SSD Speed

[jakesaunders27]

Hi,
Getting a samsung 830 128GB this week, I have a mac pro 1,1 so what kind of speeds will I be able to get as obviously my mac pro only has Sata 2, has anyone else got this setup?

Cheers

 

[paul-n]

SATA 2 allows speeds up to 250 MB/s.

 

[S.B.G]

Here is a video of a speed test between SATA 2 and SATA 3 with a Samsung 830.

 

[Tesselator]

SATA 2 allows speeds up to 250 MB/s.

SATA I is 150 MB/s
SATA II is 300 MB/s
SATA III is 600 MB/s

And in most conditions you can get exactly that too (± <1%) for large file I/O or streaming which calls for it. I read a lot of folks saying that there is some kind of magic overhead so you never get the full speed yet in all my systems I always am able to get the full speeds from them. Me thinks those folks are guessing or have something improperly configured.

That said, most of the SSD goodness happens with small file I/O and that occurs at between 60 MB/s to about 200 MB/s (depending on file size and number of I/O requests). So for most of the goodness which people select SSDs for the MP1,1's SATA II will be fine.

 

[TzunamiOSX]

SATA I is 150 MB/s
SATA II is 300 MB/s
SATA III is 600 MB/s

And in most conditions you can get exactly that too (± <1%) for large file I/O or streaming which calls for it. I read a lot of folks saying that there is some kind of magic overhead so you never get the full speed yet in all my systems I always am able to get the full speeds from them. Me thinks those folks are guessing or have something improperly configured.

That said, most of the SSD goodness happens with small file I/O and that occurs at between 60 MB/s to about 200 MB/s (depending on file size and number of I/O requests). So for most of the goodness which people select SSDs for the MP1,1's SATA II will be fine.

but you have overhead and additional informations to transfer, so 250 MB/s is a realistic rate.

 

[Tesselator]

but you have overhead and additional informations to transfer, so 250 MB/s is a realistic rate.

The additional informations to transfer are from the 8bit/10bit encoding and that's already taken into account when arriving at the 300 MB/s number. Here's the Wiki quote:

Second generation SATA interfaces run with a native transfer rate of 3.0 Gbit/s, and taking 8b/10b encoding into account, the maximum uncoded transfer rate is 2.4 Gbit/s (300 MB/s).

And that's what I get every time I test it... 300 MB/s plus/minus 3 to 5 MB/s. So for me 300 MB/s has always been actual slash realistic. You can do the math yourself, it's pretty easy. It comes out to 300. If you do the tests properly the throughput will always be right there as well 295 to 305 MB/s. And for my SATA III devices 595 to 605 MB/s.

 

[derbothaus]

Never ever seen actual 300MB/s on SATA2 or 600MB/s on SATA3. Ever. Hundreds of Mac's and PC's. If not thousands. Everything is plug and play so there is no 'improperly configured'. Those days are over. There is crappy chipsets but I have used quite a few over the years. Right now I have a Samsung 840 capable of 550MB/s or more and can get 278MB/s out of my X58 Mac Pro. That is fresh state with TRIM. I'd like to see your benches showing this with system config info in sys profiler showing link speed and model etc. I mean I believe you and all
Can you also define 'do the tests properly'? Thanks.

 

[Tesselator]

Never ever seen actual 300MB/s on SATA2 or 600MB/s on SATA3. Ever. Hundreds of Mac's and PC's. If not thousands. Everything is plug and play so there is no 'improperly configured'. Those days are over. There is crappy chipsets but I have used quite a few over the years. Right now I have a Samsung 840 capable of 550MB/s or more and can get 278MB/s out of my X58 Mac Pro. That is fresh state with TRIM. I'd like to see your benches showing this with system config info in sys profiler showing link speed and model etc. I mean I believe you and all
Can you also define 'do the tests properly'? Thanks.

Properly configured means a few things. One and the most important is that you need a device capable of more than the interconnect speed. (SSD RAID0 for SATA III and a single fast SSD to test SATA II). The second thing is to play with the files sizes and drive options (Cache, Asynchronous I/O, etc.), and then finally any of the various benchmarking utilities' settings which may have an affect.

