X25-M G3 found at Taobao

Not too bad when you think about it in $/GB. The 160GB G3 is $2.244/GB and the 600GB G3 is $1.995/GB with the current 160GB G2 drives still selling for $400+ new ($2.50+/GB). I do hope they end up being cheaper, though.
 
600GB at 800 euro would be close to 1.0 eur per GB. But don't count on cheap prices when this thing launches; you would have to wait a month or two for prices to stabilize.

Like the G2, it was cheaper than the G1 and faster (and had TRIM); so everyone wanted them; which kept its price very close to the G1 shortly after it was released. I think the G3 will be comparable; i think the prices should be very decent a couple months after launch.
 
Prices in different regions are rarely offered at the same (but converted) point, and besides, prices will come down.
 
So its clocking 220MB/s reads and 170MB/s writes - am I missing something? granted its pre-release hardware and the unit they got might've been a factory reject not fit for sale but even at the spec'd 250MB/s reads and 170MB/s writes -- not exactly vaulting me over the fascination threshold, especially with SF-2000 and Crucial/Micron gen2 on the horizon in early 2011.
 
I'm not trusting any of that info, I'll wait for production units to be reviewed.
 
odditory said:
So its clocking 220MB/s reads and 170MB/s writes - am I missing something? granted its pre-release hardware and the unit they got might've been a factory reject not fit for sale but even at the spec'd 250MB/s reads and 170MB/s writes -- not exactly vaulting me over the fascination threshold, especially with SF-2000 and Crucial/Micron gen2 on the horizon in early 2011.
Click to expand...

True, but Intel has never been competing on SSD performance, they are competing on reliability. Would still hope for better gains though.
 
understood. to me the idea of "reliability" is only as significant as the warranty period. if two drives from two different makers have the same 3-year warranty but one of them performs 33% better than the other, guess which one i'm going with.
 
Intel's writes are very fair; it can do 170MB/s of random writes; so whether you do sequential write or 4K random write; it will go at about the same throughput; which is pretty amazing and means a solid controller design.

Still, i want SSDs to give really high sequential throughput since there is little reason they can't reach 1GB/s easily, if they were not limited to SATA design and the company actually tried to optimize for performance instead. OCZ RevoDrive is a step in the good direction; but still is a Silicon Image PCI-X controller; not very sleek. Where are those hot SSDs which are faster than light and great for high-performance applications?

To match the performance level i'm setting myself on, i would need about 4 of those newer Intel's to really help my ZFS performance as SLOG; the sequential write of the SSDs combined must be higher than the sequential write the physical disks are capable of. Thus i would need > 800MB/s write on the SSDs; at least 4 brand new Intel SSDs. Whereas i might already be reaching this level with 2 new SATA 6Gbps SSDs which write at 400MB/s instead; what i expect from a real 6Gbps SATA SSD.

So while performance of SSDs is high, it can be higher, the steps we're making are quite small compared to what is theoretically possible. Though to be fair, the performance is growing alot faster than HDDs are capable of; whose performance would only scale with data density. So performance of HDDs only rises very slowly, while SSDs should have much more significant steps in performance.
 
odditory said:
understood. to me the idea of "reliability" is only as significant as the warranty period. if two drives from two different makers have the same 3-year warranty but one of them performs 33% better than the other, guess which one i'm going with.
Click to expand...

I guess that depends how much you value your data. Or how robust your backup solution is.
 
Old Hippie said:
Looks about right for real world specs to me.
Click to expand...

The read looks low. G2s get around 250 MB/s or even a little more. I guess there was either something wrong with the benchmark run, or else the prototype unit does not run at full speed.
 
The following is an opinion, not a fact.

Intel has many considerations. Observing the overall markets

1. For the people "currently one to three years" seeking FASTEST performance, you should aim at those PCI-Express SSDs, or some other vendors' products aiming for maximum write speed.

2. For the purpose of transitioning the entire storage market particularly targeted segments and light duty servers to SSD, the focus is refinement to address all round issues and bring the cost/gigabyte slightly more acceptable. mind you, slightly "more acceptable", not "easily affordable" :) It is understandable as SSD still carry substantial performance improvement for many usage scenarios over current mechanical so the premium is there.
 
