Intel Replies to SSD 'Slowness' Critique

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
It would seem that the crew at PC Perspective have grabbed Intel's attention with their article on SSDs. Ryan and the PCPer crew found that the performance of a used SSD would always be slower than that of a new, non-fragmented drive. Intel replied with the following:

In response, Intel made a statement on Thursday. "Our labs currently have not been able to duplicate these results," Intel said. "In our estimation, the synthetic workloads they use to stress the drive are not reflective of real world use. Similarly, the benchmarks they used to evaluate performance do not represent what a PC user experiences."
Click to expand...
 
I think Intel is being fair in their statement. I agree that the loads that were being used in the benchmarks weren't similar to real world usage.
 
This pretty much echos what Kyle has been telling us. Synthetic benchmark means jack.
 
Hulahoops said:
That may well be, what do you think they tend to sell them on ;)
Click to expand...

Real world performance. Does my computer boot up faster? Does it render video faster in real world use? Do my programs load faster? How does it affect my end user experience?
 
the problem is real world performance is hard to duplicate scientifically.
my old AMD X2 4600 boots in a third of the time my Quad does, because of the motherboard/on board crap.

synthetics may not be real world, but they are *specific* which makes them useful in an albeit different way.
 
Well, I have had my drive for two weeks. After an install of Windows 7, and an install of Vista, and two weeks of normal usage, my write speeds declined from the speced 80MB/s to around 40MB/s. Reads were still okay, but those two weren't nearly as good. The average access time of the drive when new (and empty) was .06. After installing just LotRO, it was .07. When full it was .11ms.

I was able to locate the HDDErase v1.3 program and wipe my drive and the performance "snapped back" to its new state. Losing half my write speed after only two weeks of normal usage, however, doesn't seem appropriate. So in my own personal experience, I'd say that PCPer is onto something. And I have gotten the impression that Intel is privately a bit more concerned about this issue than their press release indicates.
 
Hurin said:
Reads were still okay, but those two weren't nearly as good.
Click to expand...
Bah. Too early to be writing. That should read "even those". . . meh.

I wish the News Commentary forum allowed editing. I'm sure there's a reason it isn't. But for the life of me, I can't figure out why.
 
Hurin said:
Well, I have had my drive for two weeks. After an install of Windows 7, and an install of Vista, and two weeks of normal usage, my write speeds declined from the speced 80MB/s to around 40MB/s. Reads were still okay, but those two weren't nearly as good. The average access time of the drive when new (and empty) was .06. After installing just LotRO, it was .07. When full it was .11ms.

I was able to locate the HDDErase v1.3 program and wipe my drive and the performance "snapped back" to its new state. Losing half my write speed after only two weeks of normal usage, however, doesn't seem appropriate. So in my own personal experience, I'd say that PCPer is onto something. And I have gotten the impression that Intel is privately a bit more concerned about this issue than their press release indicates.
Click to expand...

Interesting if true...
 
SirKronan said:
Real world performance. Does my computer boot up faster? Does it render video faster in real world use? Do my programs load faster? How does it affect my end user experience?
Click to expand...

Those are rarely the headline numbers any SSD manufacturer uses, have a look around and you'll see that more often than not the most quoted one is Read/Write speeds.
 
Here in Holland there is a dutch site that did the same test

The bechmark after 3th run was 50% slower (the website is dutch but the Nr's speak for them self)

Intel X25-M 80GB (run 1) Single Intel ICH7R 1.134,3
Intel X25-M 80GB (run 2) Single Intel ICH7R .. 691,3
Intel X25-M 80GB (run 3) Single Intel ICH7R .. 592,1

So there is something happening
http://tweakers.net/benchdb/test/288 Boot and application performance index
(server benchmarks results added together and each given a X% of importance)

And here are all the benchmarks he use separately

http://tweakers.net/benchdb/suite/29 Server StorageMark 2004
http://tweakers.net/benchdb/suite/33 Gaming StorageMark 2006 Traces
http://tweakers.net/benchdb/suite/47 StorageMark 2006 Indices

http://tweakers.net/benchdb/suite/13 Storage benchmarks (collection of all his bechmarks also the old ones)

ps. use the filters because in some of the benchmarks he used 800 different benc test's including verrius raid setups.
 
I think the only benchmark result that could be meaningful would be a random read or write test before and after a week of usage. Something like testing application / game load time or windows boot time before and after one week of "normal" usage could work too.
 
Steve said:
It would seem that the crew at PC Perspective have grabbed Intel's attention with their article on SSDs. Ryan and the PCPer crew found that the performance of a used SSD would always be slower than that of a new, non-fragmented drive. Intel replied with the following:
Click to expand...

So a Windows install is not indicative of real world usage? Just that one thing caused degration of performance in their article.

I would say that is perfectly indicative of a real world usage scenario. Maybe not happening every day, but you have to install the OS to use it in the first place.
 
Steve said:
It would seem that the crew at PC Perspective have grabbed Intel's attention with their article on SSDs. Ryan and the PCPer crew found that the performance of a used SSD would always be slower than that of a new, non-fragmented drive. Intel replied with the following:
Click to expand...

