Which RAM is a better choice / makes more sense

That's not true at all! Absolutely wrong!!

1. I'm talking about Laptops and not Desktop replacements! A 16 or 17" "laptop" ... nothing I would put on top of my lap! )) Name one 13-14" notebook with more than 2 RAM slots? Mine is a business series one and has 2 only. Lenovo T400/T410 series which is 100% business too has 2 slots as well! My previous 14.1" had 2... MacBook Pro has 2.... should I continue?

2. The amount of max RAM is limited by the Chipset AND the CPU, not only by the available slots! i5 520M is limited to 8GB RAM, check out the specs. So are the new Sandy Bridge CPUs as well - Take i5-2540M for instance! Or look at the most expensive mobile i7 CPU (nearly 600$):

3. There are no 8GB SO-DIMM modules! the max size available nowadays is 4GB!

 

There actually are 8GB DDR3 sodimms. They are very expensive and very rare. I think preconfigured with a new computer they can be as much as $2000 per stick, but I think someone found some for as low as $500 per stick somewhere.

Samsung Gets To Work On 8GB Laptop Memory Sticks - PCWorld

Also if you go to Dell's website you can configure some of their workstations with 8GB modules.

 

Ok, let me correct myself:

There are no 8GB SO-DIMM modules freely available, I'm sure NASA do have some! Even if there were - it wouldn't help the fact that laptops (not talking about big, heavy desktop replacements!) support up to 8GB anyway.

 

The new Sandy Bridge i7's will accept 16GB RAM. That RAM number on the spec sheet is for each memory controller, so 8 x 2 = 16GB. The board handles the number of slots and capacity per slot ( typically 4GB these days ).

 

If you are right ("That RAM number on the spec sheet is for each memory controller") then Arrandale should be supporting it too!? I see no difference in the specs about memory, both have 2 Memory Channels.

But going back to the RAM usage discussion, here is my RAM usage at the moment, the PC was on sleep throughout the night and been on for about 2 hours now:



Notice the 2600+ FREE MB! Just in a few hours, this amount would melt down to 20-50MB in the interest of "Standby" memory which will become nearly 4GB big with the same set of programs, web pages, etc opened on the PC!! I don't really understand what is being cached. Going through the windows I notice there are NO hard faults per sec, which means that everything is in the RAM already

 

Intel's info on their website must (unfortunately) be taken with a grain of salt. It should be used as a guide but not the absolute truth. There is incorrect, misleading, and missing information on there all the time.

 

The new Sager / Clevo SB notebooks accept up to 16GB RAM in four slots. Max 4GB per slot, max 1333MHz DDR3. Only 8GB (two 4GB sticks) of DDR3 1600MHz is allowed.

Same for the ASUS G73JW, which is previous gen i7. The biggest thing is the mainboard only handles 4GB per slot. Most laptops have two slots, hence the 8GB limit.

 

Sager isn't really a typical example for a notebook. They produce desktop replacements / mobile workstations mainly. That's not really a laptop.

 

Not really true. 6.8 lbs. 15" GTX 485m @ 1.5" thick isn't big or heavy. Either way the point was that the CPU's do no limit memory to 8GB, it's 16GB, and the limiting factor is number of slots on the mainboard.

 

What is the CPU and what is the mainboard? Isn't it by any chance a desktop CPU?

6.8lbs is over 3kg ... that's not really a light notebook nowadays - even Lenovo T400 is lighter with its heavy metal construction... but that's not the point.

Talking about mainstream notebooks (Minding the difference between mobile workstation / desktop replacement and a laptop!):

1. Most of them run mobile "i" CPUs. My quick research made clear that neither of the i5 or i7 mobile CPUs officially support more than 8GB ram (according to Intel's specs). If it is true that this is 8GB per Channel as you suggest, which I am definitely not sure about, this would mean max 16GB, as they only support 2 channels.

2. Most of them have 2 slots. The difference between a desktop replacement and a notebook is in fact becoming subtle... but the understanding nowadays is that a machine over 3kg and 15, max 16" is no longer a laptop. And to be honest I can't think of any in this category that has more than 2 slots, including the business lines of Sony and Lenovo!

