I'm writing this just to share an experience. I'm aware that some are reluctant to jump into the SSD bandwagon. If you are one of those ... please stop reading and move on.

Just last week I got myself a new computer rig for 2D/3D use. I weighted many options and then decided to go for a 6Gb, i7 920 quad-core set-up with a very fast traditional hard disk as a systems disk (150Gb 10000rpm Velociraptor), couple of big data disks and a 60Gb OCZ SSD.

I'm running 64-bit Ubuntu Linux on the system and Poser executable resides on the systems disk. I then have one external Runtime mounted on the SSD. I must confess I was surprised at the start-up time of Poser. It takes 5-7 seconds to start Poser (after ra boot-up; 2-3 seconds if it has been running previously) with the external Runtime already mounted. It took over 1 minute in my previous installation with traditional disks. Most of the time goes to IO operations as Poser reads thru the Runtime and builds a menu system. (If anyone is interested, I decided on using the ext2 filesystem on the SSD with no saved access-times; not the fastest available file system, but very reliable and robust against file fragmentation).

The SSD solution has proven to be worth it. Loading a fully injected M3 ... 3 seconds. Loading Apollo ... 2 seconds. List goes on. Applying those 4000x4000 pixel textures from a pose file ... almost instant. The speed difference is just amazing. SSDs are at their best reading contiguous large files and many Poser files are just that. There isn't much writing need on the Runtime apart from installing new stuff and as the main app resides on a traditional disk, also the temporary directories are not on the SSD.

So ... if anyone has been thinking about placing their Runtime(s) on a SSD and has some extra bucks to spend, go for it! The 60Gb SATA II SSD set me back for over USD250, but I'm very much enjoying the speed benefits.

Quote - If anyone is interested, I decided on using the ext2 filesystem on the SSD with no saved access-times;

Did you use the same file system for the HD installation, too? Otherwise comparing file systems with and without saved access times is quite pointless for the startup time comparision...

"Wow, these apples are delicious! Much nicer than the ones we bought last week. Here, try one!"

"Ah, you say they are nicer, but did those apples come from the same orchard? Are you being fair with your comparison? Are you perhaps comparing apples with oranges?"

"Darling... shutup and eat the buddy apple".

No offence meant, just a bit of fun... and thanks for the information.

Cool! You should try linking from your render cache directory to that drive as well.

Out of curiosity, why did you use ext2 vs ext3? 

Cheers,

Rod

Quote -
Did you use the same file system for the HD installation, too? Otherwise comparing file systems with and without saved access times is quite pointless for the startup time comparision...

In fact the previous installation had an ext3 file system (improved file system over the older ext2) on the disk where the Runtime resided. I also tested this by simply plugging the old disk in the new system and pointing Poser to use that external Runtime instead of the one on the SSD. There are significant differencies on start-up and model and texture loading times between the two.

So I think the comparison is reasonably valid. I'm not actually really interested in getting to a technical feud, but rather wishing to convey a positive experience to other people who maybe thinking if SSDs would provide any benefit to them.

The startup time hasn't been an issue for me before and it is not an issue now either. However, any and all improvements in loading times of models, textures and whatnot are a very big issue for me. The newly gained smoothness of operation has a big usability impact on me as a user.

Quote -
Out of curiosity, why did you use ext2 vs ext3? 

Ext2 has no file system journaling. Ext3 keeps a journal of every write event on the disk. Journaling makes the file system withstand even sudden blackouts without data loss, but at the cost of extra disk write operations.

So ... ext3 puts more write operations on the disk and I just wanted to avoid that because SSD write operations are less effective than SSD read operations and also because SSDs have a limited write life-cycle per storage cell. I may not ever bump into the (theoretical) write limit, but nonetheless it is a factor. I could have just tweaked ext3 into writing less, but I didn't have any real reason not to use ext2.

An even faster filesystem could have been an option, but I wanted to stay on the ext filesystem wagon because ext4 is just around the corner and updating the filesystem from ext2 to ext4 can actually be made "on the fly" without formatting the drive. If it turns out ext4 is better suited for SSD (which it can be, not sure yet), I can switch to that with no fuss at all.

How about rendering an animation in a high quality firefly setting, does it have any difference?

How about your other programs like your 3d modeling application?

Quote - How about rendering an animation in a high quality firefly setting, does it have any difference?

I do not know. I've never animated anything with Poser. I'm strictly interested in toony/comicky still images myself. That being said, I do not believe disk-IO has any significance in animation rendering.

Quote -
How about your other programs like your 3d modeling application?

I've no idea how they would perform on a solid-state disk. The solid-state disk on my system is reserved for Poser only.  I would imagine that anything requiring lots of data to be read from a disk benefits from a SSD setup to some degree.

