cas latency or bandwidth?

im planning on getting 8gb of memory for my lappy in sig.
originally i was planning on getting 2x4 of 1333 cl9 but this article;

Everything You Always Wanted to Know About SDRAM (Memory): But Were Afraid to Ask - AnandTech :: Your Source for Hardware Analysis and News

made me think about getting 2x4 of 1066 cl7 since they are cheaper and accoring to the article, cl seems to be more important than bandwidth.
what do you guys think?

 

Latencies are mostly irrelevant for laptops, but I'm hesitatant to apply that to Sagers and Clevos. The bulk of laptop BIOS are far too locked down to take advantage of better timings.

 

Edit: Sorry, response was intended for wrong forum.

 

in my case. the sager 8690 supports 1333mhz. so it will definitely default to a max of 1333. my question is whether bandwidth or cas latency will give me more benefit. from the article in my op it seems that CL is more important at least on desktops.
i use video editing software like power director, ppro cs5, AE cs5, muvee; and i use photoshop cs5 with a few plugins also.
i do play games a lot too but i doubt if memory speed would have any effect on them. the same thing probably applies to my bd/dvd software, total media theatre 3 and powerdvd10.

 

If you can't set the CAS via BIOS like you can on an aftermarket/system builder desktop motherboard, you won't see any benefit. Your laptop will not necessarily default to a certain frequency - it might be capable of the frequency, but the processor has the final say on what frequency it's set at. The 720QM can run at 1333 or 1066, but the final clock speed won't be any different, just the multiplier.

My hesitation before was due to Sager and Clevo using very different styles of motherboards that might enable additional settings since they're built more like desktops in laptop housings, so more options might be enabled.

As far as benefit goes, no, you won't really see any. Those programs are somewhat processor intensive, but far more affected by hard drive bottlenecks. Maybe a couple percent different in speed. The biggest performance gains I've seen are realized in massive scientific datasets with heavy number crunching.

 

my bios itself does not have memory settings. however the new bios does come with cpu multiplier settings, ht disabling and other features. im not so sure if it will allow you to CAS latency or overclock ram, but i think it can since people with the kingston hyperx overclockable ram were excited about this bios.

i get your point about memory not providing much in performance gains. however im not sure if this hold true for the adobe cs5 programs since this are known to be memory hungry and 8gig is actually the recmmended minimum.
ive also read in some threads that cl, bandwidth will make little difference with our present laptops since the bottleneck will always be the hdd or the cpu. but then again, im not sure that holds true for the resource hungry cs5 programs.

i am however leaning towards 1066'es with a lower CL7 rather than getting the 1333'es, mainly cause thyre cheaper.

 

Ah, sorry about my previous response - I had thought that this was the forum dedicated to the Alienware M11x.

In your case, your laptop runs at 1333Mhz by default. You will definitely want to get 1333Mhz RAM. The problem with getting 1066Mhz RAM is that it may not be compatible for your laptop.

On a desktop, this isn't a problem, since your BIOS gives you a multitude of options to adjust RAM speeds, timings, and voltages. In a laptop, you do not have these options. Laptop manufacturers also likely never tested your laptop using RAM that operated at a speed other than 1333Mhz, simply because the laptop was spec'ed to run 1333Mhz.

So, you run the risk of buying RAM that is incompatible (which it most likely is) because your laptop's BIOS and chipset was most likely never designed to accomodate different RAM speeds, timings, and multipliers. The only potential upside is a miniscule bump in memory performance, which you will only notice in benchmarks - it is not something that will ever show up in real-world performance.

In short: Forget the 1066Mhz idea, and stick with default 1333Mhz at default timings. Even trying to get 1333Mhz RAM with tighter timings won't work, because your laptop will most likely use its default timings.

 

thats great advice kent +1
but just to be sure ill maybe post a few ?'s for the other guru's in the sager section.

 

Probably the best idea, since they'll be the most familiar with the particular BIOS in question. Generally, a speed bump provides far more performance improvement than a latency bump. Keep in mind - most tech articles are written for desktop components, not laptop. Our chips are underpowered, but I believe they're more flexible in terms of automatic voltage and multiplier control, so we're more impacted by fsb than desktop folks are.

