Benchmarked; DDR2 533 VS DDR2 677

I received my new ram! Since I have SiSoftware Sandra at work, I had to see the difference between CAS 5 and CAS 4 RAM.

Tested; HYNIX DDR2-533-PC2-4200 from Dell and OCZ DDR2-667-PC2-5400 from Newegg; price plus shipping $167.35.

The DDR2 667 value ram is slightly SLOWER than the DDR2-533

Synopsis; not worth it! IMHO wait for 5400 CAS 4 or if you need it, go for the 533 CAS 4 PC2-4200.

I also had high hopes with Oblivion. The disk activity is slightly better, but again, not worth it. Certainly no better frame rates if that was expected. Photoshop does a bit better when more photos are open.

Attached are the tests, all text, in WordPad.doc format if you want to pick it apart. The test show latency then bandwidth test with the 667 first followed by the 533 ram, all performed in my Dell-1705. All chip timings are shown.


View attachment Latency and Bandwith.zip

 

Excellent, thanks for taking the time/effort to do that. The DDR2-667 modules CAS latency of 5 really is high; can't say I'm surprised the 533MHz does better. Either way, I can't see how the difference would be noticable.

 

thanks for testing that! awesome research and documentation

 

Hey, I just found a huge benefit due to my $160 ram upgrade

Oblivion no longer crashes when quitting! When I quit I instantly get my desktop back without all the disk activity and redrawing of the desktop from before. And yes all patches have been applied.

 

thats why when you add 667 to a 533 system and pc mark youll get a lower score even with 4 times the ram due to the higher cas latency.

side note: my ddr pc333 in my old averatec benchmarks twice as good as my ddr2 533.

 

Yep, AFAIK this is a lot of marketing smoke and mirrors. A lot of gamers won’t touch anything DDR2 until the manufactures gets the hardware right. I read AMD is going to the DDR2 bandwagon soon.

I want my CAS3 DDR2

 

However, with 667, you do gain a decent amount of bandwidth that 533 can't match. In reality, in every day use, gaming, etc, it's probably a wash. If you had DDR2-533 with CAS2, then you would probably notice...

If Dell gave us BIOS control, then we could mess with timings and clocks ourselves...

Where you been livin'? ;-) AMD's AM2 platform is 100% DDR2 and even without really low timings, it is marginally quicker than S939. What I really want is CAS2 DDR2-1067! Oh well, in the end it all comes down to properly aligning the bandwidth/freqs between your CPU and RAM.

 

You might want to check out my bandwidth benchmarks Have you posted yours?

This has been released, when? "Where you been livin'? ;-) AMD's AM2 platform is 100% DDR2 and even without really low timings, it is marginally quicker than S939."

 

I did, the 533 beat the 667 by 12ns (112ns vs 124ns) in the memory latency test and the 667 beat the 533 by 2176MB/sec (10K vs 8K) in the maximum memory bus bandwidth test.

Benchmarks? I've only posted a few, SuperPi and 3DMark05/06. I'll rerun them after I install my 2GB.

Are you referring to AM2? It launched back in May - they've even already had a massive price reduction on the entire AM2 line. Here's a couple reviews:
Check out page five of this Anandtech review for speed comparisons between S939 and AM2. Here's another review from Tom's Hardware.

 

Wow you been busy!

"I did, the 533 beat the 667 by 12ns (112ns vs 124ns) in the memory latency test and the 667 beat the 533 by 2176MB/sec (10K vs 8K) in the maximum memory bus bandwidth test."

Read it again : "(10K vs 8K) in the maximum memory bus bandwidth test." That's ESTIMATED, Nick not MEASURED.

What's with the AM2 stuff? I only mentioned that I read they were comming out with the DDR2 platform,no? You asked where I'd been?

I see you got new 667 CAS5 on the way. Which brand did you go for?

I still stand by my 677 vs 533 statement, IMHO, not worth it. YMMV!

Cheers!

 

I understand that, since the FSB of these Yonahs can't take much above that, however, it is a proven fact that 533 is indeed slower than 667, which in turn is slower than 800, 1000, and 1067...

Well, I couldn't be sure, your statement spoke in a future tense that AMD was "going to the DDR2 bandwagon soon", when in fact the DDR2 platform is already out. I sincerely thought by your statement that you were unaware of the AM2 launch. I meant nothing else by it.

