SuperSSpeed SLC VS. Anything else

[Gilgamesh]

New Samsung 840 Pro 256GB

http://www.custompcreview.com/review...eview/15768/3/

[B Gates]

Very impressive but I can almost guarantee you that when tested with data on the drive and using non-synthetic methods the S301 will come out on top

[Gilgamesh]

I'm sure that 128GB (275$) with 50% of data on the drive can always beat an 840 Pro 256GB (270$) with 25% of data.

But 75% vs 38%?

Are you preparing reviews of Samsung 840 Pro, and I hope, an 840 TLC nands?

[B Gates]

I probably wont sample that drive until its released. I'm not sure what you mean by 75% vs 38%? If you mean do I think the S301 75% full has better performance than a 840 Pro at 38% then I would guess that it would but can't be sure until i test it but I would bet the 840 will fall short.

[B Gates]

here is a review to look at http://www.storagereview.com/samsung_ssd_840_pro_review notice workstation performance because that is similar to a typical user environment. The 840 is probably not going to do that great when its 50% full.

[Zaxx]

Can't wait for the SuperSSpeed 60s to make it out...at probably ~$160 each, I'd muuch rather have 2 of these in R0 vs a single 120GB for a slight premium, heh...who wouldn't. Will be interesting to see if Intel's latest process shrink yields lower prices...wonder if any 20nm SLC has even surfaced yet.

[Highendtoys]

Samsung didn't send anyone who tests with data on the drive. My drive(s) come in two weeks and I'm sure Jon will have some as well.

I'm tired of seeing people post bull articles and YouTube videos of SSDs with data that doesn't mean .

[m.oreilly]

Hard for me to see past the price quite frankly..

bill, doo makes a good point, since your outfit are selling these drives, if i'm not mistaken. bill, it looks like you are espousing your drive as a 'Ferrari among bricks'. sounds like a barker's hyperboli, of items you are selling. is this allowed?

[B Gates]

bill, doo makes a good point, since your outfit are selling these drives, if i'm not mistaken. bill, it looks like you are showing 'superior benching', along with your very questionable hyperbole, of items you are selling. is this allowed?

Hi Mike,
I'm not selling any drives nor is Chris they are being sold by MyDigitalSSD. It's true that they are signature editions signed by the person I work for but that has nothing to do with the drives superior performance. The reason why this drive is superior is simply because it's the only drive with SLC NAND. Chris could have just as easily picked any other drive to do a signature edition of but this drive is a stand out. The drive was benched exactly the same as any other drive we bench and its superior because it proves itself to be, not because of perceived skewing in its favor. We believe a drives performance is best determined by how it performs with data on it. This makes sense doesn't it? I made this thread for the purpose of examining whether or anyone such as yourself who may want to discuss our findings or dispute them in any way could do so directly and easily. The SLC drive is simply faster than any MLC drive to date. I would NEVER say it was if I didn't have proof to back it up and believe it to be so, even if it was my mothers signature edition. If you believe my assessment is incorrect then this is the place to prove it. If you can prove that I will gladly acknowledge that and openly declare my assessment to be incorrect. One other point. I purchased 2 of these drives these were not samples. I spent my hard earned money on them because to me they are well worth it.

[Ourasi]

Hi Mike,
I'm not selling any drives nor is Chris they are being sold by MyDigitalSSD. It's true that they are signature editions signed by the person I work for but that has nothing to do with the drives superior performance. The reason why this drive is superior is simply because it's the only drive with SLC NAND. Chris could have just as easily picked any other drive to do a signature edition of but this drive is a stand out. The drive was benched exactly the same as any other drive we bench and its superior because it proves itself to be, not because of perceived skewing in its favor. We believe a drives performance is best determined by how it performs with data on it. This makes sense doesn't it? I made this thread for the purpose of examining whether or anyone such as yourself who may want to discuss our findings or dispute them in any way could do so directly and easily. The SLC drive is simply faster than any MLC drive to date. I would NEVER say it was if I didn't have proof to back it up and believe it to be so, even if it was my mothers signature edition. If you believe my assessment is incorrect then this is the place to prove it. If you can prove that I will gladly acknowledge that and openly declare my assessment to be incorrect. One other point. I purchased 2 of these drives these were not samples. I spent my hard earned money on them because to me they are well worth it.