Get it properly configured and set up and you can see that speeds of exactly the interconnects ratings are achievable. It's in error both technically and actually to say SATA II is anything less than 300MB/s or that SATA III is less than 600MB/s.

Now if you wanna start talking about realworld averages and Application I/o that's another story. With that we're probably talking about less than 150 MB/s for either! But to rate what's possible as benchmarking I/O is meant to do that's the story. That you haven't seen that just means you didn't set up the test right is all.

 

[jakesaunders27]

Hi,
thanks for all your replies guys,

if i were to put 2 say.. 60GB ssds in raid 0 would i be able to get 600mb/s with sata 2 as some of you said you can get 300mb/s

cheers
jake

 

[Tesselator]

Hi,
thanks for all your replies guys,

if i were to put 2 say.. 60GB ssds in raid 0 would i be able to get 600mb/s with sata 2 as some of you said you can get 300mb/s

cheers
jake

Yes - assuming you meant MB/s (capitols), lower case means megabits .

But do keep in mind those are benchmark application speeds we're discussing above - you won't get that all the time full-time. Like, for example your OS X installation will probably boot from that at around 100MB/s to 180MB/s average and probably the fastest single read event might be 400MB/s and there wouldn't be too many of those - less than one second's worth for sure(?). With a single SSD that would be more like 60MB/s to 100MB/s with the fastest single read event probably maxing out the interconnect at like 298MB/s or something close to that.

With a fast-ish rotational drive most of your (boot) files would read in at like 6MB/s to 15MB/s with the fastest single read event being probably 195MB/s assuming you had a newer fast drive which could do that.

 

[jakesaunders27]

Ahh okay thanks a lot so would you say its probably not worth having 2 SSDs?

And sorry MB/S not mb/s!!!

 

[derbothaus]

Properly configured means a few things. One and the most important is that you need a device capable of more than the interconnect speed. (SSD RAID0 for SATA III and a single fast SSD to test SATA II). The second thing is to play with the files sizes and drive options (Cache, Asynchronous I/O, etc.), and then finally any of the various benchmarking utilities' settings which may have an affect.

Get it properly configured and set up and you can see that speeds of exactly the interconnects ratings are achievable. It's in error both technically and actually to say SATA II is anything less than 300MB/s or that SATA III is less than 600MB/s.

Now if you wanna start talking about realworld averages and Application I/o that's another story. With that we're probably talking about less than 150 MB/s for either! But to rate what's possible as benchmarking I/O is meant to do that's the story. That you haven't seen that just means you didn't set up the test right is all.

So your proof then? If you have to RAID0 your SATA3 then you are effectively polishing the theoretical overhead mentioned everywhere. With 1200MB/s to play with hitting the advertised speeds are easy (you have 2 effective lanes). I still would like to see your actual results. Not just your words and not just a test window with no settings or link indications. Single drive, single link, hit the theoretical. Otherwise BS. I also know you don't owe anyone anything but c'mon negating the entire web and personal experience in top tier enterprise companies and saying to believe you is a stretch, no?

 

[Tesselator]

So your proof then? If you have to RAID0 your SATA3 then you are effectively polishing the theoretical overhead mentioned everywhere. With 1200MB/s to play with hitting the advertised speeds are easy (you have 2 effective lanes). I still would like to see your actual results. Not just your words and not just a test window with no settings or link indications. Single drive, single link, hit the theoretical. Otherwise BS. I also know you don't owe anyone anything but c'mon negating the entire web and personal experience in top tier enterprise companies and saying to believe you is a stretch, no?

No. One lane. It's called eSATA.

I'm not set up to do a bunch of benchmarks right now but in the past I did this yes. I spent a couple of days messing about with two 6g SSDs in an enclosure connected via eSATA II single cable tested on 6 different machines (one was the MacPro, others were DELL and HP). By messing around with the before mentioned variables I could get the full 300MB/s (plus-or-minus a couple of megs) on some of them. I'm projecting for SATA III - It should be the same deal. I'm not aware of any studies which show I was proving something previously impossible.