Can DDR2 800 be used on DDR2 Mobo?

[lunatic359]

I was looking in getting a mobo that still supports AGP and possibly PCIe. I just bought my AGP the beging of this year and want to hold on to it for a bit longer.

I was looking at Asrock 775Dual-vsta.

The reqs for RAM is DDRII 667.

Would I be able to use DDR2 800 on the mobo? I know if anything it either will not work or the DDR2 800 will revert to DDR2 667.

Just that alot of boards are coming out with DDR2 800 and I wouldn't want to get another set of RAM when getting another Mobo when getting another Vid.

 

[Guest]

Any DDR-II (Except maybe for DDRII-400) will run on this board.

The fact that it is rated for 800mhz means you can overclock or run the ram a faster than what the CPU needs.

No problem

 

[lunatic359]

I kind of understand the overclocking theroy.

As I understand it. If you dump a DDRII-800 onto a board it will show up as a DDRII-533. Then you must make modifications in your BIOS to run the modules at 800.

So would people actually be getting DDRII-533 that manufacture support overclocking to 800 - and is why manufactures are calling in DDRII-800.

In either case, thanks for the response. My best bet is to get the 800 right off the bat instead of 667 now then 800 later. I plan on having this board perhaps for at least 2 years. By then I can budget in for a new PCIe video card.

Thanks again.

 

[sruane]

...

I was looking at Asrock 775Dual-vsta.

...

Would I be able to use DDR2 800 on the mobo? I...

.

Yes.

 

[little_scrapper]

The 800 is basically the same price as the 667 so might as well underclock for now then be able to overclock later.

 

[lunatic359]

Thanks for the info/responses.

I will get the 800MHz RAM


Although another question arises.


I understand XP pro has a 4gb limit. With the 775Dual-vsta it only has 2 DDR2 slots. I want to aim for 2GB of Ram. My understanding is that 2x1gb will run slower that 4x512mb.

If this is the case should I just opt in getting 2x512mb now and then another 2x512mb when I get my new board. Or should I just get 2x1gb and later if decided i can get another 2x1gb maxing out ram limits?

 

[Mondoman]

...

I understand XP pro has a 4gb limit.

Actual limit is about 3GB.
...

My understanding is that 2x1gb will run slower that 4x512mb.

Nope, not to worry.

 

[TabrisDarkPeace]

4 x 512 MB will be slower than 2 x 1024 MB

The 4 GB limit isn't a Windows XP limit, it is a 32-bit OS Kernel limit.

You'll be able to address between 2.75 GB and 3.50 GB depending on what PCI / PCIe hardware is installed. (Inlcuding onboard things, like Adaptec RAID-5 controllers, etc).

With a x64 Kernel (Windows XP, Vista or Linux) you won't really have any 'addresses mappable to memory' ceiling.

 

[Mondoman]

4 x 512 MB will be slower than 2 x 1024 MB

Maybe, maybe not. Depends on the motherboard.

The 4 GB limit isn't a Windows XP limit, it is a 32-bit OS Kernel limit.

Since you would have to switch to a different OS to get a 64-bit kernel, and people typically use "Windows XP" to refer to Windows XP Home, Professional, and Media Center, not to "Windows XP Professional x64 Edition", in common usage it's a "Windows XP limit".

With a x64 Kernel (Windows XP, Vista or Linux) you won't really have any 'addresses mappable to memory' ceiling.

To clarify, the OS needs to be designed to run in 64-bit mode. Vista is fine, and if you choose a 64-bit version of Linux, you're fine, but "normal" Windows XP is stuck with the limit.

 

[TabrisDarkPeace]

If it is a "Windows XP limit" why do I have the same "problem" under Windows 2000 ?

Fact is that it is a "32-bit address space limit, 12.50% to 31.25% (rarely more) of which is allocated to PCIe and PCI devices for accelerated I/O".

Regardless of the Operating System if the Kernel is a 32-bit one you can only map memory to parts of the range not mapped for accelerated I/O to other devices.