Quote - SSDs are at their best reading contiguous large files and many Poser files are just that.

Quite the opposite. Traditional harddisks perform best with sequential reads, and much slower with random access reads (and writes are still slower). The main advantage of SSDs is that their random access speed is just as high as their sequential read speed.

In fact, the sequential read speed for large files of a fast traditional harddisk should be higher than the read speed of an SSD.

It may very well be that Poser is reading files quite inefficiently, hopping back and forth through the file. In that case, an SSD would indeed be significantly faster than a traditional disk.

Well for me, price is going to outweigh speed every time. I can get a 1 Terabyte drive for around $100. Comparing that to your $250 for 60 gigs and the SSD's will loose me every time.

Quote - > Quote - SSDs are at their best reading contiguous large files and many Poser files are just that.

Quite the opposite. Traditional harddisks perform best with sequential reads, and much slower with random access reads (and writes are still slower). The main advantage of SSDs is that their random access speed is just as high as their sequential read speed.

In fact, the sequential read speed for large files of a fast traditional harddisk should be higher than the read speed of an SSD.

It may very well be that Poser is reading files quite inefficiently, hopping back and forth through the file. In that case, an SSD would indeed be significantly faster than a traditional disk.

You are very right of course on the sequential read issue. I should have formed the quoted sentence more clearly. The meaning was to convey that SSDs have superior reading speeds on regular and "normal" data fetching tasks and that Poser uses a LOT of such disk activity to even start up and very much so when "used normally; ie. doing Poser stuff.

I made some measurements on my own system to confirm my "gut" feeling about the speed. I also gathered some hard core io and cpu statistics on the time when Poser "does stuff" on startup. And also during regular use, but the startup statistics are much more tangible to grasp.

I studied the amount of data Poser reads when it starts up in a case the computer was just booted so as to eliminate any kind of data caching. During a 5-second startup period Poser read a total amount of 36Mb of data (from the 50Gb Runtime; I'm only interested in the performance of the SSD at the moment and didn't calculate anything else like the data activity to start the actual program from the Velociraptor disk). The middle 3 seconds of that 5 second period the SSD was pushing 9Mb/sec (17500 sectors of data read per second). Ok... that was fast.

I then did the same test on the same data on a regular disk (500Gb Samsung SATA disk purchased 1 year ago that housed the Runtime on my old system and has as similar filesystem). Mounting the Runtime took 49 seconds during which time the same amount of data (roughly 36Mb) was read. But the data read per second was spread throughout this time. On average the regular disk performed reads of around 2000 sectors / second (less that 15% of the speed the SSD was capable of). Reasons for this was because no matter how fast the drive read the data, it was not enough to satisfy the IO processes demanded by Poser. This is revealed by examining the so-called "drive saturation" value that tells us how close the drive is to being fully utilized. That value was in fact between 85% and 100% during the startup period. It tells us that the IO processes are merely wasting time just waiting for the disk to return data. Every read request sent to the drive had to wait in a read queu for around average to 5 milliseconds (ranging from 2-15 milliseconds). On the SSD on the hand the drive was only half saturated (50% utilisation; it can handle more, but Poser wasn't fast enough to request more stuff) and average wait time for read requests were only a steady 0,25 milliseconds.

On absolute sequential read testing the Samsung 500Gb lost only by 30% against the SSD, but on absolute random access file reads SSD outperformed the Samsung by a factor of ... a lot (0,29 milliseconds of average file seeks against 17,4 milliseconds on the Samsung).

I'm using my Samsung drive as an example drive here because that's a "normal" hard disk. 7200 rpm disks are probably the norm at the moment among the vast majority of computers. And because it was conveniently already housing my entire Runtime. Oh and data fragmentation is not an issue here.

I also gathered data on processor utilisation during Poser operation and it barely registers on startup. With this kind of data in hand it is relatively safe to say that disk-IO is in fact a major factor in Poser usability and that a SSD can provide enough throughput to satisfy Poser into performing as fast as it's coding allows. 

This was my experience and my studies on my current system based on actual use case. I'm sure the small amount of data presented here can be equally useful to a Windows-based system as well. NTFS puts a lot more strain on the CPU(s) during disk-IO activity so that part at least is going to show a difference (ie. more CPU activity when Poser reads stuff). The main item of this long-winded post seems to be: your hard drive maybe a serious bottleneck on your quest for a "smooth-use" Poser heaven. 

Quote - Well for me, price is going to outweigh speed every time. I can get a 1 Terabyte drive for around $100. Comparing that to your $250 for 60 gigs and the SSD's will loose me every time.

