Intel M25 SSD not so good after all....

[nanofrog]

Yes most people are saying here or suggesting to use them as boot drives. But that's because of the seek speed (ie. it's like, 0ms ) and not really because of the capacity.

For me, these are the reasons for sticking to OS/app use ATM.
1. The random access times are lower than any other technology.
2. Cost/GB is terrible for those on a budget.
3. OS/apps usage is more read than write, so it helps eliminate write cycle frequency.

To me, the MLC/SLC flash memory having a UBE of 1E4 or 1E5 is too low. Even with wear leveling and TRIM drivers, it still can't hit 1E14, let alone 1E15 of enterprise drives. This seems to be ignored from a lot of information I've seen, and bothers me.

If someone can prove that it can go that high, I'd be both amazed, and appreciative.

They're not faster than a 3-drive RAID when it comes to video editing so the logic here I assume is based on the price per gig. You can either have 256 Gigs at 300 MB/s or you can have 3 TBs at 350 MB/s for the same amount of money.

This is something I can't ignore either.

Also these SSD drives do indeed come in larger capacities up to 1TB but I dunno price or availability.

So far, I haven't seen any that large. I don't check fanatically though.

 

[Tesselator]

For me, these are the reasons for sticking to OS/app use ATM.
1. The random access times are lower than any other technology.
2. Cost/GB is terrible for those on a budget.
3. OS/apps usage is more read than write, so it helps eliminate write cycle frequency.

To me, the MLC/SLC flash memory having a UBE of 1E4 or 1E5 is too low. Even with wear leveling and TRIM drivers, it still can't hit 1E14, let alone 1E15 of enterprise drives. This seems to be ignored from a lot of information I've seen, and bothers me.

If someone can prove that it can go that high, I'd be both amazed, and appreciative.

I hadn't considered #3. I was told by some people here that it's a non-issue. <shrug>

Good point on the UBE too!

So far, I haven't seen any that large. I don't check fanatically though.

What's the largest commonly available these days? 128 or 256, or?

 

[TommyMacBrown]

Intel will soon be history

Apple is hiring all kinds of chip people in Calif. They hired away one of the top people at AMD, and some from Intel. They will soon be making their own chips based on their own designs.

 

[Tesselator]

Apple is hiring all kinds of chip people in Calif. They hired away one of the top people at AMD, and some from Intel. They will soon be making their own chips based on their own designs.

That's very bad news if true.

A few things need commonality and compatibility between platforms. CPUs, RAM, and storage units HDDs, etc. are among them.

 

[dr. shdw]

That's very bad news if true.

A few things need commonality and compatibility between platforms. CPUs, RAM, and storage units HDDs, etc. are among them.

Not if they can make the chips smaller and better..

 

[Pressure]

Apple is hiring all kinds of chip people in Calif. They hired away one of the top people at AMD, and some from Intel. They will soon be making their own chips based on their own designs.

No they will not.

They will be making System on a Chip (SoC) with the products they license (think ARM, Qualcomm, IMG, PowerVR and Broadcom to name a few whose chips they have licensed in the past).

But this is primarily for the iPod and iPhone.

 

[Tesselator]

Not if they can make the chips smaller and better..

Even if they're very small and much better it's still bad news! Unless of course they sell them on the open market like Intel, AMD, IBM, and Sun do. Then it won't be so bad maybe.

 

[grue]

Apple is hiring all kinds of chip people in Calif. They hired away one of the top people at AMD, and some from Intel. They will soon be making their own chips based on their own designs.

For the toys like the iPhone and iPod, perhaps. Not for the computers.

 

[bruinsrme]

Not if they can make the chips smaller and better..

Its not that easy and inexpensive to make something smaller and better.
If you thing they are going to open up a fab and start pumping out chips you have go to be kidding. It takes Intel over 4 years to really get a fab built, to "high volume" and yielding well.

I can see them designing their own chips but still have a chinese company manufacturing the product.

When the litho steppers, scanners and immersion systems costing billions to put a fab together.
if you think they can do that in China they should ask Intel how their China venture is coming along.

 

[J&JPolangin]

...SSD tech is still to new...I'll wait, thanks...

 

[nanofrog]

Even if they're very small and much better it's still bad news! Unless of course they sell them on the open market like Intel, AMD, IBM, and Sun do. Then it won't be so bad maybe.

For some strange reason, I'd think issues would surface over control. Development tools and standards could be likely examples.

This sort of situation might make 3rd party manufacturers to shy away from it if offered up as a commodity component.

For the toys like the iPhone and iPod, perhaps. Not for the computers.

Their marketshare of the iPhone and iPod seems to support such a move.

I'm not convinced of this on their computer lines however. Costs would be too high, as they don't make as many systems as the vendors using Intel parts do (minus Apple) to divide the development costs over. It would result in a system no one would be willing or possibly able to afford.

Its not that easy and inexpensive to make something smaller and better.
If you thing they are going to open up a fab and start pumping out chips you have go to be kidding. It takes Intel over 4 years to really get a fab built, to "high volume" and yielding well.