If you know hexadecimal, consider the following example: (Taken from a Tyan K8WE S2895 with 4 GB, of which only 2.75 GB can be mapped to memory):

Anyone passing themselves off as a PC tech, or enthusiast referring to it simply as a "Windows XP limit" should be shot on sight.

If I started a thread titled "Windows XP limit" and never mentioned a 2.75 GB to 3.50 GB 'ceiling' or 'wall' no one would have any idea what "limit" I was referencing.Could be processor core limit, or socket limit, or limit of allocated memory per Win32 process, file limit in root folder, or any number of other things

Even implying that the limit is in any way associated (blame placed on) Windows XP is utterly wrong. It only acts as a catalyst to encourage smacktardism én mass. Providing IT assistance is hard enough without people changing rigid definitions into sloppy mutable definitions due to sheer laziness, ignorance, incompetence or negligence. - Please keep the Hardware forums technically correct, at the very least from a hardware perspective"

8) - TabrisarkPeace

 

[Surferosa]

Regardless of the Operating System if the Kernel is a 32-bit one you can only map memory to parts of the range not mapped for accelerated I/O to other devices.

If you know hexadecimal, consider the following example: (Taken from a Tyan K8WE S2895 with 4 GB, of which only 2.75 GB can be mapped to memory)

Ahh, ok- I wondered why windows 32bit couldnt handle more than 3gb memory; whereas 64bit could. That is useful info.

Anyone passing themselves off as a PC tech, or enthusiast referring to it simply as a "Windows XP limit" should be shot on sight.

Shame you had to spoil it. I'd be careful.. its a long way down from that pedestel.

 

[TabrisDarkPeace]

If a forum user falls from a virtual pedestal in a virtual forest, and everyone hears it, sees it, and feels it.... - Does anyone get hurt or care ? (Well maybe my ego gets bruised a little, but other than that)
- Sorry - George.W.Bush has his 'War on Terror', I have my 'Crusade for IT/PC Potentialism' - :lol:

I've used Dell machines, Pentium D's on a Intel i945G chipset, that could address 3.50 GB of 4.00 GB under Windows 2K/XP, as they 'packed' all the mapped PCI/PCIe I/O address space into 512 MB.

From Dell, I was impressed to say the least.

But yeah, generally speaking 3.00 GB - 3.25 GB is around the typical figure, unless you run SLI with a large broadcast aperture, or RAID controllers.

We really need a sticky on it, but if I wrote one people would just butcher it with inaccurate data - So what would the point be ?. (Unless I made it a PDF, and accepted submissions to improve it, but then no-one would make submissions) - Lose-Lose situation = Not cool.

 

[CompuTronix]

The Asrock 775Dual-vsta is a fairly feature-baren, no frills $58.00 board, that doesn't allow for much in the way of clocking options. If you plan on using the 800Mhz memory on another board fairly soon, then it might make some sense, but on the Asrock, the 800Mhz memory potential won't be utilized, and would be major overkill. Otherwise, 533 is perfect, and 667 only if you know the tricks of how to OC that board.

Hope this helps. Good luck!

 

[Surferosa]

We really need a sticky on it, but if I wrote one people would just butcher it with inaccurate data - So what would the point be ?. (Unless I made it a PDF, and accepted submissions to improve it, but then no-one would make submissions) - Lose-Lose situation = Not cool.

Well, a sticky seems a good idea to me. Some people may pollute your valuable knowledge on the subject (I for one would have no idea of how you would begin addressing the memory); but most people on these forums are clever enough to filter out the trolls from the treasure.

My friend, a diamond with a flaw is worth more than a pebble without imperfections

 

[TabrisDarkPeace]

Gah, Sigh.

Yeah I'll write one up this weekend then, maybe I can cross utilise it for work or training purposes somewhere else.

PCs are really bloody simple when you cut back to bare metal, it wouldn't take me long to summarise it.

So you after a Presentation format, or Spreadsheet, Document, PDF, anything in mind for layout ?

Once I get going with a doc I am fine, but the blank page looking back at me is discouraging (a template or something is all I am really after).

