[ExtremeTech] Memory and Performance: The Bottom Line - Page 2 - Overclock.net

MadTaco · 08-26-08

Not a BAD article...but I get the feeling that the writer didn't really know what he was talking about in some areas. Most notable area is the Crysis test where a 9 FPS jump was noted as being "almost flat" and using Vista 32-bit for their OS.

Overall, not a good article for truly definitive information. I also agree with the above poster regarding comparing the low-speed DDR3 to high-speed DDR2.

SZayat · 08-27-08

Quote:

Originally Posted by Cheetos316

I think it's a bit too early to call it based on that review. They took the faster end DDR2 and compared it to the slower end of DDR3. Of course there isn't going to be a huge difference in performance. I bet if they took DDR3-2500 and compared it to DDR2-1066, there would be a much bigger difference.

must be uber-affordable

Urufu_Shinjiro · 08-27-08

I agree with the FSB issue as well as why the #$*% did they use a 32 bit OS?! Also it's still early in the DDR3 days, I remember everyone saying there's no point in going to DDR2 because DDR2 800 with 5-6-6 timings was no faster than DDR500 at 2.5-3-3. We all know now that DDR2 has matured and 1066 at 3-4-4 is quite fast, so I look for DDR3 to do the same.

Manyak · 08-27-08

The only time RAM speed really matters much is

a) When its limiting your ability to overclock the FSB
b) When its slow enough to bottleneck the CPU

Faster ram does give you a snappier system, its just not worth the price unless its a big jump in speed.

zaccy001 · 08-27-08

Quote:

Originally Posted by Johnny Utah

I think the FSB completely skews the results. If you're running 667Mhz ram already 1:1 with the processor, would you really see an increase with higher memory dividers? Don't think so. Memory by itself won't hardly make a difference if it's bound by FSB frequency. The ram and cpu have to work hand-in-hand for complete system throughput. You'll see the results of faster ram only if you have a cpu that can keep up and vice versa.

This is exactly the case. Beware noobs.

joemaniaci · 08-27-08

A lot of your guys' comments are the one reason I will miss my 780i, I can unlink my memory and set it independantly from my FSB, now when I get another mobo, I have to relearn how to OC the less than easier way.

Ihatethedukes · 08-27-08

Quote:

Originally Posted by joemaniaci

A lot of your guys' comments are the one reason I will miss my 780i, I can unlink my memory and set it independantly from my FSB, now when I get another mobo, I have to relearn how to OC the less than easier way.

Which can also limit your OC. Unlinking sometimes leads to undesirable divider ratios that don't work as well as others. In other words, OCing 'holes'.

joemaniaci · 08-27-08

Quote:

Originally Posted by Ihatethedukes

Which can also limit your OC. Unlinking sometimes leads to undesirable divider ratios that don't work as well as others. In other words, OCing 'holes'.

A lot of people underestimate the 7-series(at least when they work), Ive had mine to 500FSB and freeroller has had his to 600FSB

Ihatethedukes · 08-27-08

I've had mine to 500FSB as well, but I had to run 'linked' through the 475-490 range. And it wouldn't boot into 490+ but would prime for hours if software OCed from 475-->500 in windows. Really weirdo stuff. My X48 is more temperamental to get to do something, but once it does it, it works like at stock.

		Overclock.net - Overclocking.net > Industry News > Hardware News
[ExtremeTech] Memory and Performance: The Bottom Line

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)