I have a hammer! I can put things together! I can knock things apart! I can alter my environment at will and make an incredible din all the while! -Calvin
In his case, the x58 northbridge chip is responsible for 32 pci-e lanes (which are going to be routed to the graphics card slots) and the 6 lanes being handled by the southbridge. The southbridge connects to the northbridge by DMI bus and the northbridge connects to the processor by quick path bus. The quick path connection allows for 25.6 gigabytes per second. The 32 lanes of pci-e on the northbridge can use a max of 18 gigabytes per second. The DMI connection to the southbridge is 2 gigabytes per second.
The graphics slots should get their full bandwidth. The slots on the southbridge total 3 gigabytes per second and the connection to the northbridge is only 2 gigabytes per second, so they could technically fight for bandwidth.
I have noticed that PCI-express-based solid-state drives often have their flash memory cells in a RAID 0 configuration, and I wonder why that is. I am not an expert on RAID, but I do know that, while RAID 0 does provide an increase in data transfer rates over a non-RAID configuration, it also has no redundancy and thus is very vulnerable to irreversible data loss, and a flash memory device with a PCI-e interface is already so fast that I do not believe that it would need to be in a RAID configuration. What insight can anyone else here offer on this subject?
"When the people fear the government, there is tyranny; when the government fears the people, there is liberty." -Thomas Jefferson.
"Those who would trade their freedoms for security will have neither." -Benjamin Franklin
Cheaper to use slower cheaper flash memory in raid costs more to make faster flash non raid.
I am not sure about all of the SSD and PCIe devices out there, but almost all of the high end PCIe cards and SSD's have extra cells and in some cases entire duplication of memory chips for backup purposes. I am not sure of all the technology, but a cell failure will result in no loss of data. When I was at an Intel training class they briefly touched on the technology in the SSD's they sell and this subject came up. They were not too technical on it, but knew the basics. The same thing came up in our Texas Memory class, but it was mostly for the sales guys, so They couldn't answer (or wouldn't) the technical questions.
I do know that the Intel technology is considered a hybrid raid because of the power saving type of feature where the "spare" cell is not in use until a failure, or anticipated failure, so while it is technically part of the raid array, it may never be used. I will see if I can dig up the white paper for it.
"...Dumb all over, A little ugly on the side... "...Frank Zappa...
Both of those explanations make sense, even if they are completely different from each other. Perhaps as solid-state drives become less expensive, they shall use memory cells arranged in a traditional form?
Also, I have another question: why do most PCI-express-based solid-state drives have storage capacities in multiples of 240 gigabytes (240 GB, 480 GB, 960 GB), rather than the binary progression (8, 16, 32, 64, 128, 256, and 512 GB) that SATA-based SSD's and USB flash drives use? Those storage capacities are nearly universal, so it seems that the PCI-express-based SSD's are deviating from the normal trend with their capacities.
As for the main subject of this thread, I recently heard news about both ASUS and OCZ manufacturing new PCI-express-based solid-state drives, so I shall presume that such devices may indeed gain greater popularity in the near future.
"When the people fear the government, there is tyranny; when the government fears the people, there is liberty." -Thomas Jefferson.
"Those who would trade their freedoms for security will have neither." -Benjamin Franklin
Remember that 1GB isn't really 1,000MB... it's 1,024MB. And a 240GB drive isn't 240,000MB...
I have a hammer! I can put things together! I can knock things apart! I can alter my environment at will and make an incredible din all the while! -Calvin
I believe that probably reserving flash cells for wear leveling that are not visible to the system. The 240GB drive probably reserves 16 gigs and then maps cells marked bad in the primary storage to the reserve chips. Likewise the 480 reserves 32 and the 960 reserves 64.
To revive this thread after a long time of inactivity, ASUS has now released their RAIDR solid-state drive, which uses the PCI-express interface, and while they certainly hyped and promoted it greatly, the reviews of it that I have read have been mixed. One review stated that the device did not reach its advertised transfer rates in every test, and that a major reason for that is that the controller of the drive does not natively support the PCI-express format, so it instead must use an older SATA controller to handle data transfers.
Therefore, it seems to me that that greatest barrier to widespread adoption of PCI-express-based solid-state drives is a lack of controllers for them, so, if manufacturers begin to develop SSD controlelrs that can properly utilize the PCI-express format, will such drives become more prevalent? I myself am hoping that Samsung develops a PCI-express-based SSD, since their current SATA-based SSD's are so very awesome, and they develop every part of the drive themselves, meaning that they can test them for the best functionality. What does everyone else say about that?
"When the people fear the government, there is tyranny; when the government fears the people, there is liberty." -Thomas Jefferson.
"Those who would trade their freedoms for security will have neither." -Benjamin Franklin