say I feel a difference in the game that I've installed on it. Benchmarks say it is faster, but I can't feel it, in part because there are limits elsewhere and I suspect both CPU and GPU also contribute to load times.

 

[Woomack]

In local stores I see it 2x more expensive than competitive drives in similar capacity so it's far from reasonable price. Most users, if not something like 99.5%, won't see any difference so yes, these SSD are in ridiculous price for a typical user. Price can be reasonable for some users who really need and can use so fast random 4k read. On the other hand I bet that soon we will see other drives based on similar chips. 3D xpoint is not only Intel creation as other brands were also designing it. Micron was first and next they were working together with Intel on this project but I have no idea why we haven't heard anything from them for a while and they canceled their consumer grade NVMe SSD.

I guess you will see huge difference on big databases. Most other things won't show the difference.

 

[mackerel]

The point is you buy what is good for your needs. Yes, this is expensive, and for that money you get high performance in two areas (randoms, endurance). Don't need it, don't buy it. They're not intending to replace everything with it. Micron I think I've seen have started doing enterprise models. If they ever move it to consumer it'll be under their Crucial brand anyway. What else is there for competition? The Samsung thing is the nearest, but they already have said it isn't quite as fast, so they will have to be competing on price. And who knows how long before we see that in a shop.

 

[Alaric]

that aggressiveness has moved to PCs and it seems to me that it is a little less expensive and a whole lot safer.

caddi daddi might have a counter for that...

Gaming is pointless in terms of real word usage.

I game in the real world. All my stuff is here, anyway, so it seems real to me.

OC CPU cache+memory then you may notice faster access time and read bandwidth.

Is there much benefit to OC the CPU cache?

 

[caddi daddi]

caddi daddi might have a counter for that...

Let's put that to the test, shall we?

when I spill my booze in the car I get butt rot.... that's ok, I did the dead, I pay the price, no harm done.

when I have spilled my booze down the computer, fire has ensued, this has proven to be some what less than good, quite exciting, but still not good and, as a rule the computer is junk and my little bee is most upset with me............

Lets put numbers on the fun. speed is fun so let's use that.

my computer tends to make a very solid 0 miles per hour and cost around $3000, so, 0x 3000= 0, so 0 funs there, unless I'm mad at it and throw it out the window, about 5 mph. 5 x 3000= 15000 funs and a few less than fun fires but 15000 funs.

now, my cute, little fire breathing Lancia, Beta, Monty Carlo with 790 rw hp cost about $100,000 so far and makes a solid 190 mph our last outing at savanna so, 190x 100,000= 19,000,000 funs and so far no fires, other than a few blown engines as we made our way with the little 1.5 litre unit. thus I have mathmaticly proven beyond any doubt that cars are funner and safer than computers, and who really minds a little butt rot from good booze? it's better than from sweat!!!!!!