![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Intel G2 SSD still king?
- Thread starter Vexerz
- Start date
I'd still say its the # 1 choice.
The only thing that is really all around faster is the OCZ Vertex LE (Sandforce controller), however, it has an unproven track record.
The Crucial C300 is pretty fast, but it too has an unproven long term track record. Read the article at Anandtech about his dying. He had one for a long time and killed it. Granted he probably abused it, but that just simulates an extra few months or year of use maybe. So in a year, they may have a bunch of them dying prematurely.
Intel's drives have been out for a long time, especially when you consider the G2 is practically the same as the G1.
So if I was going to buy an SSD I would get an Intel SSD. (which I've done twice already for my home PC & work PC)
The only thing that is really all around faster is the OCZ Vertex LE (Sandforce controller), however, it has an unproven track record.
The Crucial C300 is pretty fast, but it too has an unproven long term track record. Read the article at Anandtech about his dying. He had one for a long time and killed it. Granted he probably abused it, but that just simulates an extra few months or year of use maybe. So in a year, they may have a bunch of them dying prematurely.
Intel's drives have been out for a long time, especially when you consider the G2 is practically the same as the G1.
So if I was going to buy an SSD I would get an Intel SSD. (which I've done twice already for my home PC & work PC)
While other SSDs may be faster in sequential benchmarks, the Intel generally does better in random I/O. That's really the most important for your system disk; having high sequential write performance is of little use for a drive not used for mass storage. Perhaps if you have a very large SSD and store other things on it too.
If you want an SSD to be the boot disk and contain your operating system, i'd say Intel is still king. Do not use Windows XP + SSD, and do not partition the SSD to use all available space.
If you want an SSD to be the boot disk and contain your operating system, i'd say Intel is still king. Do not use Windows XP + SSD, and do not partition the SSD to use all available space.
Overclock3d has a quick rundown on pricing in Europe for the Vertex 2 and the Agility 2
NCIX has the OCZ Vertex 2 50GB priced out.
They say available in April. Given the stated performance and the price seen thus far, I would think OCZ would be a solid choice.
The number of improvements and new products ocz puts out has really impressed me. I've went with ocz for my SSDs just to show my support.
NCIX has the OCZ Vertex 2 50GB priced out.
They say available in April. Given the stated performance and the price seen thus far, I would think OCZ would be a solid choice.
The number of improvements and new products ocz puts out has really impressed me. I've went with ocz for my SSDs just to show my support.
I use both Intel (G1 and G2) as well as Indilinx drives and a few others. I had a Vertex LE 100GB on pre-order for a few weeks before launch. Despite Anand's benchmarks on it, I really just couldn't tell the real-world usage difference to the point where it made sense at over $4/GB. I returned the Vertex LE and I bought a second $300 120GB Vertex and just did Raid 0 (I needed more SSD space so not getting another SSD was out of the question): http://www.servethehome.com/ocz-vertex-120gb-rma-raid-0-time/
Mind you this is just after the v1.5 firmware flash killed the first Vertex so I actually bought the second while the first was in RMA land.
Oddly enough, I can't tell much of a difference even with the second drive in there. I don't think anyone has pure 4K write usage patterns, so despite Intel's lead in that area, real world I've not seen a big difference between the two. Now, Indilinx or Intel drive v. a 7200rpm drive I could tell the difference easily.
Mind you this is just after the v1.5 firmware flash killed the first Vertex so I actually bought the second while the first was in RMA land.
Oddly enough, I can't tell much of a difference even with the second drive in there. I don't think anyone has pure 4K write usage patterns, so despite Intel's lead in that area, real world I've not seen a big difference between the two. Now, Indilinx or Intel drive v. a 7200rpm drive I could tell the difference easily.
Actually the Vertex LE beats the Intel drive in random performance, but again, it's a huge premium to pay for the sandforce controller, and it's likely that it won't make any noticeable difference in day to day usage.
I'm skeptical of some of the benchmarks I've seen with Sandforce drives. The details are sketchy, but it seems that Sandforce does some type of simple compression before writing data. I guess it is something like encoding runs of repeated bytes. But the point is that any benchmark that writes a simple pattern may benefit greatly from that compression, but real-world usage, probably not so much.
And why not? Thats what the Intel drives were designed for. They work perfectly well aligned or unaligned!!! And you can schedule the Intel SSD Toolbox to trim it weekly, so it stays at peak performance.
I love my Intel X25-M 80Gb G2 on my work PC with Windows XP. I get similar benchmarks on it as I do with my Win7 install on my home PC. And before I installed Win7, I had WinXP installed on my home PC for a few months and the drive was not aligned and it ran great that way too.
