SSD Thread (Benchmarks, Brands, News, and Advice)

Well, I don't see why they would give you the option to change the default 75% to 100% or to decrease it down to 50% if such kind of cases couldn't occur.

As I mentioned before I never had to face such a situation because I never used hibernation. I consider it useless, if anything I use sleep mode.

 

me either, never has any problem before from that site until this happens, that site is ok.

 

The only reason I can foresee is that one may either:

1. Not have enought space, still want to use hibernation, so he'd be able to reduce it to 50% of his RAM;

2. Have plenty of space and wanna make sure he won't face any problem, so he could just extend it to 100% of his RAM...

Apart from that, I am with you, I can't see any other reason MS would allow the user to stretch or reduce it...

 

In that video there is no single world about wear leveling! EOD!

Maybe You are sure about Your information, and no, You don't have to respond to any of my posts, but Your information regarding wear-leveling is wrong!

Are You saying that wear-leveling and TRIM is same thing?



I do not like to argue with anyone but information that wear-leveling works only with non-partitioned space is wrong!

 

That is true. NAND cells used for wear leveling are in a separate, unpartitioned area that you cannot see in Windows. Only the drive's controller can see and use it. But it will also use any extra unpartitioned, never-written-to space that you leave for it. If you give me some time I will find that information for you. I am sure I saw it on Intel's website or maybe in the video.

Here is a starting point for you:
Intel SSD Support Community

There is a lot of helpful information there. You can ask about the issue in question there. Please let me know if I am wrong and post the source (hopefully an Intel employee).

Thank you.

RE: RST DRIVERS
A couple snips from the Intel SSD Forums:




Comments?

 

I still haven't seen Phils question answered. I have the same. I am assuming that just because hiberfil.sys shows for example 2GB of used space, it's kind of like the pagefile.sys file where it's just reserved space and not necessarily that size. It will only utilize that space when you go to hibernate, so it's not constantly writing to the hibernate file, if I understand this correctly.

(whew, an onslaught of posts today, can't keep up, have too many other things going on).

 

can someone post CrystalDiskMark Benchmarks of 128GB C300? I need them to compare to mine as i do SSD tweaks.

 

@Tomy B.
This LINK provides a great deal of info about the X25M drives. At the bottom of the page are the 4 IDF (2009) presentation videos. The 3rd deals with endurance. Between minutes 9 and 14 of that video he talks about the "Spare Area" (7% by default) and he says it is "user configurable" to increase that area on a new drive simply by creating a smaller partition. He also says that this is the ONLY way to increase that area (cells that have never been written) other than by secure-erasing the whole drive and repartitioning (not as effective). Increasing the Spare Area dramatically increases the endurance of the drive. From what he is saying, it is obvious that this has to be space outside of any partition and never written to before.

The Spare Area is for wear-leveling. When you run the Optimizer in the Intel Toolbox, it uses this area to move "erase blocks" (512k) to and mark the original blocks as available and then re-writes those areas of the blocks that contain files that are NOT being deleted (TRIM). This is wear-leveling which is invoked by the TRIM command either through the Optimizer routine or natively by Windows 7 if you are using the default AHCI driver or RST driver 9.6 or later.

Edit: The 2nd video "Understanding SSD Performance.... " Also talks about the "Spare Area" and how a larger area can enhance not only endurance but performance as well. It also explains the term "write amplification" which is why we need the TRIM command and a spare area for wear-leveling where all of this takes place. The concept is actually pretty simple... birds fly faster and longer in a larger space and more birds can use it... (to use what may be a poor analogy).

 

Of course that You can't see any content in unpartitioned space, but wear-leveling is moving around data that You need and it must be accessible after moving to another "place".

In that video he is talking about spare area, but doesn't say any word about wear-leveling.

Optimizer and TRIM doesn't have anything common with wear-leveling.

Optimizer and TRIM keep drive in fresh, like new, state, but wear-leveling "know" which cell have how much write cycles and try to keep all cells equally worn.

