Jump to content

Windows XP: new Z68/Z77 rig in 2017


Tomcat76

Recommended Posts

I'm looking to upgrade my current computer, which has a motherboard powered by a P55 chipset and the Core i5-650 processor.  This computer has been unstable since I had it (system clock running too fast, frequent program freezes and crashes and the occasional memory error dialog), using Windows 2000 and later Windows XP.  Memory and hard disk diagnostic tools never found a problem.
 
I have thought about this for a long time, and concluded that it would be best if I split my workload onto two computers instead of one.  The first box would be a modern PC running a modern version of Windows, which will be used for video editing and storage (manual backups) exclusively.  The second box should be a high-performing "legacy" computer running a dual boot of Windows XP 32-bit and 64-bit, which I will use for basic tasks.  Unfortunately, "basic tasks" in this day-and-age have become quite hardware-demanding: H.264 decoding, browsing Facebook, etc.
 
It's this second box I'd like some input on regarding the hardware components that I'd use best.  I realize that it's possible to just take a Haswell or Skylake board and use generic drivers, but something tells me I can get better performance using "official" drivers as they are specifically written for individual hardware devices.
 
After some digging around and weighing specs against one another, I have already limited my search to the Z68 and Z77 chipsets, part of the LGA1155 socket range which is the most recent with official support for Windows XP.  The Sandy Bridge and Ivy Bridge CPUs fall into this range.
 
From what I can see, the only real difference of importance to me between the Z68 and Z77 chipsets, is the inclusion of native USB 3.0 by Intel on Z77 boards.  Since there are no Windows XP drivers for this, choosing a Z77 setup will limit me to motherboards that have an extra USB 3.0 chip by another manufacturer such as ASMedia.  But I don't like the thought of working on a system on which some things don't work normally, or need to be disabled in the BIOS or in Device Manager.
 
Question 1: the native USB 3.0 problem aside, are there any benefits for me to still pick a Z77 motherboard?
 
I already went ahead to find more information on the Z68 chipset, which raises a few more questions.  There appear to be two main versions of it: the "standard" version and the "Gen3" version, the latter supporting PCIe 3.0 if used in combination with an Ivy Bridge CPU.  Another advantage of Ivy Bridge is that it has improved onboard graphics (Intel HD 4000) compared to the onboard graphics of Sandy Bridge (Intel HD 3000).  On the other hand, I have read posts on forums stating that Ivy Bridge CPUs are running hotter than the Sandy Bridge CPUs.  Silence is an important factor for me as well.
 
Graphics-wise, I want:
- smooth playback of 1080p uncompressed video
- hardware-accelerated 1080p H.264 decoding (online videos on Youtube, etc.)
- smooth operations on multi-layered PSD images in Photoshop 7
 
Question 2: would the onboard graphics of the Ivy Bridge CPU (Intel HD 4000) be powerful enough to handle that, or is a discrete graphics card still required?
 
I have also done some digging with respect to graphics cards, as nVidia (my preferred manufacturer) has dropped Windows XP support somewhere along the line as well.  The newest supported cards are the 700 series as well as the GTX 950 and 960-based cards.  If a separate graphics card is still required, I was thinking of one of the following:
- Asus STRIX-GTX750TI-OC-2GD5 (19.6cm ~ 7.71in / 640 CUDA)
- Asus STRIX-GTX750TI-DC2OC-4GD5 (19.6cm ~ 7.71in / 640 CUDA)
- Asus STRIX-GTX950-DC2OC-2GD5-GAMING (22cm ~ 8.66in / 768 CUDA)
- Asus STRIX-GTX950-DC2-2GD5-GAMING (22cm ~ 8.66in / 768 CUDA)
- Asus STRIX-GTX960-DC2OC-2GD5 (21.5cm ~ 8.46in / 1024 CUDA)
- Asus STRIX-GTX960-DC2-2GD5 (21.5cm ~ 8.46in / 1024 CUDA)
 
I chose Asus STRIX boards because their fans don't spin below a certain point (65°C if I'm not mistaken).  All of these are marketed as PCIe 3.0 cards, but I have read on various forums that there is practically no difference with PCIe 2.0 in terms of gaming performance.  I don't do games, but does that mean they will still work as spec'd if I put them on a PCIe 2.0 slot on a "standard" Z68 motherboard?  I might save some bucks there if "Gen3" boards are more expensive.
 