Absolutely. Smaller devices are easier to design and debug. Cheaper to produce as well, as more devices fit on a wafer.

I can see them designing their own chips but still have a chinese company manufacturing the product.

Generally speaking, this is the way things are done. OEM or ODM facilities are used. In Apple's case, I think it's a mix, and might want to shift to more towards OEM for control. QC might improve this way, but that would depend on the details.

When the litho steppers, scanners and immersion systems costing billions to put a fab together.
if you think they can do that in China they should ask Intel how their China venture is coming along.

Building a fab is definitely an expensive venture. IIRC, Intel spent ~$3.3 Billion USD on their facility in Vietnam.

 

[Tesselator]

For some strange reason, I'd think issues would surface over control. Development tools and standards could be likely examples.

This sort of situation might make 3rd party manufacturers to shy away from it if offered up as a commodity component.

Excellent point. If the Apple architecture isn't backwards compatible there will be a huge lag in development during which time other platforms will gain market-share. It will also mark the 4th architecture change is a relatively short period of time which would scare me as a developer for sure. I would consider dropping Apple development at least for a few years. Heck, just the move to PPC instead of Intel almost caused Apple's death. The move to Intel put them back on the map and in the game. Now they want to undo all that? That would be insane! Truly!

 

[nanofrog]

Excellent point. If the Apple architecture isn't backwards compatible there will be a huge lag in development during which time other platforms will gain market-share. It will also mark the 4th architecture change is a relatively short period of time which would scare me as a developer for sure. I would consider dropping Apple development at least for a few years. Heck, just the move to PPC instead of Intel almost caused Apple's death. The move to Intel put them back on the map and in the game. Now they want to undo all that? That would be insane! Truly!

On the computer side, your analysis makes sense IMO.

For the hand helds, introducing a new part (application specific part or line), might be the way to go. They already license ARM, and a customized version(s) could allow for the addition of features, particularly via an SoC. Then there's the possibility of shrinking a design, which Apple seems to really like.

It should save money as well on other licensing and/or separate components. 4G for example, might find it's way on the die. New features other product lines don't offer would also be a possibility, giving Apple an edge. Development tools are more specialized in this realm anyway, and the standards aren't as ubiquitous either. Differing carriers network implementations alone make a single product line difficult at best.

 

[No4mk2]

Originally Posted by nanofrog
For me, these are the reasons for sticking to OS/app use ATM.
1. The random access times are lower than any other technology.
2. Cost/GB is terrible for those on a budget.
3. OS/apps usage is more read than write, so it helps eliminate write cycle frequency.

I wonder if this is how one will feel after use. I bought a fast Mac, but the thing that pushed it over the edge was one 80 GB Intel SSD used as a boot/apps drive. Just that alone made a difference that was not subtle in detecting. Even three in a striped RAID doesn't feel much faster, though MBS hovers around 730. I think they are the future, with refinement. I urge everyone to try one when the market or pocketbook allow.

 

[nanofrog]

I wonder if this is how one will feel after use. I bought a fast Mac, but the thing that pushed it over the edge was one 80 GB Intel SSD used as a boot/apps drive. Just that alone made a difference that was not subtle in detecting. Even three in a striped RAID doesn't feel much faster, though MBS hovers around 730. I think they are the future, with refinement. I urge everyone to try one when the market or pocketbook allow.

This is my point. They're the future, but still need time to mature and costs to fall to a more affordable level.


The other part, has to do with the write reliability. High write system use would kill them too quickly, given the limitations of SLC/MLC flash memory. OS/apps usage is primarily random reads, which is what these drives excel at. If someone has the cash and is willing, it's a nice way to go for single drive performance.

But given the current pricing, and capacity limits for some, the cash can be applied to a RAID setup. It is fast, but not as good on random access. Overall throughput can exceed a single SSD though, depending on the array type and drive quantity. For raw performance, and no redundancy (same with a single drive), RAID 0 is applicable for comparison IMO. Given a 3 disk stripe can hit 300MB/s STR, it can exceed the single SSD (Intel's X-25 series for example @ 250MB/s STR).

$300 for mechanical drives vs. $314 (80GB X-25M). Very similar cash, so the decision can be centered on use. Particularly the expected write frequency or random access reads. Greater capacity needs changes this more in favor of a mechanical drive based RAID system. High write frequency does as well IMO. Others may differ on this, but I can't help but think current SSD's in high write environments are going to experience failures much faster than their mechanical counterparts due to the SLC/MLC flash used.

 

[Tesselator]

This is my point. They're the future, but still need time to mature and costs to fall to a more affordable level.

$300 for mechanical drives vs. $314 (80GB X-25M).

Holy cow!!! I thought it was $300 for 256gig or something... $300 for 80gig??? You would have to be crazy to go for that on a desktop. Seriously.

 

[aibo]

Holy cow!!! I thought it was $300 for 256gig or something... $300 for 80gig??? You would have to be crazy to go for that on a desktop. Seriously.

Why is that crazy to have SSD on a desktop? Are your other 4 drive bays broken and unable to accept 1TB+ drives?