 

No, it's a mobile i7-2630QM, 2720QM, 2820QM, or 2920XM.

 

The above 2 statements actually make sense:

the mobile i7 CPUs support max 8GB RAM. If it is really per channel that means max 16GB. The Sager has 4 slots, that's why limitation of 4GB per slot = Max 16GB which the CPU can handle. (2 slots per channel ==> not more than 4GB per slot to avoid putting 16GB on one channel)

This could be a theoretical explanation, which would mean that laptops with mobile "i" CPUs do support up to 16GB, but as most (or even all) of them have max 2 Slots, this will not be possible until 8GB modules hit the market.

 

True, but then it will probably lose dual-channel since it's 8GB per controller, so each 8GB will run in single channel mode. But big whoop. You gain maybe 5% running dual channel anyhow.

 

Whether officially or not, Nehalem and Sandy Bridge mobile processors support more than 8GB of RAM total or 8GB of RAM per channel. You can buy a laptop with a mobile processor that has 4 slots and comes with an 8GB stick of RAM in each slot for 32GB total.

 

I'll be very interested in reading a proof of this statement... until there is one, it is just a theory, as I doubt you tested it yourself

 

He knows what he is talking about

CES: Dell goes high-end with the new Alienware M17x R3 gaming laptop - CES 2011 CNET Blogs

 

Pay attention that the discussion was about current laptops - those we already have. No doubt that laptops WILL eventually support more! But can you provide a proof that my laptop or your laptop support more than 8 / 16GB ram? I'm not really interested in individual cases that are about to hit the market. I'm interested in knowing what I have and can do.

 

It has nothing to do with CPU but available slot, ram modules. If you can find an existing model that has 4 slots and find 8GB module and the BIOS doesn't cripple it, you can.

The same CPU architecture(same memory controller) is used on desktop/server and they can go way beyond 16.

My laptops cannot as they are still using the north/south bridge(I am not using i* and all are DDR2) and all have only 2 slots.

For you, it is simple. You are stuck with a max of 8GB because you only have 2 slots and being a consumer model, it is very likely Sony has crippled it. So even you are willing to pay for 8GB module, it may not POST.

 

Proof: Dell Precision M6500 Mobile Workstation Details | Dell You can order an M6500 with 32GB TODAY.

And chimpanzee is right about being able to put 8GB SODIMMs into any Core i laptop. The CPU supports them.

Also I am sure that even if your core i laptop does not support more than 8GB total, 4GB in each slot, it will magically support 16GB total when 8GB modules become more available. The same happened with Core 2 laptops that originally only officially supported 4GB total.

 

This is the part I am not sure. we know that the CPU is not a problem(it never was). But does the BIOS play a role ?

 

I can personally attest to BIOS limiting RAM. My Vostro 1500 houses the PM965 chipset which Intel ark documentation says 4 GB max RAM but when PM965 came out 4 GB modules weren't out yet. Many accounts show 8 GB is supported on PM965. Unfortunately Dell for some reason on the 15.4" notebook BIOS restricted to 4 GB RAM as I tried putting in 6 GB DDR2 into my Vostro 1500 and it did not POST. However I tried the same RAM into a Vostro 1700/Inspiron 1720 and it worked just fine. And yes I had the latest BIOS from Dell. And it's the same chipset.

 

2720QM, 2820QM and 2920XM is the only one supporting 1600 MHz RAM and they only support 4GB of it.

 

Gracy123,

You're welcome!

Let me start answering you with a little more background in how I setup a partitioned system:

See:
http://forum.notebookreview.com/har...-hitachi-7k500-benchmark-setup-specifics.html


Note that this 'outline' has been updated since I posted it (I will make a new post to reflect my current 'best practices' soon).

The main point to take away from the above link is that the first 64GB partition is used for all the temp, swap, scratchdisk and downloaded files ( C: is not the first partition, it is the second one).