Overall performance, you might be right that the SLC SSD is faster then the Samsung 840P especially server workloads etc, and sertainly in a non-TRIM OS as SLC rewrites much faster then MLC does. I've had a few X25-E's in my time, so I know the benefit of SLC and I would still own them if Intel did not screw me on TRIM..

However, an area the Samsung 840P looks to be faster and pretty substancially so, is random read QD(1)2-6, this is where my priority are as always. RR QD2-6 will not loose steam on the samsung when it's filled to an reasonable level (XS OS levels, sub 50-60%), Sandforce on the other hand do loose abit of steam on RR QD2-6 as it fills up to the same levels, don't know if this SLC edition is more resilliant, and won't find out myself as the SLC drive is nowhere to find in norway..

Would have liked to see the SuperSSpeed use another controller like the latest marvel, I think that would have been even more impressive, and it's not to late for that..

As for the linking between the reviewsite and this SSD in this manner, it might seem a little iffy imho, add the SuperSSpeed "talkup" by you, working for the same site, and a few eyebrowes might be raised on XS No pun intended, just voicing my opinion, and I tend to love me a good conspiracy in the late night hours..

As for the SSD itself, it would, for that price, probably be in my box already if I could get one here, but the Samsung 840P will be here on the 15th and I want one.. I'm wondering why an SSD-maker like this again do not get it into the shelves in norway, we are buying SSD's like butter here, and have a few bucks to spend (!). This is a fiering offence by the resposible party imho...

[B Gates]

Overall performance, you might be right that the SLC SSD is faster then the Samsung 840P especially server workloads etc, and sertainly in a non-TRIM OS as SLC rewrites much faster then MLC does. I've had a few X25-E's in my time, so I know the benefit of SLC and I would still own them if Intel did not screw me on TRIM..

However, an area the Samsung 840P looks to be faster and pretty substancially so, is random read QD(1)2-6, this is where my priority are as always. RR QD2-6 will not loose steam on the samsung when it's filled to an reasonable level (XS OS levels, sub 50-60%), Sandforce on the other hand do loose abit of steam on RR QD2-6 as it fills up to the same levels, don't know if this SLC edition is more resilliant, and won't find out myself as the SLC drive is nowhere to find in norway..

Would have liked to see the SuperSSpeed use another controller like the latest marvel, I think that would have been even more impressive, and it's not to late for that..

As for the linking between the reviewsite and this SSD in this manner, it might seem a little iffy imho, add the SuperSSpeed "talkup" by you, working for the same site, and a few eyebrowes might be raised on XS No pun intended, just voicing my opinion, and I tend to love me a good conspiracy in the late night hours..

As for the SSD itself, it would, for that price, probably be in my box already if I could get one here, but the Samsung 840P will be here on the 15th and I want one.. I'm wondering why an SSD-maker like this again do not get it into the shelves in norway, we are buying SSD's like butter here, and have a few bucks to spend (!). This is a fiering offence by the resposible party imho...

take a look at post 3 which shows real world testing with data on the drive. look at how marvell drives fare. the samsung 830 256gb although not on those charts fares the same as Marvell drives do. What is your opinion of those results? I am just making a guess on the 840 but I suspect it will not score 69,000 at 50% full.

[Ourasi]

Vantage is fun to test, and fun to get insane results, and easy to test TRIM, but as an realworld bench it is useless. I've tested extencively if a drop in Vantage from 170k to 100k - close to half the speed according to that test - by filling the SSD's from 7% to 50%, coud be seen elsewhere, and nothing have ever come close to that, not even in the same ballpark.
Infact, I timed starting 30 apps@7% simultaniously vs. 50%, and the difference was 0.xx%. I tested all the IOmeter patterns, and the difference was almost the same as my timed results... Vantage is close to being useless from a realworld performance perspective on SSD's imho, as the guys who made it admitted. Mind you, PCM7 has gone to far, and are artificially keeping the scores between HDD and SSD to close, and are pretty useless aswell, infact even more useles as it can't be used for anything beneficial (like working TRIM) at all imho..

