I think the MB PCIe port is part of the problem. Even the MB is ony 3 years old, it's still using one PCI Express 1.0a slot and that's where the SATA III card is sitting. I think plugging the SSD into the MB SATA II controller will provide faster test results. We shall see tomorrow. At this point ,it's about finding out why the SSD performance is so low. Then worry about scratch performance.

and btw, you can't expect that windows experience to go up much just by replacing a disk.. that only works if the motherboard has a screaming chipset with tripple or quad memory channels like a core i7.. On slow chipsets, or limited memory channels you can add fast disks all you want, the bottle neck will be the chipset's main databus...

And the difference between the expensive SSD's and cheaper ones, there usually is a difference in type of memory chips used.. the more expensive ones have better chips for writing, and they are arranged for better writes as well.. But it still won't make em fly as a scratch disk on an old motherboard and/or sata controller.

Hmmm.....

"For others suffering from the same problem... miserable Write speeds with your SSD in Win7... simply go into the Device Manager and disable "Write caching" on the "Policies" tab for your SSD. "Write caching" is not needed with an SSD."

I'll have to look into this as well.

i predict you may get 50% increase over the values you posted, at most.

Either way, seems like a lot of people are having speed issues. Manufactures are showing high benchmarks and the public can't get anywhere near them.

if you want their benchmarks, you'll need a spanking brand new, top spec mobo and controller.

I don't expect their benchmarks, but I do expect better benchmarks than I am seeing on the current system. At the moment, the SSD is performing at the level of the Caviar 1 TB drive. That's not right. Okay, read was better on the SSD, but not that good.

that could mean the chipset of the mobo just can't do more.. that Caviar is prolly bottled necked as well

Tom's Hardware did some benchmarks on the Caviar and our tests were slightly better. He was at around 85 MB/s for both R/W.

My Intel 330 120GB SSD gets 250-300Mbps on an older Intel motherboard (DG43GT) with SATA II (not SATA III) controller on the board. Check that drive, something's wrong with it. I also use AHCI for the BIOS setting instead of "SATA" and that helps, too!

I doubt it is the drive, Art. As I mentioned above, the SATA card is in an older PCIe slot. I bet moving the card to a 2.0 slot will improve the times.

Dave,
Definitely hook up the drive to a SATA III port on the motherboard, if your board has them! Add-on boards will always be slower, in my experience. I'm using the Intel 330 SSD right now on my main pc, which is an Intel DG43GT board, 4GB of RAM and Q6600 CPU and GT-430 video card. The SSD has really made a huge improvement in my PC's performance!

The MB is just a bit too old to have SATA III. That's why the card. I'll be over at my parents in a bit and will continue to play. More later. :)

Believe me, I would not have gone for a controller card had there been SATA III on the MB.

Moved the sata card to a PCIe 2.0 slot, redid the speed test and there was a small improvement In r/w times, but still not what it should be 208/178. Plugged the SSD directly into the MB sata ii port, redid the test and ended up with identical numbers. Checked the bios and the onboard sata ports were in sata mode. Moved it AHCI rebooted and..... Had to leave for a family gathering. More in a few hours.

I installed an SSD after shooting with the d800 and while there was improvement, it was not that much.

More testing..

After AHCI was enabled and the SSD was plugged into the SATA II port on the MB, speed test reported R/W at 265/177. I noticed that the new SATA III card was plugged into a 2.0 PCIe port that was forced into 1.1 mode. Changed that in the bios and when I did the speed test again, it was 399/178. Write performance is still lacking, but read sure did jump.

Back to the PS tests and no real improvement. I was able to knock off 6 seconds from one filter, but the other times were the same.


I also turned off indexing for the SSD. In this case, it really isn't needed.

told ya... ssd&write performance = chitty for live action scratch disks
and old mobo's won't deliver the rest either.

Only way to make em deliver adequately for that kind of application is if you raid 0 em.

I'm getting excellent performance from my SSD's, one of which is running at SATA II (AHCI mode) and the other is in SATA III and AHCI mode. R/W scores in the 300's.

Stijn, that's just not quite right. Would you be saying the same thing if I had been talking about a platter drive that had similar performance specs to the SSD? As far as PS, what I left out, is that there is still a point where there is some serious platter disk activity, even though the scratch disk is pointing to the SSD. Our guess... PS needs to access a bunch of info during the use of these filters.

But why is the write performance so low in the other tests? I hooked up the SSD to my moms PC last night as well. Hers is a 2011 model and should have SATA III built in to the MB. Read results were a bit lower than I had expected, but the write time was exactly the same as my dads PC. Mom has an i5 proc, while dad has an i7, though two years older.

