Performance-oriented Windows tweaking

[jeh]

Agreed, EXCELLENT work. In fact I joined anandtech just to say so!

I'll bet a lot of sites end up linking to this thread.

deathkoba> However, it DOES reduce the amount of memory the system uses

Well, not necessarily even that. The memory "used by services" (and the file cache, and every other process or set of processes) is a highly dynamic quantity. If nothing else is placing demands on RAM, the services will pretty much be left as they are after startup (and even these sizes depend greatly on how much RAM is on the machine).

But that isn't a typical situation, and if it is your situation, who cares? You don't need the RAM for anything else in that case.

otoh if you do run some apps that make more demands on RAM than can be immediately satisfied, the idle services procs will have their working sets ("mem usage" in task manager) shrunk (just as with any other process).

Unless of course the service process is busy too... but if it's busy, it's presumably doing something for you, so you don't want to disable it anyway.

Disabling services to save memory does make the memory available earlier, without waiting for the OS to shrink any working sets... but that's a tiny benefit. If you see it at all you'd see it in app startup times, not in "operational" performance of the apps.

 

[okb]

There are much more useful ways to speed things up, like disabling all the bells and whistles of Windows. There was a time when I thought service "tweaking" was helping me out a lot. That was years ago. Now I'm more interested in fine-tuning useful things, like my FF settings, memory timings and shrinking my startup list to a minimum.

I have a hard time believing (as some would have us) that BV is out to corrupt the masses. Yeah his site is pretty much to stroke his own ego, but there's a few key things some people have overlooked (or ignored). 1) He's not in IT, even says so (read his FAQ), he's an electronics tech..whatever that means. Could install radios at Future Shop or whatever, doesn't matter. 2) The people who blindly follow his advice are the ones who'd blindly follow anyone else who sounds like they know what they're talking about without doing an ounce of research. If it's not BV they're listening to it's someone else, be thankful BV at least has some disclamers and warnings.

Anybody who really wants to understand the best way to tweak/tinker/modify or otherwise change their computer settings will go to a forum such as this. IMO web pages can only do so much as there are too many variables to have it condensed into a few measily html pages. Static text has nothing on open discussion. I can't tell you how many times a potential problem or completely unrelated issue has been solved/avoided thanks to open discussion. Now granted things like security settings, bandwidth maximization etc can be pretty easily slapped up on a page for all to see, but in no way would I (now) dig deeper into my system in areas I'm not familiar with based solely on what some html tells me. But then again I'm not one of the lemmings who blindly follows the current flavor of the month.

Stop giving him such a hard ride, he doesn't really deserve that much credit for the woes of the IT industry and his detractors, likewise does he not deserve the uneducated praise of his staunch supporters. Good to see there are any middle-ground voices of reason. Boycott his page? Sure, online we talk with our mouses (mice?). The funny thing about the Internet is it seems to almost have made the masses dumber, with all the information we could ever dream of at our finger tips we should all be computer gurus, though due to it's nature there are equal parts bad information. Unfortunately the bad is often easier to come by. God bless search engines. I gave up trying to save the world a long time ago and grudgingly accepted that there are people who won't believe you no matter how many references you give them, most however, will be amenable eventually. Patience goes a long, long way. They are still lemmings after all, lemming-ing is hardwired into them.

One final thought, based solely on personal anecdotes with zero factual basis. Windows seems to do a very good job now of managing it's resources, maybe that's due to the level of technology in the hardware, but I think it's also due to MS finally getting it right. We rail on MS pretty hard, and a lot of it's deserved and a lot isn't, so be it. It's still a good OS and unless you're an absolute power l33t h@x0r there's absolutely no reason to do much of anything to Windows. XP is the single best offering since 98SE. 98 will always have a soft spot for me as it got all the things right that 95 was supposed to do. Yeah it took a billion reboots to set up, so what. It was worth it. NT was a solid beast that I prefer to 2K even now (2K to me is more of a half-way house bridging NT and XP, especially the home versions). XP may have a lot of problems, but they also got a lot of things right.

 

[JackMDS]

Originally posted by: okbIf it's not BV they're listening to it's someone else, be thankful BV at least has some disclamers and warnings.

How True :thumbsup:

And mean time there is a PSU issue that has a sticky in the General forum, and much of the content there is conceptually similar to the BV claims.

Next year there would be a Thread like this proving, that the expensive PSU is doing nothing to make your system better as well.

:sun:

 

[n0cmonkey]

Originally posted by: JackMDS

Originally posted by: okbIf it's not BV they're listening to it's someone else, be thankful BV at least has some disclamers and warnings.

How True :thumbsup:

And mean time there is a PSU issue that has a sticky in the General forum, and much of the content there is conceptually similar to the BV claims.

Next year there would be a Thread like this proving, that the expensive PSU is doing nothing to make your system better as well.