Intel remaps writes so its not a big deal for WinXP and/or an unaligned partition.
YES! Why Intel SSD is sometimes preferred
For all the generic consumer SSD questions.
1, If you wanted flying colors, buy any SSD drives that advertize Highest READ/WRITE performance. Currently (year 2010) for generic consumer usage pattern, those drives and Intel SSD are generally OK and buying decision is upto your own preference.
2. If you wanted flying colors and have premium funds available. Considering you are willing to pay premium, you should get premium quality and performance, get Intel SLC-SSD or modern-day SLC from other brands, read whole post.
3. If you have lower fund, wanted "ASSURANCE" and willing to give up slightly on other area, get Intel MLC-SSD. I am neutral so if other brands can give info, they should receive the same recommendation as well.
3a. Intel SSD originally was designed not only to be sold to consumer, some of them were meant for Enterprise OEM or customers. I said original because one or two did decide to change brand. fact is fact..
3b. Enterprise OEM --> people that deploy SSD in their storage enclosure and sell to other customers..usually enterprise.
3c. Such customer groups are willing to pay premium and in turn expect answer and preferably not trying to learn reliability through postings.
3d. Therefore the SSD design has engineering limits that must be taken in consideration.
3e. An academic example
3e1 - Intel sells MLC-SSD 3 years warranty
3e2 - Intel estimates how many cycle of writes particular model MLC-SSD can handle
3e3 - To keep SSD no trouble for 3 years, Intel knows exactly the limits of the product and tells you exactly how much it expects to honor the write
3e4 - (Rumor) to make sure it can survive warranty, the firmware fine-tune itself in aggressive situation to ensure it can survive warranty (control write). (Rumor, I have no real proof...)
3e5 - Intel tells you in advance so you do not need to guess.
(technically the Intel MLC-SSD can last longer (potential total write GB). However, based on the current OS sw/hw behavior, 3 years are the limit they are willing to cover)
4. Western Digital also tells you in advance for consumer SSD. I am not sure but I think it is about 17+ GB/day. Intel-MLC-SSD is 20GB/day. SLC-SSD is usually very much higher.
5. Forum members can you please help all of us by listing consumer SSD estimated expected Write GB/day.
For all the generic consumer SSD questions.
1, If you wanted flying colors, buy any SSD drives that advertize Highest READ/WRITE performance. Currently (year 2010) for generic consumer usage pattern, those drives and Intel SSD are generally OK and buying decision is upto your own preference.
2. If you wanted flying colors and have premium funds available. Considering you are willing to pay premium, you should get premium quality and performance, get Intel SLC-SSD or modern-day SLC from other brands, read whole post.
3. If you have lower fund, wanted "ASSURANCE" and willing to give up slightly on other area, get Intel MLC-SSD. I am neutral so if other brands can give info, they should receive the same recommendation as well.
3a. Intel SSD originally was designed not only to be sold to consumer, some of them were meant for Enterprise OEM or customers. I said original because one or two did decide to change brand. fact is fact..
3b. Enterprise OEM --> people that deploy SSD in their storage enclosure and sell to other customers..usually enterprise.
3c. Such customer groups are willing to pay premium and in turn expect answer and preferably not trying to learn reliability through postings.
3d. Therefore the SSD design has engineering limits that must be taken in consideration.
3e. An academic example
3e1 - Intel sells MLC-SSD 3 years warranty
3e2 - Intel estimates how many cycle of writes particular model MLC-SSD can handle
3e3 - To keep SSD no trouble for 3 years, Intel knows exactly the limits of the product and tells you exactly how much it expects to honor the write
3e4 - (Rumor) to make sure it can survive warranty, the firmware fine-tune itself in aggressive situation to ensure it can survive warranty (control write). (Rumor, I have no real proof...)
3e5 - Intel tells you in advance so you do not need to guess.
(technically the Intel MLC-SSD can last longer (potential total write GB). However, based on the current OS sw/hw behavior, 3 years are the limit they are willing to cover)
4. Western Digital also tells you in advance for consumer SSD. I am not sure but I think it is about 17+ GB/day. Intel-MLC-SSD is 20GB/day. SLC-SSD is usually very much higher.
5. Forum members can you please help all of us by listing consumer SSD estimated expected Write GB/day.
Last edited:
Do you have any source on this? I do not believe it is theoretically possible to correct unalignment, other than removing the sector size limitation. Intel SSDs use 128KiB flash blocks; so thats their internal sector size. Any writes not aligned to this would cause misalignment.
Anything not aligned would affect two flash blocks, meaning your writes will double the wear on the drive, half the lifetime of your product, and impact on performance as well.