Okay, with that out of the way: with your 3 partition setup - how much free space is available on C:? If you have much more than 50GB free, then I would shrink it down more (20GB free seems like a good bet). Keep in mind that you don't need to use the remainder of the 'shrunk' disk space - keep it in reserve (unallocated) in case you need to expand/grow your C: partition in the future (due to a service pack update, additional software, etc.).

Set your pagefile size to the 'System managed size' radio button for each partition (make sure you click 'Set' for each partition) and reboot the computer once. Now, run an offline defrag with PerfectDisk on each partition (so that it optimally moves the pagefile.sys files we just created) and then also run an online defrag run on each drive too.

If you set the above Virtual Memory options correctly, then you should have 18018 MB showing as the 'Total paging file sizes for all drives:' on the Virtual Memory section of the Advanced tab of Performance Options. This MB amount is based on your screen shots that show 6006 MB as 'Total' RAM (x 3 partitions).


Now, all you need to do is use your computer normally - you should notice an obvious 'snap' to the newly configured system once it is fully booted.

Although 8GB RAM is the maximum a truly portable notebook can use today, that does not make it any less the minimum they should be shipped with (assuming a 64bit O/S). As has been mentioned, Dell's M6500 line can be configured with 32GB RAM for at least 6 months or more now.

PerfectDisk's 'Aggressive' performance mode is not too aggressive!

What it will do though is take substantially longer to run (at least the first few times), but what it offers in return is a system that gets fragmented much less quickly (because it aggressively fills the holes/empty spaces between files). Once set to Stealthmode, (and you leave the computer on idle once a week or so for about an hour), you don't have to worry about this again.


Finally, I will try to explain the 'why' of a system managed pagefile on each partition is beneficial:


First, just like our bodies when we try to 'diet' - Windows changes its behaviour when we limit or disable Virtual memory. Although it may seem a little faster (boot up, shutdown - yawn!), in actual use, it gets slower as we use the system because all the 'unimportant' stuff that was paged out (without an impact to us, I may add...) will need to be juggled in RAM with the apps/data we are actually working on.


Second, Windows can intelligently use multiple pagefiles by always using the one with the fastest access time (it monitors this continuously, in real time).

How does this help with a single HDD with multiple partitions with a pagefile on each? Simply put: if it needs to, it will use the pagefile of the partition it is on - not switch gears to 'jump' to another partition to use the pagefile there (even if that pagefile is on the nominally 'fastest' partition).


Third and final point I want to make: the key point is that Windows will only use a pagefile if and when it needs to - as long as you have the RAM to hold your programs, Windows will use a very, very small amount of the pagefiles for 'housekeeping', but otherwise your work overall will be RAM based and not limited to the speed of your HDD residing pagefile(s).


I hope I cleared up some of your questions?

Very curious to hear about your experiences if/when you do implement some of the things I've suggested above.

Good luck.

 

Thanks once again for your valuable input!! I still have a few questions though - I am trying to explain myself how it works, hope you don't mind
To answer the above question - currently exactly 50GB free on C:. I tend to use the desktop as a temp place for stuff, so I do need enough space on C...

Isn't it actually just the opposite? Aggressive leaves absolutely no free space between the different groups of data, which means that every time a file becomes bigger for some reason, it gets fragmented as the "new" data is written at the end of the disk because of the lack of space intervals.... In my understanding, the aggressive method is only good until everything stays the way it is, guess perfect for read-only data. That's why I chose the performance mode which leaves some empty spaces...

The above statement would be true only if the pagefile is mirrored on each of the partitions, but there are 2 possibilities:

Option 1: Each of the 3 (in my case 3 because of the 3 partitions) pagefiles contains different data, which automatically means fragmentation! The HDD will have to move the head backwards and forwards all the time if the page data is separated and fragmented on 3 different partitions on the same physical drive... hence very contra-productive!
Option 2: The pagefile is being mirrored on each partition. that means the writing load of the HDD would be tripled!... how does this speed up the system - the HDD would constantly need to move the head to serve all partitions in order to mirror the data! That should be contra-productive as well - this time in the phase of writing data!

