http://www.xbitlabs.com/articles/me...

I finally found what I was looking for to disprove this bogus 1:1 (cpu FSB to RAM speed) BS that is going nuts on the internet. Go to the link above. Notice how even on the worst latencies settings the next speed ram will out perform the next-lowest speed ram on the same freaking CPU setup?

THE 1:1 RATIO IS A BIG WIVES TALE!!!! PLEASE STOP CIRCULATING THIS BOGUS RUMOR!!

Let me explain it very simply, you have DDR2 533 running at it's normal 4-4-4-12 settings, you can pull roughly 5333mb/s
Then you have DDR2 667 running it's normal 5-5-5-15 settings pulling in roughly 5442mb/s

That's 5.44gb a second compared to 5.3gb a second. It's not to hard to figure out that you'll be able to access more information in less amount of time with 667 vs 533 now isnt it?? So why in the world are you people advising others to downgrade their systems???

For your alls info, the testbed was set at 1067FSB(that 266mhzx4) Jam's theory is to set the ram at DDR667(266mhzx2) Boy was he wrong!

So please STOP, please STOP, PLEASE STOP belittling people who know what they are talking about and stop with this bull crap once and for all!!! I'm going to get my dinner now.

Core 2 Duo 1.86
2GB DDR 667
ASUS P5L-MX
Nvidia 8500GT 500/1000

Hope you enjoyed your dinner & didn't choke on your dessert. It seems to me that article supports the 1:1 theory, not your non-1:1 theory.

"it's unreasonable to use memory with a resulting (DDR) frequency two times that of the FSB clock rate"

"At a 400MHz FSB the system with DDR2-800 enjoys a 2 to 7% advantage over DDR2-600 with the same timings, which is a good result. Switching to DDR2-1000 SDRAM doesn't provide big practical benefits: the high bandwidth of that memory type cannot be fully utilized due to the limited bandwidth of the FSB"

"Once again we are shown that increasing the memory frequency only makes sense when is performed within the limits imposed by the FSB bandwidth"

Hmmmmm, interesting concept...keeping the memory freq within the limits imposed by the FSB?? Isn't that just another way of saying 1:1 ratio???

What is that saying again -- Tis’ better to be quiet & thought a fool, than to speak & remove all doubt.

Good night!

OMG! Myth crushed! You sound like a twat. Go away.

Asus P5K SE
E6750@3.3GHz
4096MB A-Data Extreme 800
GF 7900GTX OC'd
XP Pro & Vista x64

Man you are a moron Jam.

Did you even look at any of the benchmarks??? I'm am assuming not. The only bottleneck is the current 8.5gb limit on the 1066 FSB. and the 10.6gb limit on the 1333mhz bus. NOTHING had to do with the 1:1 ratio myth as was seen in the benchmarks. AND as I originally stated it would take DDR2 1067 to match a quad pumped 266mhz bus.

Learn to read.

"More interesting are the results of the overclocked platform. In this case, there is more sense in using fast memory and the optimal memory frequency divisor is 1:1 (FSB:DRAM) as has been shown in our tests. In other words, you can achieve maximum performance by using memory with lowest possible timings in synchronous mode."

Outlander, why are you not understanding this? Yes get faster memory but run it at 1:1 at the fastest possible settings.

I usually feel bad when someone makes an ass out of themselves in public. But when they SCREAM it at you... it's just plain funny.

Slackware Linux Enthusiast

I'm not making an ass of myself, I'm making a point. A point that a few of you dont want to except.

And wemby, Jam was going around telling people to downclock their ram 2:1 cpu to ram ratio and losing considerable bandwidth. IE, telling people to clock a core2 at 1067fsb and DDR2 800 or 667 down to DDR2 533. IE effectively slowing their systems down. Some people can't seem to understand that the pent 4 and core cpu's are quad pumped and DDR2 is only double. Hense the 533 rating at 266mhz.

If one actually looks at those benchmarks, one will see this. The benchmarks keep on climbing until DDR2 1067 is matched with a 1067FSB. Plus I'm tired of Jam telling me I dont know sh*t when I'm trying to help people avoid losing performance in their systems because of one ill informed member.