I can too. But I already have 2,5Tb (and then some) so I don't actually really need any more of  the same-o-same-o. I could also have spent the money to get a better display card or the next step up of the processor family. But then I would be still be running Poser without the speed benefits I now experienced.

Naturally every user has to maintain a balance between free hard disk space and other factors - including speed issues. I respect that. That is exactly why I said what that tiny drive cost me. I too automatically compared the price tag to a mechanical regular hard drive at first. 

I may consider it when prices tumble (as they always do). But like Lost, no way I'm spending that sort of money, especially when I'm sure I have well over 60 GB of content. We'll see....

but maybe not a 60gig runtime.

You're right. I just checked and it turns out my runtime is only 40 GB. :D

But I still wouldn't buy a drive that I was going to fill 2/3 of the way right off the bat. :)

My "Multiple" runtimes are well over 100 Gig's so 60 gig's is just laughable.  And again, I'm on a fixed budget, now threatened by some car repairs. I have to go for the most bang for my bucks if I'm going to go at all.

Exactly.

For perspective, I only began building up content in August 08. If I've got 40 GB already, no bloody way I'm going to settle for that kind of bang for buck on disk space. Even though my purchasing and downloading has definitely slowed down... who knows, a 60 GB drive may not even last 2009.

It's alright... I'll settle for my present situation until I get a more powerful computer in a couple years. By then disk space should be way down in price and the options will probably be better too.

As said, balancing of spending needs to be made per user.

I'm not trying to force anyone to buy anything, just saying that in my case it worked wonders. There's some additional considerations too. The life-span of SSDs will most likely well exceed that of mechanical hard drive disks since they do not suffer from mechanical failures. If the SSD is not used as a heavy-duty systems disk, it will most probably never exceed the theoretical limits of writes per memory cell. (I've seen calculations that estimate that a normal use of SSD for 6 hours daily lasts for at least 25 years until it can no longer be written to; after that it still retains the data as a read-only drive - it doesn't go anywhere).

Thus it may be a bit unfair simply comparing traditional disks directly to SSDs since they do not share enough common attributes. Granted, both are used to store and retrieve data. Nevertheless, pricetags of SSDs are dropping fast and the tech that makes the drives tick is evolving much faster than the technology behind traditional drives. A year from now I bet a lot of users here have SSDs housing some of their data and many of those drives are superior to what I now have. In essence it also means that this post and the data I provided here is hopelessly out-of-date in, say, 6 months.

Still Poser is not going to go away and my simple calculations showed that in my system io-operations are no longer bottlenecking performance in regards to Poser use.

Quote -
For perspective, I only began building up content in August 08. If I've got 40 GB already, no bloody way I'm going to settle for that kind of bang for buck on disk space. Even though my purchasing and downloading has definitely slowed down... who knows, a 60 GB drive may not even last 2009.

Unrelated to original issue, but that's some mighty content growth-rate you got there!! Wow!
Naturally with such a growth rate you'll surely be needing one of those cheap-o 1Tb disks! A no brainer there.

I've been a heavy user since ... well ... many years, but still I've retained a relative small single external Runtime (50Gb currently). I'm also actually talking about the size of the actual external Runtime folder only and not the whole Poser folder. Everything that Poser doesn't read and use has no place in my Runtime.

Personally I do lots of things to stream-line and prune my Runtime. I move lots of stuff just out of the actual Runtime to storage areas or just delete altogether since I'm never going to use those inside Poser anyway (.bmp templates, Camera sets, readme-files, etc etc). I also automatically run pngcrunching and jpeg optimisation and all sorts of automated tasks and get rid of essentially anything I don't have a use for. My external Runtime also doesn't include any settings folders, Python folders or the sorts (those are all part of the main application folder structure and not part of the external Runtime in my setup).

I think I can keep my current external Runtime within the confines of 60Gb for at least two years with my current growth rate. (It may actually shrink some since I'm considering getting rid of M2 and V2 altogether this spring). If not I'll just spawn the exceeding part to another external Runtime and house it on a traditional disk. No biggie.

Those are my runtimes only.

Actually, not counting my Poser Pro runtime, which houses stuff like lights and python stuff.

well now, Newegg has a 120gb  SSD for $265. things are looking up :)

But who can say for sure it's the SSD's that do make the difference? It may well be the complete new setup causing the huge improvements., actually iw probably is a huge part of the new speed improvements.

Last year I upgraded my system.... except for the harddisks. My poser runtimes are external runtimes. On my old system with the same harddisks, it took 4-5 minutes to start up poser and it at times took over a minute to load an external runtime. The new system takes less then a minute to start up Poser and all of the runtimes open within 10 seconds, at the most 15. That's all due to changing motherboard, new cpu and adding more memory, it had nothing to do with the harddisks at all, they're still the same.