This is without a doubt true. More RAM is the best and most reliable solution

But how about ReadyBoost? Wouldn't for example 4GB ReadyBoost drive bring much more in terms of pagefile speed enhancement than dealing with its size and position on the mechanical drive??

 

Gracy123,

No, I don't mind trying to explain it - but trust me, you need to do it (and therefore see it's effects) to see why I'm suggesting these settings.

Since you need the 50GB free space, well - not much to do there (but on your next build I would recommend the first partition be set aside for the temp files...).

As to the PerfectDisk settings... you seem to think that data knows where it needs to go on its own?

Any new data written will try to go to the next free contiguous clusters - it won't know to go to the 'seldom used' free space section, for example.

When a file is appended to, it will always be fragmented - with the default 'SmartPlacement Performance' preset all we are achieving is simply fragmenting the free space too (for no good reason).

How I actually have my systems set up now is by editing the 'SmartPlacement Performance Aggressive' preset and taking out all free space between the different groups of data - except the space between the recently and the occasionally modifed groups.

I increase that size to around 20GB or more (depending if PD11 gives any warnings about low disk space). See the attached image with 18GB set between these two groups.

What this achieves is forcing Windows to write to the faster part of the drive while also giving it the biggest contiguous free space to work with too.

It does this by moving the occassionally and rarely modified files to the end of the drive (but, don't forget even the 'end' here is still at the fastest part of the platters) and putting all the free space as close to the beginning of the drive as possible (depending on how many boot, directory and recently modified files the drive/partition contains).


Pagefiles are all bits and fragments of different files, they are never 'unfragmented' internally. Don't let that line of thinking stop you from trying this method!

Mirroring pagefiles makes no sense (and Windows doesn't do it...) a pagefile just contains code/data needed 'just in case' of something that may or may not happen. As you noted, if it did mirror the pagefiles, it would be slowing down the system to a crawl and that is definitely not what is observed in my systems.

With Windows capability to intelligently use multiple pagefiles (it is able to do so dynamically, on the fly, with the particular set of apps/data you happen to be using at any specific time), the closer (mechanically) the pagefile is to the data/app we just invoked or are currently using; the faster the system will respond to our inputs.

While greater than 4GB of RAM will make a bigger difference than the suggestions I'm making, note that even maxed out at 8GB RAM, the above settings take a system to the next level. Not even doubling the RAM to 16GB will make it perform with the same 'snap' and directness as with the tweaks I am suggesting.

As for ReadyBoost, I'm using a very fast 8GB USB stick (and configured as 'dedicated' as a ReadyBoost drive) that I'm still evaluating.

It is not hurting performance, but I am not seeing a tangible increase either (with 8GB RAM, anyways).

Hope I've explained myself clearly enough for you?

 

You said notebook not laptop

no its not otherwise how is this possible

The Dell Online Store: Build Your System

again look above

also Dell xps 17 comes with four slots, people are adding more ram over the "maximum supported and it works fine

 

I don't mean to get in between your heated debate here but just wanted to point out that there might be more to official Intel specs in as far as maximum supported memory capacity than initially meets the eye.

Back in the humble days of personal computing I seem to recall that the 82430VX chipset (Triton III) could support 128MB but only the first 64MB were cached. The 82430HX (Triton II), geared towards workstations could cache all the way up to 128MB. The 82430FX (Triton I) only supported 64MB as I seem to recall so nothing to note here. The 82430TX (last and final, Triton IV) supported 256MB EDO, BEDO and SDRAM if I recall correctly, all could be cached. It didn't matter in those days how much L2 cache you had on the motherboard or on a COAST module, while it boosted performance going from 256K to 512K (some boards even came with 2MB back then - was a real wow factor at the time), it didn't solve the aforementioned issue.

What's my point?

If this happened in the past then I don't see why it wouldn't happen with Intel's latest i3, i5 and i5 series CPU's today. The MCH might supported more than what's officially noted by Intel but I'm pretty sure there's more to it than that.

