I know this won't be out for atleast a year but just wondering the speed increase.
25nm vs 20nm in transfer speeds
200 MT/s vs 333 MT/s
8KB pages vs 16KB pages
how does that translate to possible increase in speeds??
I know this won't be out for atleast a year but just wondering the speed increase.
25nm vs 20nm in transfer speeds
200 MT/s vs 333 MT/s
8KB pages vs 16KB pages
how does that translate to possible increase in speeds??
i7 6700K @4.8 ghz
XSPC RayStorm (very nice block)
Z170 Sabertooh ,, 32GB- Gskill (15-15-15-36 @3600 mhz) 1:1
XFX-7970 with Swiftech Komodo nickel block
Water Cooling - MO-RA3 Pro with 4 Silverstone 180mm @ 700 rpm, Twin Vario mcp-655 pumps
Samsung 850-1TB SSD,, OCZ ZX-1250W (powerfull and silent)
Crossfire 30" decent monitor for IPS too bad SED tech died
Docsis2.0 Docsis3.0
-- People who reject the idea that "government has a responsibility to reduce income inequality" give an average of four times more than people who accept that proposition.
tbh the controller involved will equate to faster overall speeds more directly than a die shrink.
"Lurking" Since 1977
Jesus Saves, God Backs-Up *I come to the news section to ban people, not read complaints.*-[XC]GomelerDon't believe Squish, his hardware does control him!
if i'm thinking of the correct thing, its not the shrink that makes things go faster, its the NAND interface upgrade(to ONFI 3.0).
EDIT: but to answer your question more directly...we've pretty much maxed out SATA 3 already. the speed improvements will be for random 4k speeds - which are the only speeds you should care about.
Controller improvements = faster and more sustainable bandwidth. Can’t really see 4K at QD1 improving with die shrinks however. (EDIT: At least not with a SATA interface). If anything 4K @QD1 will slow down. Maybe that will be offset by the large DRAM caches that are starting to appear. Some of the caches being used are huge (512MB) and they are being deployed without power caps.
Last edited by Ao1; 01-18-2012 at 06:20 AM.
I wasn't being clear my bad
ONFI 2.0 = 200 MT/s 8KB pages
ONFI 3.0 = 333 MT/s 16KB pages
I understand the controller or sata is the limiting factor just wondering how much gain between the two NAND interfaces?
found this info
THE OPEN NAND Flash Interface (ONFI) Working Group has published the ONFI 3.0 standard which doubles the current NAND interface transfer rates to 400MB/sec
does that sound right ?? the Page size doubled and the transfer rate went up 66%
i7 6700K @4.8 ghz
XSPC RayStorm (very nice block)
Z170 Sabertooh ,, 32GB- Gskill (15-15-15-36 @3600 mhz) 1:1
XFX-7970 with Swiftech Komodo nickel block
Water Cooling - MO-RA3 Pro with 4 Silverstone 180mm @ 700 rpm, Twin Vario mcp-655 pumps
Samsung 850-1TB SSD,, OCZ ZX-1250W (powerfull and silent)
Crossfire 30" decent monitor for IPS too bad SED tech died
Docsis2.0 Docsis3.0
-- People who reject the idea that "government has a responsibility to reduce income inequality" give an average of four times more than people who accept that proposition.
The page size has little to do with the sustained transfer rate.
The bandwidth of an SSD (to its on-board flash) is determined by:
1) The speed of the flash, i.e., 333 MegaTransfers / sec, 200 MT/s, etc.
2) The bus-width of the flash chips, usually 8 bits so MT/s = MB/s
3) The amount of parallelism in the SSD, which has to do with the number of flash channels supported by the controller, as well as the amount of interleaving supported by the controller and the flash
The page size mostly affects only latency. Generally a larger page size will result in higher latency.
Last edited by johnw; 01-18-2012 at 08:56 AM.
Shouldn't increased pagesize results in lower performance then a smaller one? Of course this will be off-set by the controllers, but theoretically going from 8 KB to 16 KB sounds like a step backwards to me especially with smaller files.
SweClockers.com
CPU: Phenom II X4 955BE
Clock: 4200MHz 1.4375v
Memory: Dominator GT 2x2GB 1600MHz 6-6-6-20 1.65v
Motherboard: ASUS Crosshair IV Formula
GPU: HD 5770
A page is the minimum read size that can be read. If the page size is 4K (34nm) and the read operation is 4K the 4K page is read. If the page size is 8K (25nm) and the read operation is 4K you need to read the 8K page to get the 4K data you need.
For writes an erase operation is based on a block, so smaller geometries have a much larger impact on write operations. This is of course off set by improvements in the controllers.
EDIT: Conversely if you are reading a 16K file you need to read 4 pages with 32nm. With 25nm you only need to read 2, hence sequential performance is improved with smaller geometries, but small random xfer performance is decreased. That’s my understanding anyway.
Last edited by Ao1; 01-18-2012 at 10:23 AM.
So basically ONFI 2.0 vs 3.0 will be a 66% rate increase from 200 MT/s and a 100% rate increase from 166 MT/s depending on which manufacture is being compared to.
Sounds like a nice improvement. 2013 should be an interesting year for SSDs
Pair that increase with thunderbolt / light peak interface to the mobo would be smoking fast.
i7 6700K @4.8 ghz
XSPC RayStorm (very nice block)
Z170 Sabertooh ,, 32GB- Gskill (15-15-15-36 @3600 mhz) 1:1
XFX-7970 with Swiftech Komodo nickel block
Water Cooling - MO-RA3 Pro with 4 Silverstone 180mm @ 700 rpm, Twin Vario mcp-655 pumps
Samsung 850-1TB SSD,, OCZ ZX-1250W (powerfull and silent)
Crossfire 30" decent monitor for IPS too bad SED tech died
Docsis2.0 Docsis3.0
-- People who reject the idea that "government has a responsibility to reduce income inequality" give an average of four times more than people who accept that proposition.
Bookmarks