I've got Scott Muellers "Upgrading & Repairing PCs" (various editions), and The 'New' Peter Nortons Programmer's Guide to the IBM PC & PS/2, which should be enough to cover it all, and what 'acclerated I/O' really is, why it was a good idea back when 64 MB - 512 MB was normal, and why this has bite' us in the arse later on (SLI / Crossfire did not help).

and see link: http://www.quepublishing.com/bookstore/LargeCover.asp?isbn=0789731738 ; as she looks like a Bitmap and URL is via .ASP

(I got heaps of cool books btw, just these 2 stand out for this 'sticky')

 

[Surferosa]

Well, I'd keep it in regular sticky flavour. As I suggested, others inputs would be good too, and would not devalue your input- hopefully it would add value to it. You have touched on something that I always wanted the time to research/learn;

Optimising memory performance & the impact of changing memory addressing. Ive 'tinkered' with the pagefile- which is obviously different (virtual memory)- but thats it. It is definitely something I'd like to learn more on.

Im also interested in some of the books in your collection. Any more recommends?

 

[lunatic359]

Like said, I will go with the 800mhz vs 677mhz. I plan on having this board for at least 1-2 years anyway. Reason being is that I want to get more use and money out of my AGP card. I only bought it at the begining of this year for $250-300.

By then I'll be able to by a newer PCIe card and new motherboard.


As for the 32-bit memory limit. If this limit is tied to the 32-bit system, then why is Windows 2000 Advanced (8GB) and Windows 2000 Datacenter (32GB), both 32-bit OSes, can handle more than 4GB.

 

[TabrisDarkPeace]

Aha, a damn fine question.

It is because of a CPU instruction (or set thereof) known as PAE-36.

It permits 36 bit addressing on 32 bit systems by paging huge chunks in/out of the 'active' memory pool.

The OS will hide this from applications, but it prevents on application (process) allocating a contiguous 6+ GB chunk of memory, as it would need to span multiple paged 'pools'. (Most Win32 processes can only use 2 GB, some using /3GB in BOOT.INI and special coding technique can use up to 3 GB per process).

Some applications will load more than one process into memory though, and use inter process communication to work around it.

PAE-36 has existed since the Pentium Pro (considered the 6th generation of Intel CPUs), PAE-36 is also used with the NX/XD bit to mask areas of memory as NO EXECUTE / EXECUTE DISABLE on systems with 4 GB or less or memory in Windows XP (Service Pack 2 or higher required, however a CPU does not need the NX/XD bit for DEP to work, as PAE-36 offers a software workaround to do the same job).

NEW: http://www.microsoft.com/whdc/system/platform/server/PAE/pae_os.mspx

http://search.microsoft.com/results.aspx?mkt=en-US&setlang=en-US&q=PAE+36

http://search.microsoft.com/results.aspx?q=%2F3GB+BOOT.INI&l=1&mkt=en-US&FORM=QBME1

http://search.microsoft.com/results.aspx?q=AWE&l=1&mkt=en-US&FORM=QBME1

http://search.microsoft.com/results.aspx?q=%2FNoExecute+boot.ini&l=1&mkt=en-US&FORM=QBME1

http://search.microsoft.com/results.aspx?q=Data+Execution+Prevention+&l=1&mkt=en-US&FORM=QBME1

- Anyone training IT people should bookmark the above links, Or if you are want to become a Software Developer for a 'certain grade' of software (although x64 has a flat 512 GB to 1024 GB [1 TB] memory model - it is still worth reading).

- It is pretty cool stuff, but PAE-36 has been around since the early days of Windows NT 4.0 Server, and is why many '32 bit Kernels', such as Novell Netware 6.5 SP3, can use pools of memory that span 36 bits of address space normally (but not with a flat memory model).

Past 32 GB PAE-36 is pretty inefficient though, despite the fact it can handle 64 GB 'in theory' w/o major performance issues arising.

PAE-36 is also the reason why Service Pack 2 enables a software only version of No Execute / Execute Disable (Data Execution Prevention) using /NoExecute in BOOT.INI on systems with CPUs that lack the NX/XD bit. (All thanks to an underused instruction addition, added back in the Pentium Pro era).