Lastly, as far as the motherboard goes, I have the following "prerequisites":
- 1 PS/2 port for a keyboard
- 1 PCI slot, preferably at the bottom
- 4 SATAIII 6Gbps ports and 2 SATAII 3Gbps ports (requires additional controller as Z68 only supports 2 SATAIII ports)
 
I'd prefer not to have more than 1 or 2 PCI slots as they are usually positioned in such a way that not all PCIe ports can be utilized in conjunction with the PCI ports.  If I will ever need another extension card of any kind, I plan on going PCIe all the way.
 
These are the boards I had in mind:
- Asus P8Z68 Deluxe / Gen3
- Asrock Z68 Extreme4 Gen3
 
The Asus is a bit overkill with the dual LAN and Bluetooth features, but their other boards in the Z68 series don't offer everything I need.  Long-term durability is also a concern as the PC will be left turned on 24/7.  How can I tell if a motherboard is better or worse equipped for continuous operation?
 
Anyone have any thoughts on this?

Link to comment
Share on other sites


I have an Asus P8Z68-V LX motherboard (made in 2011) with an i7-3770k on it. There are drivers for everything on it for XP, and the ASMedia USB 3.0 also has drivers for XP and all works great. I use a ginormous Zalman CNPS9900DF cooler and 3 Noctua fans (2x NF-B9 [92mm] + 1x NF-P12 [120mm]) and an OCZ ModXstream Pro 600 power source (600W), and one can barely hear the machine at all, when switched on (and the cores run below 35°C, on idle and below 50°C, upon heavy use). I'm fully satisfied with the onboard video, but that may not be of much help for you, because I'm no gamer. However I think the onboard video is good enough to do what you require satisfactorily. In general, Asus is a better hardware line than Asrock.
In any case, the standard disclaimer applies: it works great for me, but YMMV.
Warning: The board I bought came with the original BIOS version, so I had to borrow a very basic Sandy Bridge chip just to get it to boot (it was no go with the i7-3770k, at that point), then I updated the BIOS to the latest version available, and then I was able to switch the processor to the i7-3770k and it has been working great since then.

Link to comment
Share on other sites

Yeah, things can become a little tricky with the BIOS.  I assume that the "Gen3" boards are shipped with a BIOS that supports Ivy Bridge, otherwise there wouldn't be much point in labelling them "Gen3".  Still something to watch out for, though; I don't want to buy an extra CPU that I won't be using.
 
Some more digging around the internet seems to confirm your claim that the HD 4000 graphics are probably enough for what I intend to use that computer for.  Several sources also claim that the HD 4000 graphics can be roughly compared to an nVidia card with 48 cuda cores.  My current GT 610 card also comes with 48 cuda cores so I can always rely on that should the onboard graphics not be sufficient after all.
 
I just realized I also had a question concerning the RAM.  I want to use 8GB, but am doubting between 2 sticks of 4GB and 4 sticks of 2GB.  If only XP 64-bit was in the frame, I'd go with 2x 4GB; but I also have to consider XP 32-bit.  I don't know how a motherboard's memory controller handles the RAM if an OS can't use all of it.  If it disables the sticks that are above the memory limit, I'd think that having 4 sticks of 2GB would be better as it will still allow 2 sticks to be used by XP 32-bit; in the case of 2 sticks of 4GB, only one stick would remain.  Or isn't that how memory controllers work?  If it will just give me a percentage of each stick, then it's probably better to go with 2x 4GB.

Link to comment
Share on other sites

I doubt anyone could really notice the difference but, then again, I can see no reason not to populate all available slots, so I'd go with 4x 2GiB. The Asus BIOS has a quite reliable memory hoist, so that all the memory not seen by XP (my XP actually sees 3.22GiB RAM) will be moved by the hoit above the 4 GiB address, where it can be accessed using PAE. Now, instead of going for the unreliable kernel PAE patch, the best thing to do is to create a 4.78 GiB Ramdisk with that memory, which can reliably be used for a pagefile and a dumping place for "Temporary Internet Files" and other browsers caches (as well as for cookies, too, if desired). Of course this is just my 2¢...
 

On 4/24/2017 at 6:13 PM, dencorso said:

Then how do you explain I have a working 12.4GiB RamDisk on XP SP3, on a machine having 16GiB?
You just need this version of the Gavotte and to set these values in the registry:
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management]
"DisablePagingExecutive"=dword:00000001
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\RRamdisk\Parameters]
"UsePAE"=dword:00000001
and these command-line options to boot.ini (in case you want PAE but not DEP, else the default is OK):
/noexecute=alwaysoff /fastdetect /pae
Believe me: this is quite old news. :yes:

On 4/26/2017 at 11:32 PM, dencorso said:

@TrevMUN: Well, on a Ramdisk of at least 4GiB (say, in a machine with 8 GiB, of which I have a couple), the Gavotte can be the host to the pagefile and XP is happy with it, provided one sets "DisablePagingExecutive"=dword:00000001 as I posted above (to prevent the dreaded "Page Fault on Non-Paged Area" BSOD). It'll work equally well on x86 7 SP1 and Vista SP2, too. I don't know about 8+.

