HDD, SSHD, SSD comparison http://www.ign.com/wikis/playstation-4/PlayStation_4_Hard_Drive_Speed_Test_Comparison
PlayStation 4 (codename Orbis) technical hardware investigation (news and rumours)
- Thread starter Love_In_Rio
- Start date
- Status
Nice find, product brief here ...
it is SATA 2.6 (300 MB/s) <-> USB 3.0 (500 MB/s) bridge, so SSD with speeds around 550MB/s are limited to SATA 2.6 speed (300MB/s) at best ... I don't see point in buying SSD for PS4 and I'm glad that I bought cheap 1TB HDD (WD Blue) as stock HDD replacement.
And as I looking at IGN comparsion posted by scf, there is something else limiting speed of SSD in PS4, because in theory HDD read speed is around 150MB/s and if SSD is limited to SATA 2.6 speed, which is still twice as fast, I don't see correlation with IGN comparsion where SSD is only slightly faster than stock HDD and definitely not worth the cost of SSD.
Last edited by a moderator:
Most of the speedup of an SSD usually comes from the much lower latency, not the roughly doubled throughput.
At least we can see that the installation is limited by the bluray drive, and loading games on the stock harddrive is acceptable.
But ssd is clearly faster and as games grow along with assets the difference will multiply.
As everyone knows, sequential speed is only one factor of ssd drives, the real speed gain is from a seek time of almost zero, while the mechanical drive is slow slow.
yes of course if you read lot of small files SSD will be better and difference will be bigger, but when you load level in say BF4, which is probably optimised for streaming from standard HDD/BD (PlayGo system) you read one big file and we don't see any difference (if IGN comparison is not made by complete morons who didn't wait until game is installed)
Last edited by a moderator:
And since game installation and level loading is mostly streaming large blocks of data with narrow locality, access latency advantage of flash-memory is largely mitigated in this case.
An exception here would be for titles extensively using virtual texturing.
That pretty much sums it up.
... but doesn't answer question why advantage of sequential read of SSD in PS4, even limited to SATA 2.6 speed, doesn't show in IGN comparison, will be interesting to hear from developers what is limiting factor.
Last edited by a moderator:
Nope there is no such thing as too fast non-volatile storage as they are all a hell of a lot slower than the RAM they feed. Now someone else mentioned the risk of race conditions where the developer has assumed X seconds to load and timed certain systems around that but that seems like a really risky way to handle synchronisation (especially over time when disk fragmentation may slow loading). Still just because its a bad idea doesn't mean it isn't done but given that we haven't seen any reports in the wild of it being an issue this gen I'd be surprised if it showed up this go around(even with SATA 1 limitations). Tempted by perhaps a 256GB SSD myself but I'll wait for some better comparisons before I jump (the current ones don't seem that systematic).
When you only have a certain amount of bandwidth to give to devices why give more then you need to storage?
In Gamespot video, they compared Resogun boot speed from HDD and SSD, 19 vs. 9 seconds is more in accordance with assumption given by estimated transfer speeds 150MB/s vs. 300MB/s.
http://fc07.deviantart.net/fs71/f/2013/317/2/b/ps4_ssd_load_times_by_alo81-d6u4bke.gif
edit:
IGN boot time is 11.55 seconds for HDD and 9.75 seconds for SSD, why is there difference between Gamespot and IGN, someone is doing something wrong
http://fc07.deviantart.net/fs71/f/2013/317/2/b/ps4_ssd_load_times_by_alo81-d6u4bke.gif
edit:
IGN boot time is 11.55 seconds for HDD and 9.75 seconds for SSD, why is there difference between Gamespot and IGN, someone is doing something wrong
Last edited by a moderator:
It may be that the disk is encrypted and/or compressed.
It's a good point but what else has a significant bandwidth need? The camera is on it's own separate USB 3.0 bus to prevent the contention you describe and unless they'e daft enough to have all the I/O chips share a single PCIe lane it's a non-issue.
What are the chances of these speeds being improved with further firmware updates? I assume if its based on FreeBSD, most of the I/O stuff is already well optimised, perhaps? But as I say that, there strange sata on the USB thing probably changes the dynamic.
Whatever the further changes, the current firmware seems pretty rushed and rough around the edges, which is expected of course.
Whatever the further changes, the current firmware seems pretty rushed and rough around the edges, which is expected of course.
- Status