As fars timings go, they were all the rage 6-8 years ago. There aren't any serious articles on latencies anymore because the performance gain is never worth the cost. At best it's in the 0.5% range, if that.

 

Latency is part of bandwidth, so the only thing you should be looking at is which one has more bandwidth. This is usually the one running at a higher frequency. 1066/7 is only a smidge more than 1333/9, so by those numbers alone you would think the 1066Mhz memory would have higher bandwidth, but cas latency is only one of the many memory timings. On top of this, memory bandwidth is just a non-issue for modern computers, so don't get too worked up about choosing the perfect memory.

You don't know what you're talking about. Just because laptops generally do not have user-selectable options does not mean that can't run different settings. 1066Mhz ram is 100% compatible with his laptop, I guarantee it. You are just making stuff up.

 

yes i just checked in the sager forum and there are some guys actually using 1066.
so i guess im back to selecting 1066 as my ram
thanks trott

edit: would you care to comment on the badwith or latency choice? esp for my cs5 programs. or should i just get the cheaper since it would hardly matter?

cheers

 

Np. You nailed it on the head with this: "or should i just get the cheaper since it would hardly matter?"

 

Did you see This post of mine?

http://forum.notebookreview.com/6692720-post21.html

Give those a good read.

 

those articles are almost the same as the article i posted in my o.p. but theyre one year apart haha.
never new anand recycled articles. it reall is a good read though, and it was what got me thinking about CL and bandwith hence this thread.

 

the article favors CL over bandwidth in the desktop environment. i was wondering if such held true for notebooks also, or if it even mattered at all (considering i use adobe cs5 products)

 

Ultimately, you'll spend more time picking your nose in a day than you'll save in a year by upgrading for the sake slightly higher fsb or lower cas. Buy the cheaper one, save your money and buy a box or two of Kleenex.

 

You could always buy the 1333mhz ram and run a few benches, then change the frequency to 1066mhz and change the timings to cl7 (and the rest that would also change) in your bios or with thaiphoon burner.
See which way you get better performance.

Or it's possible that you could use thaiphoon burner to flash 1066mhz ram to 1333mhz @ CL9.

 

that program really looks interesting. ill have to ask if the updated sager bios does allow ram settings. if not, i can use this.
would you have any tutorial to this in nbr?

 

Here is a good guide for older ram:
http://forum.notebookreview.com/5333093-post14.html

A few things to note, thaiphoon burner might have a problem with your 64-bit OS, so that would mean you would either have to install some other OS to flash or try using SPDtool (SPDtool gives you a greater chance of bricking your ram).

The latest trial versions of thaiphoon burner don't give you all of the functions, so you can't flash with it. You need to download an older trial version (I have one I can upload if you need it).

 

i just noticed that i linked to the wrong article in the op. what i really wanted to link to was:

Memory Scaling on Core i7 - Is DDR3-1066 Really the Best Choice? - AnandTech :: Your Source for Hardware Analysis and News

which is essentially the same article as othonda points to. my bad

 

First as you figured out the person who said 1066 might not be compatiable is an idiot. If you support 1333 you can support 1066. The lower freq are always compatiable. Just liek if you support DDR 3700 or 4200 you could support 3200 2700 and 2100

Second who ever said lower latency give you better bandwidth is a moron. Almost always the lower freq has a better latency and lower bandwidth. Ussually when u increase bandwith you loose latancy.

Now Moral Hazard is right. if you get the 1333 ram and if you can downclock it, it should have the same latancy as the 1066. This would give you the best of both worlds. You can just adject the two for whatever your purpose is.

Honestly, for what ur doing i don't know which would be better or even if an effect would be noticeable. I think and dont quote me on it. If your program caches alot to ram and those things cached are small files that are read rapidly i think the lower latancy would be better (just like how flash is better than a HDD when it comes to small files. Thats when the access times matter). Now if they are large files dumped to ram than the high bandwidth would be better. At least thats how i understand it. I hope I helped a little.

