Gigabyte Claims 1 Million USB 3.0 Mobos Shipped
Quote:Gigabyte is looking to prove that there is money to be made in the vacuum of a slow economy. Like competitor Asus who is still riding the eeePC craze, Gigabyte seems determined to become synonymous with USB 3.0. They have been releasing USB 3.0 powered motherboards for several months already. Their use of NEC's USB-IF certified chipset has almost kept pace with NEC ability to produce them. Currently at 1 million motherboards, for every 3 USB 3.0 chips produced by NEC one of them is headed towards a Gigabyte motherboard.
Now I hope for their sake USB 3.0 becomes as widespread as 2.0
EXT - you said a mouthful! Nowhere is the "new, shiny, higher numbers" phenomenon more apparent (and pernicious) than in DDR3 for the 1156/1366 platform. The memory makers are raking it in for idiocy like 2133, even though:
memory speed doesn't scale in the i5/i7 - no subjectively noticeable real world improvement;
Intel officially support only 800, 1066, and (for some) 1333;
the only worth is in 'synthetic' benchmarks - but few know what the word 'synthetic' means;
and now, the idiocy is 'leaking down' into the 775/AM3 world, where people with 'T' suffix boards are trying to use i7-specific memory!
Some months back I worked on loading drivers and seting up volumes with someone here who had a roughly eight hundred dollar RAID card - and who didn't quite know what 'RAID' was - just knew "it's supposed to be fast"!!
Wow. Yeah, for the ram I was was originally just going to get 1333 ram and leave it there, however, the 1600 stuff was cheaper (silly Newegg). Up to 1600 I see a point (OCing reasons) but when someone posts with 1800+, I'm just not sure what to say (except to run it at low volts/tight timings and way under speed).
Quote:I'm just not sure what to say (except to run it at low volts/tight timings and way under speed).
I have my own solution - I just pretty much don't answer i5/i7 posts! I will look 'em over, and if their problem is that they have overclocked memory, and don't comprehend OCing the platform - well - it's their problem
I don't know if you've seen it before, but I posted a little 'tool' to compare latencies at various clocks, by calculating the equivalent latency at a 'common' clock frequency:
because one of the little bits of misdirection the memory makers use is: do you know which is better - 7CAS @ 1033, or 8CAS @ 1600? Neither do I! And, neither does, pretty much, anyone, without some calculation - and the 'gaming kids' buying this stuff use an on-line checking account, as they don't know enough math to balance a checkbook!!
Someone wrote a good article and made some nice graphs that does a clear job of explaining how memory frequency affects latency. Here is the link:
The thing that most people don't 'get' about the latencies is that they are physical periods of time - determined by the physics of the memory and memory controller chip on the DIMM. One of the 'turnarounds' might be seven nanoseconds - it doesn't care how many 'cycles' you have to wait, as the cycles are counted out in clocks (or 'ticks' as he so aptly says in the article) at whatever frequency; the fact remains that you must 'wait' the whole seven, plus whatever you 'lose' as you must count to the next higher integer, to 'cover' the actual time period! That's why I built the 'tool'; it does the 'integer rounding' and tries to compare 'physical periods' to 'physical periods' - whatever the clocking...
The best article I've ever come across regarding these things is this:
Section two gives the 'hairy' details, concisely and clearly