Forceman said:
I guess that depends how much you value your data. Or how robust your backup solution is.
Click to expand...

what does it have to do with the value of the data - you should be keeping a current disk image of your drive with Acronis/EASEUS/WHS agent/Windows 7/whatever since its easy, and people should be doing that regardless of what they believe about their particular drive's "reliability", good or bad. people seem to be of the belief "intel ssd's are the most reliable" but that shouldn't excuse anyone from keeping an up to date image, regardless of the validity of the statement.

"reliability" is kind of a silly metric which can't really be measured by a few people chiming in on a forum to talk about not having any issues with their particular drive.

reliability = warranty period + backups.
 
Last edited:
odditory said:
what does it have to do with the value of the data - you should be keeping a current disk image of your drive with Acronis/EASEUS/WHS agent/Windows 7/whatever since its easy, and people should be doing that regardless of what they believe about their particular drive's "reliability", good or bad. people seem to be of the belief "intel ssd's are the most reliable" but that shouldn't excuse anyone from keeping an up to date image, regardless of the validity of the statement.

"reliability" is kind of a silly metric which can't really be measured by a few people chiming in on a forum to talk about not having any issues with their particular drive.

reliability = warranty period + backups.
Click to expand...

What people should do, and what people actually do are two different things. Yes, everyone should be keeping nightly backups, etc, etc, but not everyone does - so reliability does matter.
 
Heres the deal guys, lets talk steady state performance. Steady state performance of these drives taht are supposedly faster than current intels (C300 and SandForce, im looking at you) is LOWER than the specs of the current intels. Once you run these drives awhile, they suck. their performance goes to shit after a few weeks. how often do you like to do HDDErase, and reimage etc. and for the regular people who dont know dick about computers, they dont ever do it and never will. of course this is the market intel wants anyway.
But, heres the point, you have SF drives in a degraded state losing thirty percent of performance easily. Very easily. Their GC is shit. takes weeks to recover, and that is company line from OCZ themselves. sprinkle in C300 degrading even faster than the Sandforce, and you have drives that in real world are slower than the current intels.
So now we have intels that are coming out and their number one stated intention in all of these IDF presentations for two years has been: improve steady state performance. Reduce degradation to nil. that is their goal and stated intention, long before these things have come out.
if they reach that goal, which my money says they have, these drives will smoke even the faster SF drives coming out.
 
Can you give some links supporting that? Specifically the poor GC on SF and C300 drives, and the 30% lower steady state performance.

I was under the impression (based largely on Anandtech's articles) that SF garbage collection was the best around (they were specifically recommended for OSX because of this). Additionally, I thought that the TRIM performance of the C300 (and SF for that matter) was pretty good. Generally I was under the impression that the steady state performance was at most a few percent less, not 30% (assuming of course that you're not hammering it so much that TRIM/GC doesn't get a chance to do their job).

If I'm wrong, I'd appreciate some data, but what I've seen so far suggests the contrary.

Computurd said:
Heres the deal guys, lets talk steady state performance. Steady state performance of these drives taht are supposedly faster than current intels (C300 and SandForce, im looking at you) is LOWER than the specs of the current intels. Once you run these drives awhile, they suck. their performance goes to shit after a few weeks. how often do you like to do HDDErase, and reimage etc. and for the regular people who dont know dick about computers, they dont ever do it and never will. of course this is the market intel wants anyway.
But, heres the point, you have SF drives in a degraded state losing thirty percent of performance easily. Very easily. Their GC is shit. takes weeks to recover, and that is company line from OCZ themselves. sprinkle in C300 degrading even faster than the Sandforce, and you have drives that in real world are slower than the current intels.
So now we have intels that are coming out and their number one stated intention in all of these IDF presentations for two years has been: improve steady state performance. Reduce degradation to nil. that is their goal and stated intention, long before these things have come out.
if they reach that goal, which my money says they have, these drives will smoke even the faster SF drives coming out.
Click to expand...
 
well there is data everywhere. but here is some for starters with incompressible data...
the SF is slower than intel with that, and that isnt degraded....

