Thread: SDRAM choice
View Single Post
  #4  
Old September 19th 14, 05:10 PM posted to alt.comp.periphs.mainboard.gigabyte
Paul
external usenet poster
 
Posts: 13,364
Default SDRAM choice

Haines Brown wrote:
Paul writes:

Haines Brown wrote:
In the past I've had no problem picking RAM according to motherboard
specifications and had no trouble, but this time the more I
investigate, the less comfortable I become.

http://www.gskill.com/en/configurator


Thank you for your very informative post. It didn't occur to me to visit
Skill to see what they might recommend for my motherboard. I see they
offers a wide selection. My intuition is that the RipjawsZ is intended
for overclocking. The Sniper SDRAM is much shorter and it is not clear
how they can recommend it for the motherboard. Am I correct to assume
that the DDR3-1600s are faster than the DDR3-1333s? I gather that the
timing 9-9-9-24 and voltage will probably have to be defined in
Gigabyte's BIOS.


The dimensions in terms of the length of the modules, they
should all be the same length and pin count.

The height is tricky. Chips are available as TSSOP or fine pitch
BGA, and the BGA ones make short modules possible. Some of the Kingston
I've bought in the past, were "low profile" and didn't have heatsinks.
Those hurt your fingers to insert, as the bevel on the connector is
too blunt and means extra insertion force. But, the low profile
modules are not very tall, and will fit under more
different heatsinks.

Modules are made taller by various heatsinks. Sometimes the heatsinks
are so tall, they conflict with the CPU heatsink. On my latest build,
I used a cooler with "extreme headroom" and I can remove the DIMMs
while the heatsink stays in place. The RAM I bought, the "crown" of the
DIMM can be unscrewed, and it still has a lower section of cooler
over the RAM. But the CPU heatsink was tall enough, to clear the modules
fully assembled.

One mistake I made in purchasing, is I didn't notice the DIMM
slots were "one-latch" variety. Good slots have ejectors on
both ends of the socket. My sockets only have an ejector on
the end of the socket furthest away from the slot 1 video card.
And that makes the RAM hard to remove. At one time, the slots
were color coded (end of slot same color as body of slot), to point
out there is no ejector. But this board used camouflage, painting
on fake colors to make it look like a latch was present. And
that caught me and I didn't notice. I can live with it, but
I'm not happy about it.

DIMMs nave transfer speed, and they have latency. The 1600 is
faster than the 1300, as there are more "megabytes/sec". If you
were doing block memory transfers, the speed might help.

The latency, is proportional to the inverse of the clock period
(a measure of the 1600), as well as the CAS number (9). The
1600 DIMM has tinier-sized CAS staps. So CAS9 at 1600 is a
shorter period of time than CAS9 at 1333.

In some articles I've read, the authors of the articles
tried to downplay CAS and look at speed instead. They
presented benchmark data, in a 2D table of latency and speed,
and the speed buys you a bit more. Their recommendation is not
to spend an infinite amount of money trying to get low-latency
RAM. It might actually be hard to find low-latency chips, so
you couldn't buy a really low value anyway. And the processor
itself has a limit as to how low the CAS setting can go. I've
had one hardware setup, where hardware other than the DIMM,
prevented the lowest possible CAS to be used.


If I go to ark.intel.com and select a 4790K processor, it lists

Max Memory Size (dependent on memory type) 32 GB
Memory Types DDR3-1333/1600

So the motherboard appears to stick with the Intel recommendation,
rather than going crazy with overclocks.


Your board has an extensive CPU support chart, and I don't
see any immediate danger signs there.

http://www.gigabyte.com/support-down....aspx?pid=4962

Notice the processor at the top of the chart requires BIOS
version B4.


Yes, the 4790K is my CPU, which requires BIOS version F4. Is the
implication that the Intel H97 - GA-H97-D3H (rev. 1.0) has it, or will I
have to start by flashing my BIOS? :-(


Once you're aware of the issue, you try to buy from a vendor with
lots of hardware turnover.

My LGA775 board, the vendor still had stock from the first
container from China, a year and a half after introduction.
That would be an example of a "stale" source of motherboards.
If there were B3 and B4 boards, I could be assured that source
would sell me a B3 and I'd be stuck. I'd need to borrow a CPU
from somewhere, to flash it up.

My new Asus board, has one USB connector with a white ring
around it, and that USB port is the "flasher" port. You can connect
a USB key with a BIOS file on it, and the motherboard can flash
itself without a processor. It means there is a standalone chip
of some sort, which works the magic of flashing the serial EEPROM
for you. And that avoids the problem of "what do I do if I get a
B3 board". So Asus has a solution to that. I hope the chip is
available so others can add the capability to theirs. As it
eliminates the problem.

snip

Paul