I bought two 1GB Transcend chips on sale at Newegg.com. It was a one day sale with free shipping - $74.99 each with $10 MIR for a grand total of $130.76 after stamps to mail in the rebates... Pretty sweet deal, I think! After Merom hits and the prices stabilize, I'll pick one up and then load a hacked BIOS and be able to fully utilize the RAM's speed... assuming it all goes to plan. LOL

 

Thanks for taking the time to do that, it was an interesting read.

 

[nick_danger]I understand that, since the FSB of these Yonahs can't take much above that, however, it is a proven fact that 533 is indeed slower than 667, which in turn is slower than 800, 1000, and 1067...

Well that ESTIMATED stuff bugged us last year. It is not explained the Syssoft Docs and several people have questioned this. The best we can guess is that it's theoretical? Now as you say proven fact or not, I have to weigh what I have seen as the measured results;

OCZ-667-PC2-5400 Bandwidth
RAM Bandwidth Int Buff'd iSSE2 : 3639 MB/s < YES!
RAM Bandwidth Float Buff'd iSSE2 : 3604 MB/s

HYNIX 533-PC2-4200 Bandwidth
RAM Bandwidth Int Buff'd iSSE2 : 3630 MB/s
RAM Bandwidth Float Buff'd iSSE2 : 3631 MB/s

Only these numbers are shown on the cute color graphs and are used to compare other RAM.

And in real world I just don't feel any real difference. I was kind of hoping I would have to chain it down or something.

If you can, please BM your new ram. I think the free Syssoft lets you do this or another utility of choice. If we start a thread of ram BMarks, other users could make some choices from them?

I bought two 1GB Transcend chips on sale at Newegg.com. It was a one day sale with free shipping - $74.99 each with $10 MIR for a grand total of $130.76 after stamps to mail in the rebates... Pretty sweet deal, I think! After Merom hits and the prices stabilize, I'll pick one up and then load a hacked BIOS and be able to fully utilize the RAM's speed... assuming it all goes to plan. LOL

Awsome deal! And I'm on the same Merom road as you.

 

Yeah, that can be disappointing, but like you said - Oblivion doesn't crash! Because I ordered my lappy with 512MB RAM (no "free" 1GB offered), BF2 barely loads, HL2 gets chunky when accessing the swap file, and I can't have too many apps open at once. I will notice a massive change in real-world performance with 2GB.

According to Newegg.com, my RAM is on a truck in my city, so it should be waiting for me when I get home. I'll download SiSoft and run a test before along with a couple timedemos for HL2, Halo, and UT2004, then install the RAM and rerun them. Shoot... I could even just install 1GB and run the tests and then install the 2GB and run again... That could be interesting, but it won't really be a solid comparison since the 1GB test won't be dual-channel. Look for the results tonight!

 

This thread deserves a sticky, since we get a lot of questions about 533MHz vs. 667MHz.

 

Hmm, just want to add one little point.
Of course 667mhz is faster than 533mhz. It's faster by exactly 25%, assuming equal timings.

But, that's just the RAM's speed in itself. The performance difference when actually installed in a motherboard is an entirely different beast. And it depends on a lot of other factors.
It seems that few CPU's need the extra bandwidth offered by 667MHz memory, and so, the effective performance difference is tiny.

The point is, though, that depends. Another system might launch next week that can actually use the full 25% extra performance offered by 667MHz ram.
Or if not next week, then maybe next year.

My point is just that you have to be careful when saying that "667 isn't really faster than 533". It might be the case *now* with *your* system, but it's not an universal truth. The RAM is quite a bit faster, so the potential is there for a noticeable performance boost. As soon as we've got CPU's and motherboards that actually make use of it.

 

Jalf

Point taken and understood. You know that 667 ram is faster, heck, ask anyone

The benchmarks reflect what was available on those dates and on a lone machine. Hopefully OCZ and all the others will release 667 CAS 4 (or 3!) soon. You can bet I will benchmark those as well

Everyone, please submit your benchmarks!

Cheers

 

Oh, it wasn't really directed at you. I just figured, since this thread has been stickied, it was probably worth pointing out for anyone who just read the thread.

 

Have you tried lowering the timings on the 667?

 

AFAIK this is locked on the dell laptops.

 

Are they locked on Sager laptops as well?

 

You can't fool with specific settings like that - laptop BIOS' are very primitive.