 

 

 

Link to comment
Share on other sites

There are some interesting reads online about ramdisk.  Gavotte and ImDisk seem worth the try.  Too bad I only have 4GiB ATM so I can't test them in depth.
 
I take it you'd rather disable the pagefile than to put it on the ramdisk.

Link to comment
Share on other sites

4 hours ago, Tomcat76 said:

There are some interesting reads online about ramdisk.  Gavotte and ImDisk seem worth the try.  Too bad I only have 4GiB ATM so I can't test them in depth.
 
I take it you'd rather disable the pagefile than to put it on the ramdisk.

But you'd still be able to use the Gavotte as a place to dump things you'd like to disappear, like the "Temporary Internet Files". However, with a smallish Ramdisk, you'd have one possible issue: if you download a file bigger than the Ramdisk size, it'll fail silently. So that you'd have to remember to move the "Temporary Internet Files" before using IE to download a file bigger than the Ramdisk or, much better, never to download such a big file using IE (other browsers and wget allow one to decide where the download should go, without 1st bringing it forcibly to the "Temporary Internet Files", then moving it to some indicated destination, like IE does).

Link to comment
Share on other sites

8 minutes ago, dencorso said:

... much better, never to download such a big file using IE (other browsers and wget allow one to decide where the download should go, without 1st bringing it forcibly to the "Temporary Internet Files", then moving it to some indicated destination, like IE does).

Or - possibly much-much ;) better - don't use IE at all.

Recent news, JFYI:

https://hotforsecurity.bitdefender.com/blog/internet-explorer-bug-can-reveal-the-contents-of-your-address-bar-19020.html

https://www.brokenbrowser.com/revealing-the-content-of-the-address-bar-ie/

jaclaz

 

Link to comment
Share on other sites

2 minutes ago, jaclaz said:

Or - possibly much-much ;) better - don't use IE at all.

That's impossible... unless one forgoes using MS Update/Windows Update entirely. :dubbio:

4 hours ago, Tomcat76 said:

I take it you'd rather disable the pagefile than to put it on the ramdisk.

Not if your machine has 8 GiB RAM and is running XP x86... you either use the Ramdisk (and then there's no reason not to put a system managed pagefile inside it) or let ca. 4.7 GiB RAM otiose, to use a word jaclaz likes... ;)

Link to comment
Share on other sites

32 minutes ago, dencorso said:

Not if your machine has 8 GiB RAM and is running XP x86... you either use the Ramdisk (and then there's no reason not to put a system managed pagefile inside it) or let ca. 4.7 GiB RAM otiose, to use a word jaclaz likes... ;)

My thought exactly after reading the article by tompsonn.  A 64-bit XP would have all the RAM at its disposal, so using a ramdisk would just utilize RAM that was already available to the system.  As such, it isn't useful to host a page file in it.  However, in the case of a 32-bit XP when used with a ramdisk tool that uses the unused RAM portion, nothing is taken away from the system so I don't see a problem in putting the pagefile in there.

Link to comment
Share on other sites

 

56 minutes ago, dencorso said:

Not if your machine has 8 GiB RAM and is running XP x86... you either use the Ramdisk (and then there's no reason not to put a system managed pagefile inside it) ...

Sure, but ... 

19 minutes ago, Tomcat76 said:

However, in the case of a 32-bit XP when used with a ramdisk tool that uses the unused RAM portion, nothing is taken away from the system so I don't see a problem in putting the pagefile in there.

A pagefile that will likely never be hit, so a largely otiose use anyway.

Seriously, if the programs you use (or your particular usage of the system) consistently hits the pagefile, then you will be much better served by an operating system allowing more "direct" RAM, even if Imdisk or Gavotte are fast (faster than a hard disk and possibly even faster than a SSD) they are much slower than direct RAM access.

On the other hand, if the programs you use (or your particular use of the system) never *need* a pagefile you are introducing a (remote possibility as over the years both Imdisk and Gavotte's has been demonstated as being very stable) an unneeded possible point of failure.  

jaclaz

Link to comment
Share on other sites

21 hours ago, jaclaz said:

On the other hand, if the programs you use (or your particular use of the system) never *need* a pagefile you are introducing a (remote possibility as over the years both Imdisk and Gavotte's has been demonstated as being very stable) an unneeded possible point of failure.