- Hows that for innovation ?

Riveting stuff I say

 

[Surferosa]

Well, as hopefully our good friend will explain much better than I can- it is something to do with the Physical Address Extension that can be enabled in these server OS (which, I assume, are not available in standard Windows OS).

Large memory support

Edit: See ^, I knew he would

 

[TabrisDarkPeace]

What are the ****ing odds that is both my 1086 th post, and your 86 th post ? (x86 - get it).

Seriously the chances of that are something like 1 to 93,396 - assuming it was known you would post immediately after me.

 

[Mondoman]

If it is a "Windows XP limit" why do I have the same "problem" under Windows 2000 ?

It's also a limit in Win2000. By saying "it's a Win XP limit", I mean that it is a limit when running WIndows XP, and not (necessarily) when running other OSs. Other OSs, such as Win 2000, may very well impose the same limit. The REASON for imposing the limit was originally based on memory-address space limits imposed by the hardware, but most new computers have hardware that allows greatly-increased memory address space (both physical and virtual). If I run WinXP on such a system, the limit remains; thus, it seems reasonable to me to call it a Win Xp limit.

Fact is that it is a "32-bit address space limit, 12.50% to 31.25% (rarely more) of which is allocated to PCIe and PCI devices for accelerated I/O".

Yes, exactly.

...

Anyone passing themselves off as a PC tech, or enthusiast referring to it simply as a "Windows XP limit" should be shot on sight.

I respectfully disagree. We all use commonly-understood shorthand. "Windows XP" is commonly understood to refer to the Windows XP Home, Professional, and Media Center OSs. It IS a limit imposed by these OSs.

If I started a thread titled "Windows XP limit" and never mentioned a 2.75 GB to 3.50 GB 'ceiling' or 'wall' no one would have any idea what "limit" I was referencing.Could be processor core limit, or socket limit, or limit of allocated memory per Win32 process, file limit in root folder, or any number of other things

Could also be a limit on the number of backup copies allowed, a limit on the number of CPUs/cores, etc. I don't quite understand your point here.

Even implying that the limit is in any way associated (blame placed on) Windows XP is utterly wrong.

First, it is a fact that it is associated with Win XP, as it appears in Win XP.
Second, association does not imply blame.
No wrongdoing here.

It only acts as a catalyst to encourage smacktardism én mass. Providing IT assistance is hard enough without people changing rigid definitions into sloppy mutable definitions due to sheer laziness, ignorance ... [and so forth]

You forgot to add a few more, including the "decline of civilization and the dispossession of the Aborigines."

 

[little_scrapper]

Tabris just likes the sound of his own Im smarter than you voice. :roll:

Anyone who says someone should be shot for simply speaking incorrectly about minor technicalities/symantics should be shot.

 

[TabrisDarkPeace]

Any DDR2 will fit into a DDR2 slot......you would think thats a no brainer, but apparently not...

Sure it'll fit, but the question implied was "will it work on a Asrock 775Dual-vsta mainboard".

Consider most DDR2-800 people recommend around here runs at 1.9 to 2.2 Volts.

Not all mainboards support more than 1.8 Volts to VDIMM w/o problems.

It is for these reasons I'd heavily recommend 1.8 or 1.90 - 2.00 Volt DDR2-800 (eg: G.Skill sell some decent 'lower than extreme' voltage stuff that works on most mainboards w/o issues).

Anyone who says someone should be shot, for saying someone should be shot about simply speaking incorrectly about minor technicalities/symantics should be shot. - :lol:

Saying someone should be shot is forum slang, I don't mean it literally. (Forums, Internet, Anonymous, helping each other, and all that good stuff).

Since it would be best to use MSDN or www.microsoft.com searches:
http://search.microsoft.com/results.aspx?mkt=en-US&setlang=en-US&q=windows+xp+limit

Vs "Address Space Limit", which is technically correct regardless of OS being talked about, and gives accurate search results on both Linux and Microsoft / Windows sites: eg:
http://search.microsoft.com/results.aspx?q=address+space+limit&l=1&mkt=en-US&FORM=QBME1

I just wish people would stop 're-defining' industry terms to suite their own definitions, that's all.