 

It is my understanding that the Spare Area is used for the Optimizer and TRIM and wear-leveling is simply a record of memory cell usage which is tracked by the controller so the same cells don't get used over and over again. But in practice, all of this actually takes place (the block moves) in the spare area which is used only for this procedure in order to clear blocks with deleted files in them so they can be freed up for new writes.

Your original objection was my claim that this spare area had to be unallocated space outside a partition that had never been used. I think you can see and hear from the videos that this is true.

 

This is my original objection.

Source

Source (page 5)

 

I have no problem with what you're saying but those are just generalized statements that don't go into the details of how it is accomplished, physically. The "Spare Area" size has a direct performance and endurance impact on the drive via the whole process of wear-leveling, TRIM, etc. That's where it all takes place.

Maybe we're just talking "apples and oranges" here and our concepts are not aligned. In that case there is no point in continuing the discussion. My understanding of the process works for me. I hope yours works for you.

Thank you for your time and responses.

 

Just one more question: do You know that the same data (say one song that you copied to SSD after installing OS and never open, copy, delete, rewrite of what ever) doesn't stay always in same physical cells?

 

I don't know, but I suspect that the wear-leveling process will eventually move it somewhere else to even out the usage of the memory cells. Or, if another file in that 512k erasure block is deleted, all the files in that block get re-written, most likely to another block or series of available blocks. I would bet that few people truly understand this except the guys who write the controller firmware.

 

No hard fillings, but You really doesn't know that and come here to argue with someone about wear-leveling?

 

I never came here to argue about wear-leveling or anything else. You contested my statement that the spare area had to be in unallocated space. You called it "nonsense". As you can see now from the information from Intel, it DOES have to be in unallocated space.

You are trying to goad me into an argument which serves no one and is not appropriate or productive. I have other threads to read on other forums. Good luck finding a combatant, I'm logging out.

 

I contested this statement:

Source

All I'm trying to do is to teach someone something that I'm 100% sure I know what I'm talking about!

 

You are correct in both cases.

1. Hiberfil.sys is only a reserved disk space for the content of the RAM to be written to in case of a very long winter (hibernation).

2. If the long winter comes, then it will write to the said space the content of the RAM, and as Stamatisx says, it will compress the compressible data in oder to write everything in RAM onto the hiberfil.sys file, thus, more likely not needing as much space in the hiberfil.sys file than it needs of RAM, because of the compression it will operate when writing the data to the disk.

So, yeah, I would say you answered your own question perfectly !

 

When should we expect to see 20nm class SSDs on the market? I'm trying to hold off on buying until these are out, and according to news reports on the web most manufacturers have changed over to this manufacturing process already (saw a report that confirmed Intel changed over to this process already in late July) but doesn't seem like any of them are on the market yet?

 

Not before Q1 2011 according to recent news.

 

And then they will probably be priced out of this world for a while until the older stock is exhausted. For some reason I don't see them lowering the current stock pricing even when the new stock comes out.

 

They're (Intel) claiming the 25nm Gen 3's will be double the capacity for the same price. Hahahahahaha... in your dreams

Edit: Their production costs, not OUR price.

 

Are you sure Intel said that? Do you have a link?

Intel may have said it halves production costs of memory chips.

 

The manufacturers are claiming that costs go down due to increased yield (both because of more chips per the same physical production size and also due to less failure rate of production). I do think you are partially right in that manufacturers will use this to increase profit margins. However, 20nm class SHOULD be win win - the higher capacity shouldn't be accompanied by exponential cost increases....at least one would hope.

 

You're probably right. But halving production costs won't translate into what my original post said. They were "hinting" or implying that you would get double the capacity for about the same price. No direct promises, statements or specific pricing info was offered. This came from one of the IDF (2009) videos.

Other post amended.

 

Well yes that and the fact that the costs of an SSD are made up of other components than just the costs of the memory chips.