I) Provided one sets "DisablePagingExecutive"=dword:00000001, in my experience, the chance of failure for any reason changes very little due to this setup. In fact, XP is really hard to crash, unless one runs experimental or undebugged programs or uses a program deliberately written to cause a crash (the most common was written by Matt Pietrek, IIRR, but there are many others... and other ways, but that's beside the point).
II) Suposedly, when a program is first loaded and converted into a process image able to actually run, if there is a pagefile, that should be done directly on the page file, and then just some pages are brought into the actual frames of the real memory and let run or a quantum... at least that's the theory of virtual memory, IIRR... In case I'm right, then the pagefile is always hit, in a way, provided it exists. Then again, even if I'm wrong, setting a pagefile in such a Ramdisk should be harmless at worst and beneficial at best (quod abundat non nocet). Remember there are some programs that do require a pagefile to run (although I cannot provide an example of such a program from the top of my head, at this moment).

Link to comment
Share on other sites

@dencorso

As said it depends on programs used and on specific use of the programs,

@Tomcat76

Photoshop is often mentioned as an example of one that *requires* the pagefile, but AFAIK/AFAICR it is only a (stupid) "cold" requirement, i.e. you *feed* it with a 100 Mb pagefile and it will start just fine, it will simply refuse to run without a pagefile, which does not automatically mean that it *needs* a pagefile, the difference is very slight, but it is there alright.

The general point that I was trying to make is that if one runs the "normally" installed system in his/her "normal" way, using the programs "normally" used in the "normal" way he/she uses it and monitors pagefile usage for a few days, the decision on size (and *need*) of a pagefile can be taken correctly, based on actual data.

And I will repeat myself - if needed - we are not talking of an alien OS just landed on Earth from an outer galaxy.

We are talking of a 16 years old OS, that has been consistently run for years by millions of people with "automatic pagefile setting", which means - according to Mark Russinovich:

https://blogs.technet.microsoft.com/markrussinovich/2008/11/17/pushing-the-limits-of-windows-virtual-memory/

Quote

You’ll notice that the default configuration is for Windows to automatically manage the page file size. When that option is set on Windows XP and Server 2003,  Windows creates a single paging file that’s minimum size is 1.5 times RAM if RAM is less than 1GB, and RAM if it's greater than 1GB, and that has a maximum size that's three times RAM.

Now, it is evident (at least to me) how the algorithm used is - to say the least - approximated, only a little more sophisticated than a generic (and "safe") rule of the thumb, and since the minimum size is required only by the *need* for enough size for a full crashdump the lower limit is in itself m00t.

If we assume that the formula is correct (reductio ad absurdum) it plainly means that a 4 GB RAM system with a pagefile of - say - 512 Mb has (roughly because of the known issue with reserved unused memory under 4 Gb) the same total capacity as a 1 GB RAM system with a fully committed/used 3 GB pagefile or as a 2 GB RAM system with a pagefile expanded up to 2 Gb or DOUBLE that of a 512 Mb machine with a fully committed/used 1.5 GB pagefile .

The above (machines with 512 Mb or 1 Gb or 2 Gb of RAM running XP) represent probably 70% (or more) of the hundreds of millions machines that ran XP in the last 5-10 years of life of the OS.

 Yet the reported crashes are not so common, it would reasonably mean that the provision is "big enough".

What I personally suggest is - when the "full" 4 Gb of RAM are present to make a variable sized pagefile, range 100 Mb to 4 Gb, then run the system normally for a few days, checking how much the pagefile expanded.

Once determined this expanded size, make a fixed size pagefile of corresponding size and be done with it. (and of course, you can well move it to the "past 4 Gb" Ramdisk if the size is suitable).

If the size did not grow, try running without a pagefile for a few more days, if everything works as before then you don't *need* a pagefile.

As said IF there is the need of a pagefile, AND the resulting size is in the range 1-3 times the 4 Gb RAM, it means more or less that you would be better served by a 64 bit OS.

jaclaz

Link to comment
Share on other sites

Well, now that we've beaten the poor dead horse into beef tartar, let me give one more unrelated sugestion:
@Tomcat76: give the machine a smallish, fast SSD (a 120GB OCZ Vertex, or any other SATA III OCZ from before takeover by Toshiba would be great, but do avoid the older Octane series) for OSes and one bigger HDD (I suggest a WD2002FAEX or any WDX *FAEX, which is the last 512-byte sectored series from WDC and was findable new/used on eBay till quite recently) for data, and you're all set, IMHO.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...