 

[dr. shdw]

Holy cow!!! I thought it was $300 for 256gig or something... $300 for 80gig??? You would have to be crazy to go for that on a desktop. Seriously.

1 X25-M 80gig is equivalent to like 10 15K SAS drives in random write speeds..

 

[nanofrog]

1 X25-M 80gig is equivalent to like 10 15K SAS drives in random write speeds..

I never said SAS was a bargain either.

SATA still rules for low cost drives. For now anyway. SSD will almost certainly take over at some point, but not today.

 

[Pressure]

I never said SAS was a bargain either.

SATA still rules for low cost drives. For now anyway. SSD will almost certainly take over at some point, but not today.

If you like performance, it has already taken over.

$300 is a bargain for the performance increase you are seeing with the Intel X25-M installed. It is the single best purchase you can make to speed your computer up on a day to day basis.

 

[VirtualRain]

If you like performance, it has already taken over.

$300 is a bargain for the performance increase you are seeing with the Intel X25-M installed. It is the single best purchase you can make to speed your computer up on a day to day basis.

I agree... just 4 or 5 years ago I was paying over $300 for a 74GB first-gen Raptor with 8MB cache that pales in performance compared to the SSD's I can get for that money today.

Unlike Raptors, I think SSD's are already much more mainstream than their dinosaur predacessors were in those early days.

I really think that anyone buying a Mac Pro owes it to themselves to spend an extra $300 on a X25-M... it is the most dramatic performance improvement you can make... bar none!

 

[Tesselator]

OK, OK, all you SSD lovers and defenders... I just meant it was too expensive for me. Not at all worth it. It's basically the difference between a 1min. boot time and 1.5 min. boot time. It will also save you about 5 min. a day under HARD use as an OS X drive. Remember that with 12, or 16 gigs of RAM every time you use an app just about all the program code goes into the system cache and stays there typically until you reboot. If I load PS with 100 plugins it takes 32 seconds the 1st time. Then if I quite it and use iMovie, Lightwave, and Illustrator for 3 or 4 hours and then load PS again it takes 7 seconds to load the second time. Going back to Lightwave, Illustrator, or iMovie is the same thing again - ultra fast the second time. Initially I thought Apple was daft for keeping such huge chunks of app vectored in RAM but with hard use it began to make more sense. It does the same thing with bits of the OS as well. And this is not even considering that with 12 or 16 gigs of RAM the user typically does not quit the apps after use. I don't if I know I will be using it again in the next hour or so. With 16 GB RAM you can do that easily with 5 or 6 large applications.

With SSD you're paying approximately 35 to 40 times the price of standard SATA media and all just for that 5 min. per day advantage. To me that's crazy - nutz-o, bonkers. I could see it if it were only 5 or maybe 10 times which is where it would fall if SSDs were $300 for 256GB but they're not. I guess I could also see it on a laptop where RAM is usually low and LT HDDs are hella-slow. But on a desktop at $300 for 80GB which is not even large enough to put OS X on - unless you keep your system very very vanilla and don't install almost anything, I think not. Considering that, we're also limiting ourselves severely for that same 5 min. a day advantage. Yup, sounds "crazy" to me - on a desktop anyway!

I'm not in the habit of lighting hundred dollar bills on fire just to watch them burn and SSD on a desktop is the buyers analogue to doing just that - in my opinion. Also if you're going to consider time in this equation then let's go back to when 100 MegaByte hard drives were $300 and we can justify anything! Pretty silly logic IMO. YMMV. BTW, 4 or 5 years ago it was 250 GB HDDs for $100. We would have to go back quite a bit further in time to see 74GB drives for $300 in common media like IDE, SCSI II / III, and etc.

 

[VirtualRain]

I agree... It's definitely hard to justify on any kind of ROI analysis. The ironic thing is that we are talking about this in a Mac Pro forum where obviously value for money is not necessarily the highest priority (or let's face it, you probably wouldn't own a Mac Pro). I mean everyone who owns a Mac Pro is paying a premium for the privilege so it's odd to hear people in this forum having a hard time justifying 5-10% more money for one of the single biggest improvements you can make to your system.

Just to share more of my enthusiasm for these things...

Over the last 15+ years, every time I built a new PC with the latest processor architecture, I was usually thinking "Wow! what a stunning improvement!"... upon booting it and using it for the first time.

After a few months, I just got use to the new performance until eventually it felt as slow as ever and I ultimately built a new rig... then it was like "Wow!" all over again.

My last couple of PC builds (including the purchase of this 2009 Mac Pro) have not had that "Wow!" factor... ... UNTIL I installed my SSD's over the weekend! Honestly, it's the first time in the last several years where I've said "Wow!" after booting my computer.

In a few weeks, it will all be lost on me as I will become desensitized to it, but for now, it's worth every penny!

 

[Topper]

Gaming

.
Will the Intel X25-M improve gaming performance?
.

 

[dr. shdw]

.
Will the Intel X25-M improve gaming performance?
.

Load times.