http://www.bit-tech.net/hardware/storage/2010/10/12/crucial-realssd-c300-review-128gb/3

here is some testing from a guy @ 2cpu.com....bear in mind this is READ ONLY performance with the
SF drives degrading:

http://forums.2cpu.com/showthread.php?t=97485


other stuff...here you can see the difference well
http://www.bit-tech.net/hardware/storage/2010/09/03/ocz-revodrive-review-120gb/3



here is a thread where we talked about some of SF weaknesses and real performance. really read john4200 posts he tells it better than I,

http://hardforum.com/showthread.php?t=1552424

as to the GC sucking, well there is a big link floating around where Tony @ocz explains to people to quit freaking out about their drives and that the GC for SF can take a WEEK to repair the drive after big write sessions. i have seen it linked in like twelve forums LOL i will find it for ya gimme a bit...
 
Last edited:
alrighty then finally found it....this is from OCZ themselves....but guess i over reached with the whole WEEKS thing...but A WEEK is more like it.

http://www.ocztechnologyforum.com/forum/showthread.php?73537-Windows-Write-Cache-Flushing-discussion.&p=528034&viewfull=1#post528034


ok and here is the thread that this is really gone into in depth over at XS we even get some OCZ rep action responding to the issue...very good thread :)

recap of thread here...(kinda, just meat of the debate)
http://www.xtremesystems.org/forums/showpost.php?p=4581793&postcount=11

then whole thread here

http://www.xtremesystems.org/forums/showthread.php?t=260409&highlight=knock+ocz
 
The goal of mass-market SSDs are always

1. Enthusiasts as test bed, everybody is aware. [both vendors/users]
2. To address the initial storage ecosystem surrounding 10K-RPM, entry-level 15K-RPM infrastructure for light duty servers
3. Then to address heavy duty large scale enterprise 15K-RPM infrastructure
4. When all premium segments done, finally compete on absolute price in orderly fashion for high-to-low mass market consumer segments.

The point 1 can pay, but the total revenue is not large.
The point 2 is the high-margin volume play. Companies can pay, but almost all of them have been conditioned by the consistency in mechanical enterprise storage technologies and expect same functionality and reliability with minimal care, probably for 4-6 years for storage
The point 3 is the premium market with full scale sw/hw/services infrastructure, probably with absolute minimum tolerance of unexpected failures. Well-tested, recovery and probably many more complicated issues must be addressed.
The point 4 needs no further explanation.

While SSD technology is still being developed, some companies/users are trying to address all segments in the intermediate steps in parallel to overall macro transition. Nothing wrong. For those seeking the bleeding edge, they are doing pioneering works.

Cheers
 
ty lightp2 for your absolute genius, after i read that i decided that my whole view of life as a whole was skewed, and i will never question anything, ever again. you are like the Jesus of computers. thanks man for clarifying life and everything thereafter in one post.
 
Dear Computurd,

There you go, as stipulated previously, it is an opinion, not facts. I think the issue is perception.

Remember, do not believe everything written on the Internet. :)

However, further info to explore.

Technology investment are gearing for high growth, high return for many years. All the tier-1 mega-corps or start-ups follow the same path of milking enterprise first to get their money back in the shortest amount of time.

The ONLY current wellknown mega-corp alternative to this scheme is Apple, where it is operating from the complete reverse of extremely high premium milking consumer 1st. After that, many mega-corps with consumer interest now line up to attempt the same recipe.

To be fair to Apple, it seems they are contributing something worthwhile to premium customers such that they(customers) are willingly forking out their own money to make Apple a success.

Cheers
 
Last edited:
what i dont understand is the relevance. you are so off topic it is ridiculous.

i say this

THE SKY IS BLUE

this is something that we all know, no one in here is disagreeing about it, and no one is even TALKING about it. this is the same as your post.
information everyone has a general idea about, but does not care to talk about, because they weren't in the first place.
 