If all you do is sequential benchmarks then a misalignment won't be visible; but it'll be there. To work with XP without misalignment, either correct the partition yourself manually or use a Vista/Win7 setup dvd to create the NTFS partition; then install XP on that partition without formatting.
Very interesting, thanks for the link.
Though only Intel seems to be relatively unaffected by this, as with other SSDs the performance gains of using aligned partition are huge:
OCZ Vertex LE (Sandforce-1500) goes from 51MB/s to 162MB/s, more than 3x the performance.
Crucial RealSSD C300 goes from 18 to 141MB/s, more than 7x the performance
The only SSD that seems relatively unaffected, is the Intel X25-M, that improves its performance from 37.4MB/s to 46MB/s - still a decent 23% faster.
I would like to confirm this on my own though, and particularly i'm interested in how the SSD would detect unaligned writes; for all it knows this is one big file in which specific parts (LBAs) are updates. It is possible to fix alignment when mapping LBA to flash cell blocks, but then it doesn't explain the 23% speed difference if that was done correctly. With such a remapping feature i would expect close to zero different in performance.
Either way, you do not want to use Windows XP + SSD; as all modern SSDs appear to lose performance due to the unaligned partition; just like with WD EARS harddrives. If you want to use Windows XP with such a drive, using Vista or Wïndows 7 to create the partition first would be a sensible thing to do.
Though only Intel seems to be relatively unaffected by this, as with other SSDs the performance gains of using aligned partition are huge:
OCZ Vertex LE (Sandforce-1500) goes from 51MB/s to 162MB/s, more than 3x the performance.
Crucial RealSSD C300 goes from 18 to 141MB/s, more than 7x the performance
The only SSD that seems relatively unaffected, is the Intel X25-M, that improves its performance from 37.4MB/s to 46MB/s - still a decent 23% faster.
I would like to confirm this on my own though, and particularly i'm interested in how the SSD would detect unaligned writes; for all it knows this is one big file in which specific parts (LBAs) are updates. It is possible to fix alignment when mapping LBA to flash cell blocks, but then it doesn't explain the 23% speed difference if that was done correctly. With such a remapping feature i would expect close to zero different in performance.
Either way, you do not want to use Windows XP + SSD; as all modern SSDs appear to lose performance due to the unaligned partition; just like with WD EARS harddrives. If you want to use Windows XP with such a drive, using Vista or Wïndows 7 to create the partition first would be a sensible thing to do.
I think that was exactly the point that C-rizzle was making. The X25-M was designed to work well without 4K alignment, but other SSDs are not.
I do not think the X25-M "detects" unaligned writes. I think it always uses an algorithm that does not slow down very much due to unaligned writes, but has the downside that the algorithm overhead prevents the high speed writes (that are possible for other SSDs) when they are 4K aligned.
Anandtech said:
I've personally run an Intel X25-M 80Gb for a few months on a WinXP install, it was unaligned until I upgraded to Win7 just recently. And from all the benchmarks, etc... it preformed almost identically across the board.
Then just recently I upgraded my HDD is my work PC to an Intel X25-M 80Gb also, but this time I aligned the partition properly (know more about SSD's). And benchmarked it and used it the past few weeks. And I can say its pretty much the same performance wise across the board.
I've tested the drives like 6 different ways:
WinXP unaligned with IDE, AHCI and Intel Raid drivers
WinXP aligned with AHCI
Win 7 aligned + AHCI/Raid
And the benchmarks ranged +- 8Mb/s, so the total range of scores were maybe 15Mb/s on the seq. read, and on nearly every other benchmark the range was much less... maybe 5Mb/s.
http://www.anandtech.com/show/3618/intel-x25v-in-raid0-faster-than-x25m-g2-for-250
Anandtech said:
Last edited:
What exactly have you tested? Random I/O or sequential I/O? Unaligned partitions will not affect sequential I/O; or at least should not.
If Intel indeed remaps unaligned writes, it also wouldn't wear the drive excessively; but other SSDs will. So for those SSDs it is not a question of performance; their lifespan would also be cut in half.
Perhaps you simply did not notice a 20% performance difference; which is not all that big. Still if you have a chance to make your partition aligned, the 20% extra random IOps is a nice bonus.
As more and more storage devices become affected by alignment issues, older operating systems like Windows XP will have a hard job performing well. The new EARS harddrives use 4KiB sectors and thus need proper alignment, any striping RAID array needs alignment, and Solid State Disks too; perhaps with the exception of the Intel controller.
One good method would be to increase the sector size to a large value, like 128KiB. That would make sure everything is aligned and the SSDs will never have to do a read-erase-program cycle. The downside is that small writes would take up to 128KiB; so you need a smart filesystem to do with such a large sector size.