If some of the memory can't ever be cached (thus portions of it exist in the CPU's L2 and L3 caches) then sooner or later (ideally under heavy workloads which use ample RAM) they'll be a significant penalty hit in the form of additional CPU idle cycles for the given execution thread. Is Dell going to come out and announce this publicly if true? I wouldn't bet on it.

Just wanted to point that out as I'm pretty well sure it holds true, even in this day and age. Sometimes more is less.

 

Just have to remember that the i* thing now has the memory controller inside the CPU. It doesn't make economic sense to cripple the max memory it can handle at this level.

In the past, i.e. the north bridge/south bridge case, you and grace123 may have a point. MCH/chipset is a thing of the past as far as memory is concerned.

edit:

this is the top of the line Sandy Bridge i7 2920xm spec. it said Max RAM 8GB.

http://ark.intel.com/Product.aspx?id=52237

this is a notebook using it

http://www.techpowerup.com/138722/E...Notebook-on-the-Planet-The-Eurocom-Racer.html

it said max RAM 32GB.

Now, should we literally interpret the ark page and said Eurocom is selling snake oil or accept the simple answer that the CPU can take as much memory as you can find for it ?

 

So people are too stupid to realize that all the RAM they put in their laptop is going unused? There is really no reason to believe that since the same thing happened with Core 2, where some chipsets were supposedly only capable of 4GB of RAM, but then magically overnight they became capable of 8GB of RAM. Also mobile processors are the exact same silicon as desktop and server processors, just in a different form factor. If desktop and server processors are easily capable of handling huge amounts of memory, what is stopping pretty much the same processors on laptops from doing so?

 

Personally, I use 17" DTRs on my lap all the time... and I'm a little person.

Aside from that, I'll also point out that the HP 8440w (14" notebook) has 2 memory slots, and can be purchased with 8 GB SODIMMs in them for 16 GB total... although it requires a Clarksfield (1st generation quad-core) to do it. You'll also be paying $4325 just for the memory, but... it's possible. Just expensive. And this is 1st generation Core i. I doubt Sandy Bridge would significantly drop the memory capabilities (although the fact that dual-core apparently doesn't support the 8 GB SODIMMs may mean that Sandy Bridge dual-core might not either). The 8540w is a 15.6", has 4 memory slots, and can take 32 GB total with a quad-core as well, with a listed starting weight of 2.89 kg. Oh, and as pointed out, the 8740w has 4 slots, and can take 8 GB SODIMMs in each slot for 32 GB just like the M6500.

 

Don't be so quick to dismiss the dual cores. The Dell M6500 comes with a dual core CPU as standard, and can still take 32GB.

 

Not quite correct. M6500 Dual-cores ship with a dual-core board, with only 2 memory slots (debguy actually ordered a quad-core, which came on a dual-core board with only 2 memory slots). According to havoctex (Dell representative), Arrandales won't accept more than 2 memory slots. Now, we do know that you can mount an Arrandale in a 4 memory slot board (and vice versa), but I don't think anyone ever got around to testing an Arrandale in a quad-core board with all 4 memory slots populated. If you want sources, this is all covered in the http://forum.notebookreview.com/del...ision/439109-new-m6500-discussion-thread.html.

 

Unfortunately further experimenting and testing will have to wait

I had a major system crash yesterday - after about 7h of work on the PC, it turned out that one of the 2 reasons was bad RAM (the new 4GB module)... so it is going back now and I am back to the "middle ages" with 4GB RAM....

 

Have you run memtest?

 

Yeah that's the first thing you gotta do with new RAM, run memtest/IBT/Prime95 to ensure stability. Nothing like finding a good deal on defective hardware.

 

Yes. The problem is the RAM came in a damaged packaging and the module was slightly bent. I ran Memtest when I put it in initially and did not find anything. It worked perfectly fine until this morning, but that wasn't the problem for the crash anyway... (more info HERE) As people say - "evil never comes alone" :-/

 

Gracy123,

Sorry to hear about your problems!