To know for sure you have to do the same tests on a harddisk that's in the system. I've got the feeling that if you do so, you soon find out that the extra speed improvements (if any) isn't worth the huge amount of extra money paid for the SSD's.

Quote -
To know for sure you have to do the same tests on a harddisk that's in the system. I've got the feeling that if you do so, you soon find out that the extra speed improvements (if any) isn't worth the huge amount of extra money paid for the SSD's.

Or you could have just read the post a few posts up to see that I did test it that way. 

I've now tested the exact same Runtime in 4 different disks (I got 4 disks in this setup) in this same system. SSD is by far the fastest and this is not just an empirical "feeling", but also based on facts that are gathered by inspecting the actual i/o-traffic that happens when Poser "does it's thing".  

Sorry didn't read that one, only the original post. 

Thanks for the data, 12rounds. Since Newegg has those 30gig SSD's for less than $100, it looks like the next upgrade of the main machine will have a pair of those in RAID 0 for a boot drive (wonderful speed up, putting the OS on its own drive(s) and no apps or data. Keeps things nice and clean).

The only issues I can see coming up are power consumption and read/write reliability.

I just hope the upcoming mobo's have enough SATA connectors...... 

Thanks for the heads up. I knew silicon drives were coming, but that this level of function and quality is available now is very good news. I wonder how much how many gigs will cost by Christmas?

For comparison shopping to NewEgg I suggest also TigerDirect.com. Can anyone else suggest a source with exceptionally good prices?

For agressive runtime maintainance, remember that Poser will respond to a Short-Cut the same way it will to a folder. So putting a short cut to a directory that contains content is the same as installing that content in the runtime. Short-Cuts to seldom used content (stored on other disks) take very little disk space, but can save the time of reinstalling things rarely used. It occures to me suddenly to wonder if Poser can access content in a zipped file by way of a Short-Cut? I will have to try that out.

LMK

Here are some stats I've gathered thus far.

Fastest traditional hard disk I have:
Western Digital Velociraptor (10000rpm, 150Gb SATA/II disk, ext3 filesystem). Powered on for 68 hours. For home users the Velociraptor and Raptor disks are perhaps the best ones out there. Investing into SAS disks would give even better results, but then again the costs get easily out of hand as well.

Poser mounts the Runtime in 11 seconds. Disk delivered  a very uneven load pattern (sectors read per second varied from 1920 (1,0Mb) to 7340 (3,7Mb) sectors per second). Disk utilisation was very even (51% to 68%). Average I/O wait times between 0,67 to  2,14 milliseconds per request. 

My old disk (Samsung SpinPoint T166, 500Gb, 7200rpm SATA, ext3 filesystem). Powered on for 2088 hours. Purchased less than a year ago so it's not exactly an old drive.

Poser mounts the Runtime in 40 seconds (a slight improvement over yesterday's 49 seconds... I've no idea why). Disk gave it's best giving a disk utilisation level of 82-92%. Read processes were forced to wait for around 6 milliseconds (which is 3 times worse than the worst cases of the Velociraptor). Data read and delivered to Poser was between 976 sectors/second (0,5Mb/sec) and 3775 sectors per second (1,9Mb/sec). Performance is piss poor compared to Velociraptor.

My SSD (OCZ Core V2 60Gb). Formatted to ext2 filesystem. 
There are huge performance variations between manufacturers and disks, but this disk model is not the worst of the bunch - it's also quite far from the best models out there.

Poser mounts the Runtime in 5 seconds (exactly the same performance as I got earlier).
Disk utilisation was at a constant 50%. Read processes were waiting at worst 0,32 milliseconds before being served (note also that the time to both put the request on the chip and take it for processing is counted in the wait time).  Data delivered to Poser was a constant 17000-17400 sectors per second (9Mb/sec).

Considering the 150Gb Velociraptor disk costs USD180 and the Samsung 500Gb costs around USD80 currently, there's a very significant performance difference between the two. Housing the Runtime on 10000rpm disks would still be a very expensive solution compared to 7200rpm disks, but would give a good performance boost as well. Using the SSD to house the Runtime has proven to be at least twice as fast as the Velociraptor and is very tangible in some things like loading up models and textures. It's also VERY expensive when compared against a normal 7200rpm disk. 

Okies. That's it for now. Of course these stats are from MY system and on a Linux box. There is perhaps (?) a significant impact imposed on by the NTFS filesystem in Windows-based systems. It would certainly be very interesting to see what kind of results Windows users get on a similar setup. I would think the details differ a little but the stats would remain in the same ballpark.