 

Maybe someone can clarify a point about NAND memory cell longevity. One of the Intel videos said that the cells can hold a charge for "about" 10 years.

Someone here said that only applies to writing data and not reading. Does this mean that what has already been written to the cells will be accessible for longer than "about" 10 years? If so, I wonder how long the data can be accessed with confidence (no loss of data integrity).

Anyone seen any studies on this?

 

I don't think they know for certain since the technology hasn't been around that long. Flash has been around a while, but not with the technologies they use now.

 

IRST 10.0.1043, tested and works perfect in PM55+i7:
intel drivers pour Raid/Sata/Ata/Ahci
[ Non official drivers, use at your own risk. Phil ]

 

Works fine in my HM55+i5 laptop. Seems to boot up faster.

I'd post CDM scores but I've already run that benchmark numerous times and want to start cutting back on unnecessary writes.

Speaking of that, I just got a 16GB Class 10 SDHC card [Sandisk Extreme III] and now direct a lot of writes (downloads, temp files, etc.) to that card to save wear & tear on the SSD. Seems to be working well, so far.

 

That's exactly what I was thinking of doing. Does the SD card sit flush with the chassis, though?

 

Not quite. Sticks out maybe a millimeter or two. Not enough to worry about -- I can just leave it in there since I carry the laptop in my backpack with the other, opposite side facing downward.

 

anyone else try them? Do they improve SSD performance.

 

sean473,

I have been running on those new beta drivers for a couple of days now and definitely know that the system is snappier.

I was curious to see if CDM would show this: nope - it actually shows slightly less 'scores' than the official 9.6 version did. But... there is no doubt that the system feels faster....

So, I tried AS SSD too - again the 'scores' are almost identical - but, what has improved considerably is the access times AS SSD shows.

0.221 ms read access time and 0.270 ms write access time is definitely an improvement from 0.288 (r) and 0.360 (w).

Can hardly wait for the WHQL' certified drivers!

 

i now can confirm this version works great with HM55+i7, HM57+i5, GM45, GM965 chipset laptops here.

 

How about the temps?

Btw, this tweak will greatly improve your bench scores including access times (Read = below 0.02 ms)

 

Had to find out.. so I ran the CDM again:

With IRST 9.6:


With IRST 10:


Some scores declined while the all-important 4k scores went UP.
I have the registry tweaks applied that are referenced in the post above mine by lousygreatwallgm.

I realize different test parameters were used in the two tests... (5/50MB) -- I don't think that would make a great deal of difference, would it? Both tests in regular boot (not in safe mode).

 

I have a question, I read on OCZ's website in reguards to the Vertex 2 on a Intel Chipset system, it's better to use Microsoft's drivers than to use Intel's when useing the drive under ACHI nonraid (my setup). Anyone know if this is true? I usaly do not like Microsoft's drivers.

From their site:
Which SATA/Chipset drivers are best to use with Vertex?

1 Intel chipset IDE mode = Intel driver

2 Intel chipset AHCI mode = Microsoft driver

3 Intel chipset raid mode = Intel driver

4 AMD chipset IDE mode = Microsoft driver

5 AMD chipset AHCI mode = Microsoft driver

6 AMD chipset raid mode = AMD driver

7 Nvidia chipset IDE mode = Microsoft driver

8 Nvidia chipset AHCI mode = Microsoft driver

9 Nvidia chipset raid mode = Nvidia driver

 

I may reverse the tweaks from "lousygreatwallgm" because now my CPU will not go above x11 multiplier (1463mhz). I do miss the 1.73GHz. So, it's a trade-off, better SSD performance versus top CPU performance. Must be a way to have both. Someone say yes. Please?

 

I suspect you tweaked the wrong one (Maximum instead of Minimum).

Can you check from your selected power plan Maximum Processor State is at 100%?

 

Can't answer directly on your question regarding IDE since I never tried it.
But setting it to IDE will only give you slower performance vs. AHCI.