The only other real world test I did, I could not time. I tried opening a 50 meg psd file from the caviar drive, where his pix normally are stored and opening a copy from the SSD. In both cases, they opened in the blink of an eye. I could not hit start/stop fast enough on my timer to get accurate readings.

Even more fun!

Went out and bought a second SSD drive with the intention of installing CS6 on it. Now PS opens from one SSD and uses another for its scratch disk. While PS now opens very quickly, the PS tests did not increase much. Only one filter sped up by ten seconds using the 500 meg test file. The other numbers pretty much were equal to PS being on the slowest platter drive (C: ).

Pretty disappointing. Should not matter if it's a faster platter drive or SSD. Speed should increase. Maybe not light speed, but noticeable.

One last test tomorrow and I am out of ideas. I'll uninstall PS from the SSD. Since I have two SSD's at my disposal, I'll RAID 0 them, reinstall PS on a platter drive and set the scratch to use the RAID drive.

After that, I am at a loss. This MB can only handle up to 12 gigs RAM in triple channel mode, 16 in dual. The CPU is an Intel i7 950 quad with HT.

I'm still going to play with my above post, but...

I've come to the conclusion with Photoshop (PS), that no matter what you do to speed up performance, there are internal issues with PS that virtually nothing will fix in most home users PC's, even if they have the latest and greatest. You can have all the RAM in the world and PS will still use a scratch disk. I think the only way to get blindingly fast performance crunching large files - 500 megs plus, are multiple processors. Not just cores on a single processor, but 2/4/6 each having multiple cores....

EDIT: I tried to cripple PS by lowering that RAM it could use to 1 gig. The efficiency meter drop to 99% for 2 seconds, but the total time to repeat the test was identical to when it was allowed to use 10 gigs.

It's the databus on your Mobo..

Take the same disks on a motherboard with a Core I7 and tripple channel memory and you'll get much better results..
Your current motherboard will bottle neck cause it's memory is only dual channel, and it shares the bus to the cpu with the pci ports
For instance an X58 chip the memory is direct talking in bathes of 3 instead of 2 + not sharing the bus with the peripherals..

http://content.hwigroup.net/images/n...ockdiagram.gif

Right now, all your memory and stuff is comparable to that bottom group, sharing that single 2GB bus to the cpu

This is a MB with a Core i7 and triple channel memory. Just a bit older and with not SATA III onboard.

Still, after watching what is gong on, it seems to be PS doing calculations with certain filters that just do not get faster with any amount of RAM.

Seriously, the R/W times on the SSD's that there should be a noticeable gain in time to crunch data with those filters. PS is obviously using the scratch disk, I can watch the file grow. All the while, everything I read says that PS should be using only RAM. The 500 meg image should not run out of the 10 gigs of allocated RAM. The meters show that, but yet the scratch file grows, but even so, it should perform faster on SSD.

Right now, opening PS from a secondary SSD is very fast. PS does have an issue with opening and saving files, in that it uses only one core. When working on the file, it uses multiple cores.

Ok, in that case you'll have to agree that SSD performance is no good for scratch disks..
It's only good for fast read operations


I did the SSD thing, and in the end dumped em and got myself a hybrid instead.
Much more cost effective, fast reads on the stuff you frequently need, and normal writes to the platters.

The Adobe community does not agree with you (they do agree with RAID 0/5). Something else is going on with PS that everything tried will not speed up dramatically. There is a gain, but not what one would expect.

I don't think hybrids would help in this case. They seem to store roughly 4 gigs on the "SSD" side. That's not going to help with PS in reducing processing speeds.

They agree with the Raid 0 because they know SSD has $hitty writes. and Raid 0 is the only way to improve on that.. I'de stay well clear of Raid 5, cause that is a write bottle neckitself. Unless maybe you got 5+ disks in a set, but when you get to that price range there are other , faster solutions like a hardware ram disk..

Actually, they think using SSD for a scratch drive does improve performance, so does Adobe. They just also agree that RAID 0 is even better. Adobe suggests RAID 5 as an option. I'd go with RAID 0 (though I can't try RAID 5 currently anyway) and will with these last tests later.

I'm done testing and am sticking with my above conclusion.

I setup the RAID 0 on the SSD's and redid the speed test. 405/289 R/W. No gain on the read side, but 100 MB/s faster on the write side. Redid the picture tests and no gain. PS was also reinstalled back on the slow C drive. We redid some of the tests moving the scratch file from the SSD's to both large platter drives and times increased by 20 seconds for the large 500 meg pic.

We used a normal image that my dad would be working with and moved the scratch file around to the different drives and reapplied the same filter. No matter is it was the SSD's under RAID 0 or the two platter drives, the time was exactly the same, I mean exactly.