EDIT: you can also double check your ram latency with cpu-z. if your curious on what your latency is now. Also DDR3 has a huge bandwidth....i am not sure if you can even max it out. As i said your biggest clue is wether or not your dumping large files or tiny files to ram.


EDIT: also i believe there are 5 different timings for ram. They tend to be the same number minus the last one is ussually the highest i think...its been a long time so don't quote me. Just go to wikipedia or newegg and check the rams latency out.

 

thanks dcmaker, your explanation makes much sense.
i guess in the end, as shown by the anand articles the very best you can hope for is a minor increase, which maybe i will just see with after effects - the program i use the least.
i think ill just get the 1333/cl9 instead. looking at prices its just a 10/20 dollar difference right now.

 

There are actually a lot more than 5. If you run memset you will see them.

Some might say the most important is the CAS latency.

 

Ah i did not know that. I was going off what you see when buying ram and when twiking it in the BIOS.

Also travlbug a big thing to look at too is if your MB supports triple channel. I ahve seen some highend laptops support triple channel. That might be benificial but i am not sure.

EDIT:+rep for moral hazard...i didn't know there were more than 5 ^^


EDIT:as some comments in that article said you might also want to get an SSD. That'll boost your performance alot. Especially load times.

 

unfortunately it doesnt support triple channel

 

Try the memory advisor from Crucial's website.

For your laptop with Intel PM55 chip says:
Each memory slot can hold DDR3 PC3-10600 with a maximum of 4GB per slot.

It says the memory inside your laptop needs to be DDR3 memory with support for DDR3 PC3-10600 speeds or faster. They don't mention PC3-8500 anywhere for your model.

For my laptop with Intel HM57 chip says:
Each memory slot can hold DDR3 PC3-8500, DDR3 PC3-10600 with a maximum of 4GB per slot.

 

thanks twiz but many users of the sager 8690 use 1066 ddr3 and just checking now its a standard option for memory in sager resllers like xoticpc so it will definitely work with 1066

 

Wow, again someone who thinks the know it all but doesn't. Memory bandwidth is a product of frequency and latencies. Frequency =/= bandwidth. And if latencies did not affect bandwidth, what do latencies even do?

 

If you have one 4Gb chip in your laptop already just find its matching chip. If you have two 2Gb chips I guess they become useless as you have to replace both of them. What's inside right now? PC-8500 or PC-10600?

 

The relationship between memory latency and memory bandwidth is the same across several computer concepts. Latency is the time it takes for the first piece of data to arrive after it has been requested, while bandwidth is the rate at which it transfers.

If this was a network, "latency" would be ping, and "bandwidth" would be the network bandwidth.

If this was a hard drive, "latency" would be seek time, and "bandwidth" would be the transfer rate.

If this was a highway, "latency" would be the speed limit, and "bandwidth" would be the number of lanes in the highway.

 

How am I "making stuff up"? Find one statement in my post that is false.

"May" indicates "possibility" or "chance". It is not a definitive.

"Risk"... again, indicating "possibility" or "chance".

Anything false there? Only *potential* upside (not guaranteed upside) is better memory performance, which would not yield any noticeable benefit.

Don't get 1066Mhz because it introduces risk for no reward? Something wrong with that advice?

Just because you "guarantee" that "1066Mhz ram is 100% compatible with his laptop" doesn't mean it's good advice for the OP.

A slow 4200rpm SATA hard drive is 100% compatible with the OP's laptop as well... it doesn't mean he should run out and buy one.

 

i have 2 x 2gb ddr3 pc3-10700h

 

The analogy only works for networks since you're talking about distinct packets. For hard drives, it's got to be a massively fragmented hard drive. Seek time is mostly irrelevant for large contiguous files. Highway analogies don't work particularly well. Latency does not limit speed. It's more akin to a car on the ramp trying to get on the highway and needing to wait for a gap. Bandwidth would relate to the to speed limit and lanes, but that's still not quite right. It doesn't work unless you're on a theoretical highway with all vehicles moving at the same speed and no lane changes, and that doesn't really happen.