Edit: My spelling laying down in the dark on the sofa typing with the oddest of fingers, might produse some odd words, but I do see them and try to edit while rolling my eyes in shame and annoyance...

[B Gates]

Vantage is fun to test, and fun to get insane results, and easy to test TRIM, but as an realworld bench it is useless. I've tested extencively if a drop in Vantage from 170k to 100k - close to half the speed according to that test - by filling the SSD's from 7% to 50%, coud be seen elsewhere, and nothing have ever come close to that, not even in the same ballpark.
Infact, I timed starting 30 apps@7% simultaniously vs. 50%, and the difference was 0.xx%. I tested all the IOmeter patterns, and the difference was almost the same as my timed results... Vantage is close to being useless from a realworld performance perspective on SSD's imho, as the guys who made it admitted. Mind you, PCM7 has gone to far, and are artificially keeping the scores between HDD and SSD to close, and are pretty useless aswell, infact even more useles as it can't be used for anything beneficial (like working TRIM) at all imho..

Edit: My spelling laying down in the dark on the sofa typing with the oddest of fingers, might produse some odd words, but I do see them and try to edit while rolling my eyes in shame and annoyance...

what do you think of PassMark testing? I see what you are saying but thats not the point of Vantage fill testing or how it works. What you are saying is that if Vantage fill testing shows for example a 40% performance drop off when data is added then why dont we see a 40% slower boot time or 40% application loading time? Vantage fill testing simply gives us an accurate representation of one drives performance verses another with data on it not the percentage of performance lost as the drive is filled up. Fill testing with Vantage and ranking drives with its results consistently provides results that are corroborated by other non-synthetic methods used in the industry. Take PassMark for instance. PassMark creates a test file on the drive and plays it back reading and writing to determine performance. This is as real world as it gets. Passmark Workstation results are always 100% in line with Vantage Fill testing. All SSD's slow as data is added to them and some more than others. Vantage fill testing simply exposes realistically which drive actually has better performance in a measurable way with data on the drive.

[Ourasi]

I haven't testet it myself, but from what I can see, they have an IOmeter aproach to their testing, looks ok to me from a server perspective if it uses correct compression levels, allthough I do not agree that the Workstation pattern copy real world correctly if they do intensive reads and writes at the exact same time. The reason, is that a typical workstation may have the same daily pattern of the test's 80/20 R/W, but not R&W at the same time like I think they do here. A typical SSD freak like me, have RR(+some sequential read) QD#1-6 in short burst, all day, with a tiny amount of mainly 4-32kb RW QD#1-2 more or less always in the background. Some writes during work/surfing and other stuff (downloads goes to HDD storage), but seldom or never at the same time as the intensive read. This, in my case, amounts to ~90/10 R/W per day, so not far off the workstation 80/20, so in theory the test is correct, but not in execution. This is why the QD1&QD4 RR results of the 840P looks so tempting for my usage, writes RW&SW have been more then good enough for a long time among the good SSD's, and have lower priority for me, for an OS SSD... My old C300, wich up to now has been the king in RR QD1-6, is as fast or faster then a Intel 520 at boot and apploading, thanks to it's strengths at low QD RR. And, the Intel is twice as fast in vantage ..

The X25-E SLC was also a monster in workstation, so if heavy intensive reads during intensive writes is the usagepattern in question, SLC is sweet.. Thats why everyone should know their usagepattern before they buy...