 

[djgandy]

Aha, a damn fine question.

It is because of a CPU instruction (or set thereof) known as PAE-36.

It permits 36 bit addressing on 32 bit systems by paging huge chunks in/out of the 'active' memory pool.

The OS will hide this from applications, but it prevents on application (process) allocating a contiguous 6+ GB chunk of memory, as it would need to span multiple paged 'pools'. (Most Win32 processes can only use 2 GB, some using /3GB in BOOT.INI and special coding technique can use up to 3 GB per process).

Some applications will load more than one process into memory though, and use inter process communication to work around it.

PAE-36 has existed since the Pentium Pro (considered the 6th generation of Intel CPUs), PAE-36 is also used with the NX/XD bit to mask areas of memory as NO EXECUTE / EXECUTE DISABLE on systems with 4 GB or less or memory in Windows XP (Service Pack 2 or higher required, however a CPU does not need the NX/XD bit for DEP to work, as PAE-36 offers a software workaround to do the same job).

NEW: http://www.microsoft.com/whdc/system/platform/server/PAE/pae_os.mspx

http://search.microsoft.com/results.aspx?mkt=en-US&setlang=en-US&q=PAE+36

http://search.microsoft.com/results.aspx?q=%2F3GB+BOOT.INI&l=1&mkt=en-US&FORM=QBME1

http://search.microsoft.com/results.aspx?q=AWE&l=1&mkt=en-US&FORM=QBME1

http://search.microsoft.com/results.aspx?q=%2FNoExecute+boot.ini&l=1&mkt=en-US&FORM=QBME1

http://search.microsoft.com/results.aspx?q=Data+Execution+Prevention+&l=1&mkt=en-US&FORM=QBME1

- Anyone training IT people should bookmark the above links, Or if you are want to become a Software Developer for a 'certain grade' of software (although x64 has a flat 512 GB to 1024 GB [1 TB] memory model - it is still worth reading).

- It is pretty cool stuff, but PAE-36 has been around since the early days of Windows NT 4.0 Server, and is why many '32 bit Kernels', such as Novell Netware 6.5 SP3, can use pools of memory that span 36 bits of address space normally (but not with a flat memory model).

Past 32 GB PAE-36 is pretty inefficient though, despite the fact it can handle 64 GB 'in theory' w/o major performance issues arising.

PAE-36 is also the reason why Service Pack 2 enables a software only version of No Execute / Execute Disable (Data Execution Prevention) using /NoExecute in BOOT.INI on systems with CPUs that lack the NX/XD bit. (All thanks to an underused instruction addition, added back in the Pentium Pro era).

- Hows that for innovation ?

Riveting stuff I say

Heh i remember this argument in another thread heh. Some fool was arguing whats the point in 64bit when we have pae. He failed to understand that PAE is something that is transparent to user mode applications and that a single process can still only address memory within the width of the cpu registers.

You're correct though.
Most programs can only address to 0x7FFFFFFF. Above 0x80000000 is kernel space.
You have a small chunk for stack too.

 

[TabrisDarkPeace]

:?: Why are there so few people like you on my buddies list ?

And where do you all hang out ? (A Hardware forum isn't exactly my first guess, but TomsHardware Forumz is still pretty International, I need a forum to hang while I learn C++ and Win32 / Win.x64 application development).

There isn't much left for me to gain in hardware, only to try and help other people out (for free), but as scarce as (absolute technical) truth is around here at times, the supply has always been in excess of the demand. One decent sized How To would cover 99.7% of this forums questions IMHO, thus it would be better off to make the HowTo into a piece of software that people can just run. (As complex as it is, I might just have a simple solution to it).


Coding for a platform like it has no limits is what got software in the ****hole that it is in today.

If the operational ceilings were more publicised then more developers would appreciate it from the get go.

(That's the summary of my theory on it anyway).

So where do people like you hang out on the Internet ?, I need to emmigrate from TomsHardware eventually.