Paul wrote:
Don wrote:

WOW.

Thanks Paul. I will read your note carefully and try it all out.

Thank you very much. I appreciate your advise.

Don

One thing I missed, is your processor has a default
expectation for memory operating speed.

http://www.cpu-world.com/CPUs/K10/AM...0OCGQBOX).html


It says "DDR2-1066 and DDR3-1066" at the bottom. Or
perhaps DDR3-1333 with two sticks in dual channel, and
DDR3-1066 when four sticks are installed in dual channel.

*******

ftp://download.gigabyte.ru/manual/mb...gmt-usb3_e.pdf

"Memory Clock

This option is configurable only when Set Memory Clock is set to Manual.

X4.00 Sets Memory Clock to X4.00.
X5.33 Sets Memory Clock to X5.33.
X6.66 Sets Memory Clock to X6.66.
X8.00 Sets Memory Clock to X8.00. "

It'll probably default to X6.66 for DDR3-1333 on the first
startup, and then you can set it to X8.00 to get DDR3-1600.

The board promises DDR3-1800 as the maximum overclock, and since they
don't show a divider suitable to make 1800 (it would take "X9.00"),
you'd need to increase the CPU input clock a bit, over its nominal
200MHz value. Then, you could use the X8.00 setting plus a little
CPU overclock, to get to DDR3-1800. But since your memory is
DDR3-1600, you don't have to worry about this part.

The other thing to consider, when the manufacturer makes certain claims,
is to download the memory compatibility chart, and see whether they're
being honest or not. Since they support a variety of single sided (SS)
and double sided (DS) modules at DDR3-1600, it looks like you're safe.

http://www.gigabyte.com.tw/FileList/...85gmt-usb3.pdf


The reviews here says...

http://www.newegg.com/Product/Produc...82E16813128432

"Remember to change the memory clock multiplier in BIOS if you are
running DDR3 1600 as default is 1333 (this is noted on the Gigabyte
site as well)."

Another thing I was just looking for, is whether the USB3 is half-speed
or full-speed. If the USB3 chip is connected to a PCI Express Rev2.0 x1
lane, it runs full speed. If connected to a PCI Express Rev1 x1,
it would run at half speed (due to the 250MB/sec limit of a Rev1 lane).
The documentation remains quiet on the subject, choosing not to name
it. In the picture here, there is an asterisk next to "6x1 PCI Express
2.0",
but without explaining under what circumstances it isn't Rev 2.0.
Rev 2.0 requires a low jitter clock, which may not be available
"for free" on the chipset. That was a limitation of some of the Intel
boards with that USB3 chip. It might take a while, before truly
worthy USB3 stuff shows up, to use as a test.

http://www.madboxpc.com/wp-content/u...ipset-diag.png

Have fun,
Paul

*****

As you know, I put the Sapphire HD 5670 in my new GA-785GMT-USB3 (see
[1] below). I experienced periodic black screens of death, and vertical
gray line screens of death. This occured with Win7 32 and 64 bit
versions. The memory checked out fine with memtest86, and all of Paul's
suggestions were great. I have the latest bios (F3a). The on-board video
runs fine, and has never crashed the system. The on-board video used the
same ATI 10.4 drivers as this card (now 10.5). It is the Sapphire card
that is the problem. It is defective. Without it, the system runs fine
without the lockups and crashes.

I RMA'ed the Sapphire HD 5670, but with all of the fooling around to
diagnose the problem I may have missed Newegg's return date (30 days
from the date of the ORDER for them to RECEIVE the product back --
thanks a lot). Plus $5.99 in shipping to me, plus $10.22 to ship it
back, plus a $13.50 "restocking" fee (to do what? throw it away and make
a claim against the manufacturer?). Thanks again. Well over $100.00 +
everyone's time wasted. I will see what happens. Proceed accordingly.

Thanks.

Don

[1] SAPPHIRE 100287VGAL Radeon HD 5670 (Redwood) 512MB 128-bit DDR5 PCI
Express 2.0 x16 HDCP Ready CrossFireX Support Video Card.