 

Ahh well at least I know the DDR2 533 won't go to waste... When I get a new lappy I was hoping to re-use at least 1GB of the ram I have but was thinking much better to get DDR2 667 to "sync" with the FSB speed.

 

here are some 533vs667 gaming results on a desktop Core 2 Duo system

http://www.anandtech.com/mb/showdoc.aspx?i=2810&p=6

lower timing seems better for gaming... don't know if it holds true for laptops.

 

Good find! I was too lazy to try that with FEAR, I had already finished and deleted the game

 

I just added 2 x 1GB PC2-5300 in my Sony SZ.
The BIOS/SPD (dunno who is responsible for it) clocks the RAM down.
Now the RAM runs with the same speed like the memory controller (533Mhz DDR instead of 667), what means the CAS is now 4 instead of 5.

So the benchmarks didn't change at all for me between DDR2-667 and DDR2-533. It changed a lot between the slower PC2-4200 my laptop was shipped with and the new RAM though, because it slowed down the memory-controller.

 

Those Anandtech results don't correspond to what you'll see on a notebook, though. They're using a desktop motherboard with highly tweaked timings. They got their DDR2-533 to run at CAS3 and the DDR2-667 at CAS4, vs. the CAS4/CAS5 stock settings that the memory is rated for and which you're pretty much stuck with on a notebook.

Their settings
CAS3 @ 533MHz = 5.6ns/access
CAS4 @ 667MHz = 6.0ns/access
i.e. the DDR2-533 does have lower latency

At stock settings, though, the latencies are exactly the same
CAS4 @ 533 MHz = 7.5ns/access
CAS5 @ 667 MHz = 7.5ns/access
More clocks, yes, but the same amount of actual time.

 

So.... I have just purchased a Dell D820 with Core Duo T2500. It has the Mobile Intel® 945PM Express Chipset. It has a 667MHz Front Side Bus. Also indicates "Dual-channel DDR2 667MHz Memory Support." Which is better for me 533 or 667? And, if there is a difference, will I notice it in everyday operations or is it just a technical measurement difference? I ordered it with 667, but could easily (I hope) return the 667 for 533.

 

Which ever you got the best deal on You won't feel any difference in real world. Try to stay with a brand name.

 

What are some good brands?

 

Yep, no way we can use that ram, those timings and get those FPS on our stinky old laptops

I do find it interesting, after all the tweaking, is the actual FPS across the different speed ram vs games.

 

Alright, we have an incredible amount of questions regarding DDR2-533 vs. DDR2-667, so this thread gets a sticky again.

 

well i think this thread is not too accurate to be taken as a prove for which is better

why is that ?

cause u never buy a machine according to benchmarks results ( or att least me )

u buy a machine when u see scores from practical results, like compare the fps of gaming using DDR2-533 and DDR2-667, compare the time need to compress or extract a compressed file ( 1GB or something ), compare the time it will take the system to calculate 32mb in SuperPi

but comparing using sandra scores is not something very good since those scores are mostly on paper and won't show real life performance as it is

i will be glad to see anyone comparing 533 to 667 in gaming and office use

then we can know which si better to buy for ur system
DDR2-533 @ Cas4
OR
DDR2-667 @ Cas5

 

I agree. Sandra, at least, let me return repeatable numbers.

I tried to do a test with my then game of the month, F.E.A.R, but could not repeat the frame rates, plus or minus 3 or 6 FPS to decide a winner.

The main thing was i could not feel any difference between the two. I would still suggest to buy the cheaper of the two unless someone slaps me with CAS 4 - 667.

FWIW, the OCZ ram has been faultless to date.

 

would it be bad to have a gig of 553 and a gig of 667 in the same computer

 

so what si ur question ???

1gb 667 @ newegg for 98$

 

You will be fine. The 667 module will "step down" to 533 speeds.

 

Just to clarify, it is possible to add a 1x1024 MB of PC5400 in a computer with a 533 FSB and 1x512 MB PC4200, correct? It will then simply function as PC4200, no odd issues? The PC5400 is cheaper than the PC4200 these days.

 

As long as both are listed as usable on your motherboard, the faster ram would slow down to the slower for backwards compatibility.

I haven’t shopped ram in a while but can only guess as to why the 5400 is cheaper than the 4200

 

Yeah, upon closer inspection it is due to the higher latencies of the PC5300. Think I'll stick with P4200, and wait until after xmas until prices drop a bit.