Did you actually read the article or did you just look at the pretty pictures?

I have NEVER told anyone to run their Intel CPU:DRAM ratio at anything other than 1:1, provided their system allows it to be setup that way. And I always recommend getting DDR2-800 & then downclocking it so that the memory timings can be tightened up, thus reducing the latency.

What you have to realize (& you have been told this several times) is that you should deal with FREQUENCIES, not quad pump or double pump figures. Those numbers are bogus, trumped-up, theoreticals based on FULL bandwidth saturation & 100% efficiency! 100% effiency is NEVER attainable (there's always SOME loss) & full bandwidth saturation rarely happens under "real world" useage. Benchmark data can be manipulated anyway you want, but we deal in the real world & that article states it clearly:

"the results of DDR2-533 and DDR2-1067 differ by only 5-10%, i.e. installing the twice faster memory leads to a negligible performance increase even in games"

Also, a fact you are totally ignoring is that the RAM is being run in dual channel mode. That effectively doubles the RAM throughput. So with the CPU frequency at 266MHz (1066MHZ FSB) & the RAM clocked at 266MHz (DDR2-533) in dual channel mode, the RAM throughput & CPU bandwidth are equal (aka 1:1). I know I've done that math for you before, but here it is again!

CPU bandwidth calculation:

266.6666MHz (freq) x 4 (quad pump) = 1066.6666MHz

1066.6666MHz x 8 (64-bit bus / 8 bits per byte) = 8533.3333MB/sec or 8.53GB/sec

RAM throughput calculation:

266.6666MHz (freq) x 2 (double pump) = 533.3333MHz

533.3333MHz x 16 (128-bit bus** / 8 bits per byte) = 8533.3333MB/sec or 8.53GB/sec

** RAM bus width is 128-bit due to dual channel mode (64-bit x 2)

Thanks for making my point for me again jam.

Dual channel increases performance by 5-10% at the most, so if you are puling 4000mb/s at DDR 533, than you will have about 4,400mb/s, not 8,000mb/s I dont know where in the world you are coming up with this 100% increase from.

We are talking "real" world Jam , look at the benchmarks Jam, it took them till PC-1067 to get to around 8000mb/s (sandra reported about 6500)

"Real" world performance Jam, "real" world

http://www.xbitlabs.com/images/memo...

So if you want to lose FPS in any game, follow Jams instructions

It may be close to 30 yrs since I'v been in college, but as far as I know, math hasn't changed over the years. The calculations I did above are sound.

You will continue to believe what you believe, regardless of the facts that are staring you right in the face. The article you posted backs my claims, not your's & there are several others involved in this thread who agree.

And every single time I find you pushing your 1:2 CPU:DRAM ratio agenda, I will do my best to convince people otherwise. This thread & the article you posted in the OP should be a huge help. Thanks.

BTW, where's jboy when you need him? ;-)

Outlander
Look at your example
Let me explain it very simply, you have DDR2 533 running at it's normal 4-4-4-12 settings, you can pull roughly 5333mb/s
Then you have DDR2 667 running it's normal 5-5-5-15 settings pulling in roughly 5442mb/s
If you tighten the setting on the ddr533 you get the same throughput as ddr 667.
Now to buy 1067 ram for a 5% increase is not worth the money. You are not doubling your throughput not even close, now if you are overclocking you will of course want to buy the bext grade of ram but set it to a 1:1 ratio and tighten those settings as far as possible for the best performance as your link clearly shows. The gains in the graphs clearly show that the most gain is from the ddr 533 being tightend up and as you use faster ram the gains are less and less although the latency goes up and this I believe is where you are getting confused.

Outlander, I think you'd be great to join Apple's advertising team. I've noticed that you sound a lot like those Apple ads in my old magazines! They loved using phrases such as "boy were they wrong." They also loved using Photoshop speed tests to assert outrageous claims (e.g. that their G3/G4/G5s were "supercomputers").