[B Gates]

I haven't testet it myself, but from what I can see, they have an IOmeter aproach to their testing, looks ok to me from a server perspective, allthough I do not agree that the Workstation pattern copy real world correctly if they do intensive reads and writes at the exact same time. The reason, is that a typical workstation may have the same daily pattern of the test's 80/20 R/W, but not R&W at the same time like I think they do here. A typical SSD freak like me, have RR(+some sequential read) QD#1-6 in short burst, all day, with a tiny amount of mainly 4-32kb RW QD#1-2 more or less always in the background. Some writes during work/surfing and other stuff (downloads goes to HDD storage), but seldom or never at the same time as the intensive read. This, in my case, amounts to ~90/10 R/W per day, so not far off the workstation 80/20, so in theory the test is correct, but not in execution. This is why the QD1&QD4 RR results of the 840P looks so tempting for my usage, writes RW&SW have been more then good enough for a long time among the good SSD's, and have lower priority for me, for an OS SSD... My old C300, wich up to now has been the king in RR QD1-6, is as fast or faster then a Intel 520 at boot and apploading, thanks to it's strengths at low QD RR. And, the Intel is twice as fast in vantage ..

so are you saying that a C300 is faster that a 520 in real world?

[B Gates]

look at this drives 4k read


this is the fastest drive when reading compressible 4k but the drive is slow

[B Gates]

Can you show that a c300 is faster than a 520 at booting and app loading than a 520 because I would respectfully disagree. The c300 has terrible latency.

[Ourasi]

Your CDM picture is 0fill, 100% compressible, and it's the C300's QD#2-6 that matter the most, QD#1 is not as important. The 520 have both read and write latency at almost the same levels as the C300 when used a little, atleast in AS SSD..

C300 boot:

520 boot:


The difference is around a second+ on average, over 20+ boots on the same fresh image. I was abit dissapointed and surpriced to say the least, and the same story is told by the 30+ simultanoious apptest, allthough the difference there is smaller in %...

[B Gates]

What are you using to measure boot time? I would like to try it.

[Ourasi]

The Intel SSD 520 240gb scores 63mb/s vs. C300 256gb at 93.5mb/s in Random Read QD#3 4kb (Anandtech, and the C300 is up there with the best of the best even today, in this test), and as I've said, with my usage pattern, the C300 is better on average, the reason the 30 apptest is closer between them, is that there is some sequential reads in some apps, and the Intel is stronger there.

With this said, I do not mean the C300 is better for everyone compared to the Intel, I'm just saying it's better for me (and I might be one of not so many, as usual)..

[B Gates]

I enjoy discussing stuff with a person as well informed as you I really appreciate your time

[Ourasi]

Added Bootimer in my previous post, hope it works as it should, I know UAC might prevent it from rebooting, but my guess is that this "feature" is pretty much gone in your setup
Doubleclick, answer yes to reboot, wait for result after reboot without touching mouse & keyboard, enjoy..

[B Gates]

here's what I'm getting with slc sandforce
on sata 2 port

[B Gates]

If you don't mind my asking what did you use to test app launching?

[Ourasi]

Just manual timing from my shootingtimer, starts when i click the mouse on the .bat file and stops when i click the mouse when all apps are loaded. A little margin of error is there, since my reaction is slower then my shooting timer (that is within 5ms from i start the .bat) when clicking at the end. But on the other hand, starting 30 apps takes a bit of time, so a few hundreds is of little consequence when the difference is several seconds..

Good bootime BTW, you will notice that the full aero and GFX/sound drivers will slow the bootime, I have a 3.xxx time, but not fully operational and eyecandy OS was one of my rules in my SSD-bench thread, so I stick to my own rules... The thread btw is now at 200k views and 247 pages, so we norwegians do like our SSD's

You can try it yourself: start a notepad, add these lines:
start /d "C:\Program Files (x86)\Internet Explorer" iexplore.exe
start /d "C:\Program Files (x86)\Windows Live\Messenger" msnmsgr.exe
start /d "C:\Program Files (x86)\Windows Live\Mail" wlmail.exe
start /d C:\Windows explorer.exe

I guess all these are installed, remove those that are not. Save the notpad as Favorites.bat, make a shortcut on you destop or startmenu, and launch as a normal shortcut. This is the proper way to start apps on SSD's, not one at a time... The apps above is startet when im booted onto the desktop every day, the hassle of launching one app at a time is over for me...