Just a thought though:

While you're waiting for the replacment RAM to arrive, can you try putting all of your partitions with a 'system managed pagefile'?

This will be interesting to know if it resolves, minimizes any of your 'pausing' issues you were experiencing with the original 4GB RAM installed.


Note that if you do try this setting (and leave it there), when you install the bigger RAM module that you will need to perform an offline defrag with PD again (the pagefile will grow to match the new RAM installed).

On a side note: you have more guts than me 'downgrading' the BIOS - that is one thing I have never done (successfully, anyway).

Good luck.

 

Interestingly I was just going to report about that I did that the other day after bringing my PC back to life... I have enabled pagefile on all 3 partitions (nearly 12GB).

However, to be honest I feel no difference - HDD seems to be overloaded as before. I am not saying there isn't any difference... I'm just not feeling it - neither in positive, nor in negative direction, so I would rather consider it minor...

Just as I switched to this tab, I had to wait for nearly 5 seconds for the content to load... terrible! Can't wait to get my RAM back

 

Hey, thanks for the update!

At least we know for certain that when more physical RAM is needed, then local (my word) pagefiles do not make a difference.

You did an offline (boottime) defrag and an online PD defrag too, right?

Would love if you could report back with your full 6GB RAM again.

BTW, does IE slow down that much for you (as much as Firefox does)?

 

I will report again when I get my RAM back Yes - I did offline defragmentation, I am about to start online as well with the "aggressive" setting for a change and try that too

Regarding IE - I can't really compare, as in the rare cases I use it, I use one instance only... can't really compete with the 20-40 websites opened in Firefox But I have the feeling Firefox is lagging the most, then comes IE and on third place Opera, which works the fastest among those 3.

 

Ok, I'm back in the game and enjoying the noticeable difference for the second time

However, after some experimenting I still have a big question mark in my head!!

Whenever I copy a large amount of data from/to the HDD, ALL of my available RAM seems to be used for that process, forcing windows to start caching everything in the Page file instead. This makes working on the PC during data transfer of big folders (say over a few GB) very hard. It seems to be sliiiiiightly better than with 4 GB, but still nothing I have experienced with Win XP!!

Why does a file transfer need SO MUCH RAM? If it is kind of a buffer, it is really huge to take up nearly 4GB RAM!! Do you also have this problem?? Try copying say 10 or 50GB folder from one place on the HDD to the other. Or even from external HDD to the internal or the other way around!

Here some pictures:

1st one: During the file transfer (looks like that a few minutes after I start - All of the free memory is used up and becomes "standby")



2nd one: Right after the file transfer ended.



You can see the Hard Faults per sec counter .... everything is very slow just as if I had 2 or 4GB RAM!

Any explanation/suggestion? Is it the same on your PCs? Try it out...

PLEASE CONTINUE DISCUSSION HERE!!!

 

Gracy123,

This is a Windows 7 'feature' afaik.

Glad you got back to your 6GB RAM - but you really seem to be able to use 8GB!

Have you disabled the pagefile(s) again?

Or, are they on 'system managed' for each partition?

 

1. buy cheapest you can. if you dont swap out both of them... I think you might have an issue of compatibility. sometimes memories just fight and work none because they're not from the same family.

2. pay attention to speeds. rule of thumb.. if you have a slower mem and a fast one, even if they run, they run at slower one's speed automatically . so 1333 is still 1066 with the 2GB

 

Dude.... 10 pages thread and you read just the first post? Thanks anyway

 

Windows "Superfetch" service also uses as much idle ram as it can to cache programs and files you frequently and recently opened to speed up loading in the future.

 

Sorry for bringing back an old thread. I have same problem as the OP because seller shipped me 10600S by mistake (I ordered PC3-8500). I don't want to go through RMA process though...

What is PC3-10600'S'? I noticed it has 'S' after 10600? What's the difference between 10600 and 10600S?

So PC3-10600S will have no problem right?

Is it better than PC3-8500 in terms of future use? My laptop is 3 years old so I may be looking to buy a new laptop and maybe put this 8GB RAM on my new future laptop.