He has an i7 720QM. The 720QM runs on 1066 or 1333, and the Intel spec require the motherboard to run both. No risk or chance involved.

I stand by my nose-picking-Kleenex solution. There are benefits, but not worthwhile as far as I'm concerned. Better memory performance CAN yield benefits, just not in this particular scenario.

Again, there's no risk. It's fairly well established on the laptop and desktop platforms that performance gains between 1066 and 1333 frequencies are negligible - on the other of 0.5% - 1%.

 

I agree with kents analogies...he is dead on. latency is response time...just like the seek time on the harddrive. Its the time for it to reach the file. and bandwidth is the rate it can send it. RAM does the same thing. The latency is how fast it can respond to a request. Hence why i said if his tasks for his programs require many many small files and rapid accessing than a lower latency would show benifits over a larger bandwidth. Its just like a harddrive or your internet connection in a game. Would you rather have a 10Mbps/10Mbps with a 250ms ping and a 50ms jitter or a 1Mbps/1Mbps with a 45ms ping and a 4ms jitter? You would want the lower bandwidth and lower latency. If your dealing with 1 gig files in your ram and lets say codeing them into a movie or whatever than high bandwidth would be better. Though if your dealing with 4gigs of 512k files than a lower bandwidth with a lower latency would be better. Its the same if you got a harddrive that can transfer at 130MBps and has a 19ms response time and you got a harddrive that has a 90MBps transfer speed and a 7ms response time. The faster transfer rate will do wonders with contineous files. But if you got 64k or 512k files than your 7ms drive will kick that ones . Thats why SSDs destroy even 15K rpm hardrives. 15K hardrives have like a 4-7ms but a good SSD has a .3ms. Does this make sense?

 

Wall of text is painful.

There's no disagreement in there. I didn't say the analogy didn't work for hard drives. The analogy just needs to be limited to one type of activity. RAM timings are extremely consistent and the analogy needs to be the same way. With networks you still have the request-wait-transmit-wait-confirm-wait-request-etc pattern that's somewhat similar to the RAM timing patterns, regardless of speed or scale. With hard drives, the pattern is similar, but highly inconsistent depending on whether the head seeking to the next sector in the cluster or if it needs to jump to an entirely different track.

 

Now your just being nite picky. The basic concept is all were getting at....we are not talking about wether the file is fragmented or spread throughout the drive.....honestly your just being ridiculous. If its that big of a deal use SSDs for the example...they dont have fragmention issues and they work just like RAM but slower. We are just helping him understand the basic concept of what the acces/response time does and bandwidth....no need to make this more complicated....your being too technical....you are being like Data lol

 

I'm sorry, I was wrong about the exact effect of ram timings. I was extrapolating decreased bandwidth from benchmark results of timing and frequency settings. I knew there was the max theoretical bandwidth based on frequency and I wrongly believed that looser timings made it harder to get close to that than tighter timings. It is a bit more complicated than that though. I think I've figured it out now though.

No, almost all of it is false.

You defend your statements by saying they were not definitive. But you say this and that may or are likely or whatever, when they never are or are extremely unlikely, so therefore those statements are erroneous.

Yes, there is no risk. I do agree there is little reward though, the only thing you said that was not false.

I wasn't advising him to get the 1066Mhz ram, just telling him that what you said was mumbo jumbo:

 

This thread is hung up on CAS timings, frequencies, and memory bandwidth and I'm being ridiculous and too technical? That's amusing. Few people can do analogies right and the explanation is usually a hell of a lot easier without them. If you're gonna do it, do it right or don't bother. Or at least stop while you're ahead (why the heck were highways in there?).

Both Anandtech articles linked in the original post explained what CAS timings are in the grand scope of SDRAM. Simplifying it beyond what Anandtech provided produces misconceptions when bandwidth latency timings and frequency are related back to each other.

Regardless of any debate over frequency vs timings, we don't know what, whether, or if his motherboard even supports the timing in question (regardless of adjustability), so this entire debate is moot. It's a question of frequency alone and even then, it's still a marginal improvement at best.