Last edited:
Computurd said:
Heres the deal guys, lets talk steady state performance. Steady state performance of these drives taht are supposedly faster than current intels (C300 and SandForce, im looking at you) is LOWER than the specs of the current intels. Once you run these drives awhile, they suck. their performance goes to shit after a few weeks. how often do you like to do HDDErase, and reimage etc. and for the regular people who dont know dick about computers, they dont ever do it and never will. of course this is the market intel wants anyway.
.....................
So now we have intels that are coming out and their number one stated intention in all of these IDF presentations for two years has been: improve steady state performance. Reduce degradation to nil. that is their goal and stated intention, long before these things have come out.
if they reach that goal, which my money says they have, these drives will smoke even the faster SF drives coming out.
Click to expand...

Dear Computurd.

The quoted portion is from your post (I think 16).

There are a lot of expectations floating around. You tried to explain around from this angle. My posts serve similar purpose. Basically opinion such that

1. Intel is not after initial maximum read/write rate. Read from various SSD threads some people are high in anticipation of G3 jump

2. However, just opinion point 1 is not enough, many users will ask WHY? "you can even sense those in this forum/thread already have somewhat much greater perf expectation" so your explanation above about steady state performance,

3. And steady state performance is not a new problem, some competitors products already have lesser steady state performance even to current G2 for many scenario, why Intel needs to continuously focus on your suggested scenario? and not push spec just like everybody else? My "extremely plain posts" to postulate where Intel's attentions are. OK, my suspect. Basically, Intel next focus, to address all these stuff in anticipation for greater push into server and high-end, opinion, not facts.

As a result, if you are looking for crazy 6G monster benchmark run, maybe this is still not the focus yet.

Cheers
 
Last edited:
lightp2 said:
The goal of mass-market SSDs are always

1. Enthusiasts as test bed, everybody is aware. [both vendors/users]
Click to expand...
This seems like it was the goal of the G1/G2s in large part.

2. To address the initial storage ecosystem surrounding 10K-RPM, entry-level 15K-RPM infrastructure for light duty servers
Click to expand...

This is where Intel has been concentrating a lot of effort in the past two IDFs. Getting people to understand write endurance, getting standards in place for measuring write endurance, publishing write endurance numbers and metric/graphs of options to trade off capacity for endurance. Getting infrastructure in place both in software and hardware for monitoring and prediction reliability and endurance.

Intel is currently the only SSD maker that I'm aware of that has published detailed information on write endurance and endurance issues.

3. Then to address heavy duty large scale enterprise 15K-RPM infrastructure
Click to expand...

Intel from last years IDF has apparently teamed with Hitachi/HGST to target this market with native dual port SAS 2.0 drives with 10+ PB write endurance. The slides/talk references that while intel/HGST would co-design the drives, they would only be sold and branded by HGST.

For reference this is the HGST SSD FAQ: http://www.hitachigst.com/tech/techlib.nsf/techdocs/F133CBBB37AD6E968625763B00054905/$file/SSD_FAQ_final.pdf
 
My first SSD was a Patriot Warp with a (early) Jmicron controller. You should see the poor thing now. Still works....but you guys have no idea what "degraded performance" feels like....... :(

The poor thing can muster about 20MB/s writes and it thrashes alot now-a-days (its in a laptop now).
 
odditory said:
understood. to me the idea of "reliability" is only as significant as the warranty period. if two drives from two different makers have the same 3-year warranty but one of them performs 33% better than the other, guess which one i'm going with.
Click to expand...

So you bought some of the Patriot Torqx drives with a 10 year (it now says limited lifetime) warranty then? The same ones that got numerous 1-egg reviews on Newegg for their questionable reliability?

I don't care if the thing had a 20 year warranty and performed better than Intel...I'd much rather go with the proven solution and be confident that my data will still be there in the morning. I'll gladly sacrifice some performance in sequential reads/writes in order to know that I have a very minor chance of my drive failing. Backups are great, yes, but most users don't care about increased performance if the product has questionable reliability.
 
john4200 said:
The read looks low. G2s get around 250 MB/s or even a little more. I guess there was either something wrong with the benchmark run, or else the prototype unit does not run at full speed.
Click to expand...

You're right. I was just looking at some of my past tests and the G2s were 240-252 MB/s.

Testing done on an AMD chipset?
 
You must log in or register to reply here.
Back
Top