Thanks to eBay, I could afford to SLi my PC and eat artistic toast.
Opteron 185 @ 3.0GHz
2X 8800GTS-640MB
Asus A8N32-SLI deluxe
4GB of PC3200 OCZ
800GB local storage
X-Fi, Vista X64

"Did you actually read the article or did you just look at the pretty pictures?"

Hey be nice jam .... that is what I do when watching FAUX news, especially the orly segment.

Now back to CN, I have often wondered if the word outlandish has any remote connection to our friend's username of choice or if it's just pure coincidence.

I must agree though that there is certainly a propensity for bending the truth with this OP & even dissemination of outright fallacies. While this does not happen 100% of the time, it does happen around 99.5% of the time.

Hehe not been here for awhile,thanks for the chuckle outlander.

Hi jam,saber and all

Tt Lanfire
MSI K8N Diamond
X2 4600+@225x12=2.7ghz 1:1
1GB PC3200/4400
OCZ Powerstream 520w
7900GTO
WDCaviar 160gb sata x2

Fox News rules sabertooth and don't you forget it! Yeeeeeeehawwwww. :)

I've found this site after googling for "ram myth" or something. Actually, I was looking for benchmarks comparing PC6400 to PC5300 and 5-5-5-15 to 4-4-4-12.

I wanted to say that I've found this site hilarous. Seems like everybody on this forum is trolling the f*ck out of Outlander lol. Even that jam guy, who manages to say that 533 is faster than 667 with a straight face when there's a link on the same page proving exactly the opposite.

Well, anyway, that was very funny. Everybody please run your RAM @ 533 and if you find your PC is too slow, buy a better CPU. Hell, clock your CPU to 533Mhz too, I heard it was faster.

haha Outland changed his name to skidelydoo.

In my case, I have a C2D e4500 (runs at 2.2Ghz standard, thats 200Mhz x 11 multiplier) and DDR2 800 (with 4-4-4-12)

so what I did, (since I'm a total noob at this, and just kinda guessed but I thought 1:1 was good)

I set my multiplier to 6x (low as it goes)
and put my bus speed up from 200Mhz to 400Mhz, getting and FSB of 1600Mhz and a cpu running at 2.4Ghz (a small overclock) and i didn't touch my RAM. But I have a 1:1 ratio, right?

So what I want to know is, is this an ideal setting? Or am I retarded?

Is an FSB at 1600Mhz ridiculous or am I happy that I could achieve that?

I would also like to OC closer to 2.6-2.8 (not to much since this PC is for my GF's mom, lol) the best way to get 2.7Ghz out of my CPU would be to just OC the ram as well? to something near DDR2 900Mhz?

Again I'm real new to this, and this was that first thing that came up on google. sorry if I don't know what the deuce I'm doing. just tell me before I break something.

No, I am me, there's just no point in arguing with morons that don't understand basic arithmetic thats all, and also getting into trouble at work. Downgrade your system if you like, it's not hurting my performance any.

And I think your on the wrong thread btw.

Core 2 Duo 1.86
2GB DDR 667
ASUS P5L-MX
Nvidia 8500GT 500/1000

Outlander, I think you'd be great to join Apple's advertising team. I've noticed that you sound a lot like those Apple ads in my old magazines! They loved using phrases such as "boy were they wrong." They also loved using Photoshop speed tests to assert outrageous claims (e.g. that their G3/G4/G5s were "supercomputers").

Well I use to be a mac tech from 94-2002 :) But I stopped being one when that crap OS-X came out and stopped using them when x86 macs came out. A PC in sheep's clothing.

And the G4 and G5 ran circles around the PC's of the time. P3's, and P4's at the same speed were no match, especially the P4. Then when using altivec enabled apps was added to the mix no PC stood a chance! But, times have changed and steve freaking jobs has killed the macintosh. Nothing more than a damn expensive PC! God I miss the system 7 days! Ahhh.. system 7. How I miss thee.

I especially miss the 604e with it's multiple FPU units, it took a 333mhz G3(750) to match a 233 604e in terms of math performance. I still run my 604e machV 9600 for 3D art with Strata, 400mhz machV 604e!

Core 2 Duo 1.86
2GB DDR 667
ASUS P5L-MX
Nvidia 8500GT 500/1000