Is sata 2 actually a bottleneck?

Soldato
Joined
22 Dec 2008
Posts
10,369
Location
England
I know that a lot of the excitement of sata 3 is that it's possible to saturate sata 2 with a few well chosen ssds. I think it tops out around 300 mb/s or so.

What I'm wondering is if this is actually a limiting factor in a pc. Say 4ghz i7, 12gb ram, one vertex. Would adding a second vertex actually make any difference? I'm not convinced it would, but I'm open to persuasion either way. Since I'm struggling to see raid 0 of vertex drives helping, I really struggle to see a saturated sata 2 link holding a system back

It limits the system under hard drive benchmarks, but does this translate into any other usage?
 
SATA II has a maximum data rate of 300MB/s

SSDs are getting awfully close to this. Remember even if you RAID SSDs, each one has it's own SATA connection so you double the maximum rate to 600MB/s.

Its only when a single drive actually surpasses the 300MB/s mark will it be a bottleneck ;)
 
Even though SSD's peak read rate is approaching that of Sata II, its average throughput when dealing with random reads and random writes and general use is much lower.

Plus as mentioned already Sata's performance is per port, so raiding works well.

Its really IO performance that counts, thats why even with the Sata II bandwidth I am looking forward to seeing the internally raided OCZ Collussus drives, they should have some quite excellent 4k random write performance.
 
Ah, I feared this response. Thank you, but that isn't the answer to the question I am asking. I shall try rewording.

Given an sata 2 vertex ssd, running on a standard overclocked desktop pc. And given two sata 2 vertex ssds, running in raid 0 on an identical pc. Is there any difference in usage? Not in hard drive specific benchmarks, but in any other circumstance? Do levels still load quicker, windows still boot quicker? Is it in any sense more responsive?

I don't think moving data to and from disk is a limiting factor once you're already hitting around 200mb/s, which leads to the conclusion that sata 3 is a gimmick.

However there are people on here with ssd raids, who hopefully used the comptuer with just the one drive for a while first and will be able to comment. I don't think sata 3 is worth paying any attention to, but if sata 2 ssd raid makes a difference then it may be. Do you follow?

I know we haven't hit the limit of sata 2 yet. I also know four vertex in raid 0 look bottlenecked by the southbridge. What I want to know is if this matters or not, since having a storage array that is far quicker than the rest of the computer just doesn't matter
 
yes, the harddrives themselves, each port has its own limit, the chipset has a limit aswell but 2 decent ssd's aren't breaking that limit on the newest chipsets.

So 2 vertex's wouldn't be limited, but they also wouldn't be altogether that much faster than a single vertex, check some game loading bencmarks, a Intel G1, a Vertex, a Crucial and Raptor all gave load times in most games within a second or two of each other.

THe problem is loading a level ISN'T about the data load only, textures are being read then decompressed, stored and moved, ai is being loaded then essentially put into motion. THe data might load faster but if its still taking the cpu as long as it needs to do its thing, you're limited elsewhere.

Sata 3 really won't bring instance performance benefits, though to a degree as memory gets smaller and cheaper and produced in larger quantities, well, consider a 500gb ssd with a new controller with 5 times as many channels so it can read and write hugely faster.

But the main thing here is, a southbridge with 4 sata 2 ports will "aim" to be able to cope with 1200mb/s throughput, theoretically they can, in reality, they can't. There was no need for a chipset that can do that much in reality because 99.99999% of users won't ever max out more than a single drive at a time so really wouldn't be pushing more than 100mb/s sustained for any length of time. Its just a waste of money, silicone and time making a better chipset.

However a chipset designed for the future needs of using 6 sata 3 ports. We won't be maxing them either, but even a conservative outlook in terms of chipset design, they'll be upping the realistic limit of the chipset to cope with the possibility of newer/faster drives.


No, we don't need sata 3, yes we do want chipsets with more capacity, the way it works is it all comes in together in one neat package. The fact is, sata 3 won't make anything more expensive than if it was branded sata 2.

Sata 2 mobo's didn't carry a price premium, same way as USB port numbers increased we didn't get charged extra, its just a very basic thing.

THen theres the other angle, while a game isn't loading faster TODAY, in two years it could be making a significant difference, maybe in 6 months, who knows. Goalposts move and theres no downside to sata 3 being introduced, sata 2 gave us nothing either in reality, 150mb/s really wouldn't limit "most" of what we do. A Vertex can read faster than sata 1, but 98% of your real usage, loading games, surfing, loading windows, is done in the random read/writes at the lower end of the scale that frankly, Sata 1 wouldn't make a difference to either.
 
Sound. That basically agrees with my thoughts on the matter, and shows better arguments for getting to the useful conclusion that I shouldn't get another vertex, or move to sata 3 when it comes out.

Apologies for using gaming as an example, it's tricky to think of a standard usage which primarily stresses the hard drive. I suspect nothing I do does so, which would rather answer my question for me I suppose.

Thank you, it's always nice to see a response which makes such sense :)
 
Back
Top Bottom