Once used NVidia driver (DELL) on my XPS = performance dropped like hell
Can't recall seeing NVidia driver support TRIM feature.

Intel chipset... this is the interesting part
MS drive controller driver vs Intel RST driver will give you close to none performance difference in real word usage til you start doing some bench tests (Intel wins)

PS: Only use Ver:9.6.0.1014 or later to have TRIM work.

Have fun on your amazing SSD!
"My SSD Is Faster Than Your SSD" Still trying to decide where to stick mine

EDIT: Did you buy the Vertex 2 yet?

 

My boot became slightly slower. 4K read and writes are identical.

 

For AHCI and Windows 7 setup, I've found that the default MS controller beats out all of Intel's drivers.

This particular driver introduced TRIM support for RAID setups. It doesn't apply to AHCI. MS default controller supports TRIM as well.

Mine became slower too. I have yet to find an Intel driver that performs better than the default MS driver.

 

Ive had the exact same SD card in my laptop for months now. All my documents, pictures, and temp files go there. The media files remain on my SSD (Intel G2 80gb). My SD card also sticks out a millimeter or two. No problems, though I've accidentally ejected it only once, but fortunately it was while the lid was closed.

 

How does the speed of that SD card compare to a ssd speed?

 

The default MS driver is fastest with booting because the Intel RST driver takes some time to load.

In other tasks the Intel RST driver should be faster though. Have you measured it?

 

I had done most of my measuring a while ago using HDTune Pro. Though it's not the best tool for measuring, it gave me a good baseline for 64k reads. MS driver consistently performed better. It was better than IRST 10 by about 10MB/s. I still have to use HD Tune to check the drivers, cuz it was my original basis for comparison. Way too much work to re-bench using CDM or AS SSD. Not worth the writes either. Would be nice if either of those tools could have an option to disable writes. They're not necessary EVERY time the programs are run.

 

I did some more real world testing: duplicating files, scannning folders, multi tasking. No differences between IRST 9, 10 or MS driver.

The only thing that did give a small difference was booting, on average MS driver boots 0.2 second faster.

 

My real world testing is simply using the computer in an entirely 'normal' (for me) way for a few days and comparing to the previous version. The v10 IRST drivers deliver much more 'snap' in my system (i3 350M, 8GB RAM) with an 100GB Inferno SSD.

I would not expect it to increase performance per se, (actual 'work' is not done by a storage device - 'work' is only done by a CPU in RAM) - but the response time is certainly better which give a more connected feeling to the O/S for me.

But then again, I'm very sensitive to how much 'snap' any system has too.

 

Maximum is set at 100%, as listed in your post. I double-checked all of them. This may be a Turbo Boost issue. Changing certain settings seems to disable or limit Turbo Boost. After registry tweaks were reversed, CPU back to 1.73GHz when under load but CDM scores went back down. Can't seem to have the best of both worlds. I'll keep the higher CPU clockspeed, for now. Real-world usage, for me, doesn't really give a noticeable difference in SSD performance. I prefer having the CPU speed because I do a lot of archiving, compiling, and encrypting.
EDIT: I didn't actually change back the registry settings, I just put the High Performance Power Plan back to "default settings" and my CPU was immediately able to go to 1.73GHz (full Turbo Boost).
Note: I just realized that my laptop has a service running called "Acer ePowerManagement" which may be causing the throttling with your settings applied. I'll disable that service and try your settings again.

Not even close. But it seems to be the fastest SD card available and since SanDisk & Lexar are the only manufacturers of these cards, the SanDisk comes highly recommended. Mine is rated at a max speed of 30MB/sec but it takes a few seconds to get up to that speed, it's not instantaneous like the SSD:
Sandisk Extreme

Memory card readers don't have the speed advantage of a SATA II controller like the SSD does. I think they're on the USB hub, or maybe PCI-e controller/bus... don't really know.