Weird, color went black on me there. DOH! Here it is again, hopefully in a readable color this time!

So a Windows install is not indicative of real world usage? Just that one thing caused degration of performance in their article.

I would say that is perfectly indicative of a real world usage scenario. Maybe not happening every day, but you have to install the OS to use it in the first place.
 
What's with the black text on a black background? The default is set to white for a reason.

SSD's aren't even on the radar for me yet anyways and with only one page of replies so far and already one user who can attest to degraded performance, I'll be watching from the sidelines for a while yet.
 
Azhar said:
This pretty much echos what Kyle has been telling us. Synthetic benchmark means jack.
Click to expand...

Some might be less then others but how else are you going to be able to do it? What. have a stop watch loading up windows? Kyle's stand at least from what I have been able to understand is he would rather use real games to bench which is absolutely great until you see what resolutions and settings he uses, then that "real world" argument goes right out the window. Still makes me laugh [H] actually thinks the absolute highest known to man resolutions with the absolute highest AF and fsaa settings is somehow real world...real word to whom exactly? But that is just my opinion which means jack so....

Sparky!
 
maleficarus said:
Some might be less then others but how else are you going to be able to do it? What. have a stop watch loading up windows? Kyle's stand at least from what I have been able to understand is he would rather use real games to bench which is absolutely great until you see what resolutions and settings he uses, then that "real world" argument goes right out the window. Still makes me laugh [H] actually thinks the absolute highest known to man resolutions with the absolute highest AF and fsaa settings is somehow real world...real word to whom exactly? But that is just my opinion which means jack so....

Sparky!
Click to expand...

Well can you tell us what you think real world is, because I can't recall the last time someone spent $1000 on a quad sli config to play at 800x600 on a 17" monitor.

As for what Intel's saying, I don't buy it. I think they're covering their ass until they can dig deeper. Synthetic benches may not prove dick all, but data is data and this is a raw performance issue that I would like to see addressed before I even consider going SSD. Till they do, I'll deal with platters. The cost per gigabyte ratio is more favorable anyway.
 
snaggletooth said:
What's with the black text on a black background? The default is set to white for a reason.

SSD's aren't even on the radar for me yet anyways and with only one page of replies so far and already one user who can attest to degraded performance, I'll be watching from the sidelines for a while yet.
Click to expand...

Heh, I have no idea what the hell happened there with the black text, I never changed the color setting at all. I even reposted it, but I did copy and paste it, so the color came through again all weird.

Anyway, same here. I just got one of the new VelociRaptor 300GB drive about a month ago, so that will hold me for quite a while.
 
Miscommunication said:
Interesting if true...
Click to expand...
My X25-M after about 10 days of normal usage. Including one installation of Windows 7, followed by an installation of Vista. I even moved my temp files, browser cache, and swap file onto a spindle drive:
ssd-bench1.jpg


The same drive in its "factory state" (having been reset to that point by HDDErase 1.3 as Intel recommends):

ssd-bench2.jpg


I'm not really interested in arguing things. Those are just the numbers. Though it seems silly on the part of some to bring up [H]'s stance on synthetic benchmarks for video cards in this situation. That's a totally different scenario and the concerns that prompt [H]'s policy on video card evaluations really have very little to do with this current situation.

Take it for what you will. But also ask yourself why people tend to be defaulting to one side or the other with very little information. I just know what I saw myself. And what I continue to see.

H
 
maleficarus said:
Some might be less then others but how else are you going to be able to do it? What. have a stop watch loading up windows? Kyle's stand at least from what I have been able to understand is he would rather use real games to bench which is absolutely great until you see what resolutions and settings he uses, then that "real world" argument goes right out the window. Still makes me laugh [H] actually thinks the absolute highest known to man resolutions with the absolute highest AF and fsaa settings is somehow real world...real word to whom exactly? But that is just my opinion which means jack so....

Sparky!
Click to expand...

So what do you shoot for when playing games? Lowest resolution and no AF that gives you playable frame rates?? :confused:

Obviously if a video card will game at 2560x1600 with 2xaa and 16xAF, it will be good for anything below that. Do we really need to show you lesser settings to see if they work?

What is laughable about that exactly? And you might head over to our Displays forum and see just how many folks here game at 1920 and 250 resolutions......lots.
 
I wouldnt even waste my breathe on him, Kyle. He either gets it or he doesnt. His logic is so flawed its futile to even bother with arguing with him. If he cant see the gaping huge holes in his argument for himself then no matter of arguing or reasoning with him is going to change that. I mean there are some pretty big holes in argument.

That said, this is yet again one more reason not to buy ssd right now. I've heard about being on the bleeding edge of technology but god damn, that edge is sharp enough to cut diamonds. Its one issue after another with these ssd drives and then when you start to look at how much they cost you have to ask yourself, "Just exactly what are these coocoo's thinking pricing their drives at assine gb/$, when hdd's cost a helluva lot less, dont have nearly the crippling level of bugs, and are proven reliable."