For example, if you setup a RAID5 under nVidia onboard RAID controller, it will perform like 10MB/s write speed. But if you make the writes aligned, it will be like 400MB/s write speed; very good. So alignment is getting more important as more and more devices need it; both newer harddrives and SSDs.
Besides, XP is like 10 years old, it has various I/O limitations (such as MAX_PHYS of 128KiB, no native AHCI support, no NCQ for SSDs, no TRIM support). The reasons to run XP have diminished now that Windows 7 is released. I would hope the next wave of SSDs, particularly the new Intel controller in Q4 of this year, would not be hampered in performance too much, by making it compatible with older operating systems like XP.
If Intel indeed remaps unaligned writes, it also wouldn't wear the drive excessively; but other SSDs will. So for those SSDs it is not a question of performance; their lifespan would also be cut in half.
Perhaps you simply did not notice a 20% performance difference; which is not all that big. Still if you have a chance to make your partition aligned, the 20% extra random IOps is a nice bonus.
As more and more storage devices become affected by alignment issues, older operating systems like Windows XP will have a hard job performing well. The new EARS harddrives use 4KiB sectors and thus need proper alignment, any striping RAID array needs alignment, and Solid State Disks too; perhaps with the exception of the Intel controller.
One good method would be to increase the sector size to a large value, like 128KiB. That would make sure everything is aligned and the SSDs will never have to do a read-erase-program cycle. The downside is that small writes would take up to 128KiB; so you need a smart filesystem to do with such a large sector size.
For example, if you setup a RAID5 under nVidia onboard RAID controller, it will perform like 10MB/s write speed. But if you make the writes aligned, it will be like 400MB/s write speed; very good. So alignment is getting more important as more and more devices need it; both newer harddrives and SSDs.
Besides, XP is like 10 years old, it has various I/O limitations (such as MAX_PHYS of 128KiB, no native AHCI support, no NCQ for SSDs, no TRIM support). The reasons to run XP have diminished now that Windows 7 is released. I would hope the next wave of SSDs, particularly the new Intel controller in Q4 of this year, would not be hampered in performance too much, by making it compatible with older operating systems like XP.
Well from running Crystal Disk Mark on all of the installs across the board the scores were as follows:
---------- read ----- write ----
Seq.--248-260----85-90
512k--170-190----80-85
4k------22-23-------63-65
I have screenshots from nearly every configuration, but that table above summarizes the benchmarks. And in everyday use, I think the unaligned IDE in WinXP "felt" the fastest, but thats very subjective!
---------- read ----- write ----
Seq.--248-260----85-90
512k--170-190----80-85
4k------22-23-------63-65
I have screenshots from nearly every configuration, but that table above summarizes the benchmarks. And in everyday use, I think the unaligned IDE in WinXP "felt" the fastest, but thats very subjective!
I'm particularly interested in random reads, could you try the AS SSD benchmark? And also HDTune Pro (evaluation though; beware) has a nice Random Access test that measures IOps.
The 4K test seems to be on the low side. Intel is capable of 200MB/s+ random reads with enough queued I/O's. The AS SSD benchmark should have a Q64 random read benchmark as well. If this is still a rather low value, that means you need NCQ (AHCI) to go any faster - as that would enable multiple flash channels to work in parallel, when doing random I/O. With sequential I/O; the dirve predict contiguous reads and performs read-ahead so all channels work in parallel.
The 4K test seems to be on the low side. Intel is capable of 200MB/s+ random reads with enough queued I/O's. The AS SSD benchmark should have a Q64 random read benchmark as well. If this is still a rather low value, that means you need NCQ (AHCI) to go any faster - as that would enable multiple flash channels to work in parallel, when doing random I/O. With sequential I/O; the dirve predict contiguous reads and performs read-ahead so all channels work in parallel.
I ran those benchmarks on Crystaldiskmark 2.2.
version 3.0 has a 64-thread (I think) version of the 4K, which I'm sure would have been higher across the board, but especially on the AHCI, and not so much on the IDE as it doesn't support NCQ.
My 2 drives are now installed as follows:
Win7 with 9.6 chipset (raid drivers b/c I have 2 other drives in Raid on same controller) - native TRIM
WinXP aligned with 9.6 chipset drivers in AHCI mode
version 3.0 has a 64-thread (I think) version of the 4K, which I'm sure would have been higher across the board, but especially on the AHCI, and not so much on the IDE as it doesn't support NCQ.
My 2 drives are now installed as follows:
Win7 with 9.6 chipset (raid drivers b/c I have 2 other drives in Raid on same controller) - native TRIM
WinXP aligned with 9.6 chipset drivers in AHCI mode