Intel needs to get off its high horse with the whole $3/gig+ when their E series has been proven to brick on enterprise level tasks and now its M series is shown to be about as bankable as a one minute man. Computer use's typically are marathons not sprints and in a marathon I'd much rather have a tortoise than a hare.

I was hoping that ssd tech would be good to go for 2nd gen, however it looks like it wont be worth it until 3rd gen.
 
Azhar said:
....Synthetic benchmark means jack.
Click to expand...

Which is why I always skip past synthetic tests when shopping for new hardware, especially video cards. I always lol at people who say a certain card is really good or bad because how it does in a synthetic test. n00bs.
 
hofan41 said:
has anybody confirmed whether this occurs with the X-25E drive?
Click to expand...

We briefly had an E series drive to test, and it does fragment (by about the same ratio seen on the M drive), but it immediately snaps back, just as the M series drive does most of the time. We only had the drive for a week so we were not able to find out if it hits the 'point of no return' we saw with the X25-M.

Keep in mind this fragmentation is really just a down side to write combining. It is a simple tradeoff. The X25 drives achieve crazy high small write IOPS using this technique, in fact the M even beats out other SLC drives in this very area. If you do nothing but hit it with small writes, it is eventually going to slow to the point where it acts more like a non-write-combining drive, and larger writes will take a hit because of the fragmentation that came about from the initial write combining. Hitting it with a mix puts the average write speed at some point between 40 and 80 MB/sec, all depending on the particular mix of large and small files written (i.e. your usage pattern).

Even after writing the article, I still use my X25-M as my primary OS drive, even with all of the other SSD's at my disposal (including SLC units). Its insanely high IOPS performance makes it the best choice. You just have to realize that hitting it with a constant power user workload will slow it down. It is *supposed* to bounce back with larger writes, and we are working with Intel so they can easily replicate the 'bad' cases we saw. I suspect that once the Intel techs can dissect what we stumbled on, that a simple firmware fix will solve that particular issue.

A side note: Some sites have taken the Intel statement and spun it as if the drive does not slow down at all. This is not accurate, and I believe does everyone a disservice. Intel and PCPer have a mutual understanding that usage will drop performance to a steady state value based on that usage. You have only to read the "Ask and Intel SSD Engineer" article on [H] Enthusiast to figure that much out... It is also worth mentioning that the official write speed spec of the X25-M is 70 MB/sec, where a 'new' drive will write at 80/sec (closer to 90/sec with AHCI enabled). Clearly Intel considered this slow down when they published the spec, and it was very honest of them to do so.

Hope this info helps...

Allyn Malventano
Storage Editor, PCPer.com
 
This is Not News(TM) and it is Not New(TM).

Yes, SSDs of certain design types do have slower write performance. This is because, duh, they have to search for free blocks to write to. The blocks are not preallocated or premapped. This is the same in Hitachi USP and USP-Vs, HP XP24k, XP20k, EVA, IBM SVC 2145, IBM DS4200, DS4700, DS4800, DS5100, DS5300, the list goes on. (And the list starts at about $100K.)
If blocks are not preallocated, or during the process of allocating additional blocks for a new LUN in an array with free blocks, write performance is impacted because the controller has to locate free blocks. SSDs which do not preallocate all blocks will have the same issue. When there are more free blocks, there is less seek - when there are less free blocks, there is more seek.
 
I think the question here is, if synthetic benchmarks are showing a degrade in performance, and what looks to be a bretty good bit for a short amount of use, what is the "real world" degrade proportion? There must be a degrade somewhere. Sure a synthetic load is more then the average use, but if synthetic benches are showing a degrade in performance, is it to say there is none for "real world" use?

The one thing for sure is, Intel is not about to say that there shiny new and expensive solid state drives are degrading in performance. Anyone out there with a Solidstate drive, Intels uberfast 10 channel jobs or not...what is your take on the daily use of the drives you have? Do they feel slower now? Boot slower? load programs slower? Has the shine worn off and rust is forming? It would be cool to start a thread just for this "Performanc eof my Solid State/Intel Solid State drive over time"
 
Aside from any benchmarks, the thing that started me down this path to begin with was a DeadSpace install that took over 15 minutes, where the X25 had solid access and my Plextor appeared to be just sipping off of the disc. The very same install was repeated to a 74GB raptor and was much faster. The same install to a 'clear' X25-M was just as fast.
 
Allynm said:
We briefly had an E series drive to test, and it does fragment (by about the same ratio seen on the M drive), but it immediately snaps back, just as the M series drive does most of the time. We only had the drive for a week so we were not able to find out if it hits the 'point of no return' we saw with the X25-M.
Click to expand...
Awesome, thanks for the informative post!
 
Really good information down here. I hope this is a fixable thing, ie. within the firmware, and not something all the SSD drives will be prone to. It would suck if a drive would become significantly slower by using it a couple of weeks.
 
You must log in or register to reply here.
Back
Top