:sun:

I agree with you to a point, although a _quality_ PSU goes a long way in keeping a machine stable. And quality is typically a bit more expensive.

 

[Slikkster]

Interesting results, and worthwhile to post. Good work. That said, I think it's worthwhile to say that this was a very limited test on very dated hardware. While it may indicate what could be expected on any PC, no matter its hardware, I think that remains to be seen.

Notwithstanding BV's tips, I hope FD isn't trying to imply that no manner of tweaking XP services/applets/configs would improve performance. I think MS has done a fine job with XP, by and large, but the "one size fits all" default setup can't possibly create the best performance for each individual machine. That's just common sense. When using older video cards like a TNT's, I'm not sure any tweak of the OS is going to help once the card's own processing power is maxed out. Results might be different on a high-end system with a high-end card. Note key word "might".

Best I can tell, this was a 2-PC test. While it's useful to show the results on these boxes, I'd like to see a much, much wider survey done in a controlled environment on a wide variety of test platforms, including high-end systems. I'm sure Fresh Daemon's an up and up guy, but let's face it, this is a forum, and anyone can say anything here. I'd like to see some place like PCMag, ExtremeTech, Anand, HardOcp, Tom's, etc., put their personal integrity on the line testing these tweaks, OR tweaks of their own suggesting. I think there would be a greater sense of credibility and acceptance if these guys did their own controlled tests, just like they test new hardware.

If that's already been done, I plead ignorance and apologize in advance. But since this thread seems to be the first to directly take on BV (and others who seemingly just regurgitate the same info), I'm going to assume it hasn't been done.


Bottom line is that I'd like to see a comprehensive test on a wide variety of machines that painstakenly looks at the value of tweaking default settings of this service or that service, as well as other user-configurable settings to the OS. That would be of the most value to me.

 

[okb]

Slikkster, good points. But on the flip side isn't this the same thing BV is doing? I'm not overly familiar with his methods or his test bed(s) as I've never given his site more than a cursory browse but I doubt he's done what you're suggesting. And I certainly didn't see any benchmarks other than a couple for seeing how the swap file affected Unreal Tournament 2K3. IMO FD's results lend more credibility and BV's are simply, well he shows no results which I think speaks volumes. Should FD's experiment be taken as definitive? Hardly, but he's taken a big step in providing solid evidence that BV's so-called "Super Tweaks" are nothing more than crippling modifications. And let's be honest, if we thought they would speed up our systems 99% of Anandtech-ers would be using them, don't say you wouldn't.

A question though, how many systems do PCMag, ET, Anand et al use when they benchmark a new piece of software/hardware? I'm not familiar with their testing policies.

 

[Nothinman]

That said, I think it's worthwhile to say that this was a very limited test on very dated hardware.

That's the point, if the changes don't make a noticable difference on older hardware the effect will be even smaller on newer hardware.

I'm sure Fresh Daemon's an up and up guy, but let's face it, this is a forum, and anyone can say anything here

So do some tests yourself, if your results are that different from his say something. I would be willing to bet money that they won't, though.

 

[SoulAssassin]

One thing you failed to test was startup times, I wholeheartedly agree that these tweaking guides lead people to falsely believe they're properly configuring their system but it would be nice if you reran these tests and showed the time from POST to desktop. Keep in mind that XP starts alot of services after it's shown the login screen so it's probably a tough (but fair) thing to measure.

 

[Fresh Daemon]

I have a hard time believing (as some would have us) that BV is out to corrupt the masses.

"Never attribute to malice that which can be explained by stupidity." -- Napoleon Bonaparte

And mean time there is a PSU issue that has a sticky in the General forum, and much of the content there is conceptually similar to the BV claims.

Agreed, enthusiasts pick PSU's way in excess of their needs. Now I have a 535W Enermax, however, DFI doesn't give any support unless you have at least 480W and I found the 535W on sale for less than a ~480W, so I plead not guilty.

SPCR has good threads and articles on PSU sizing. I think the misconception is that when nVidia says "you need 550W for SLI", they mean 550W from a crappy PSU, which will fluctuate and not deliver good amps. They could say "you need a 400W PSU for SLI from a good manufacturer" but then they would have to start defining a good manufacturer, and for nVidia to start endorsing or condemning various third parties is going to be a legal nightmare.

Not that it's really relevant, but that's my understanding.

That said, I think it's worthwhile to say that this was a very limited test on very dated hardware.

What nothinman said!

Look at it this way. Shave 20MB off a 160MB box and you've increased available RAM by 33%. Shave 20MB of a 2GB box and you've gained 1%. Which has a bigger impact on performance, 33% or 1%? I doubt you could even measure the difference on a 2GB box if there was one. What is BF2 going to do with 20MB? Is that even enough to load a texture?

I think MS has done a fine job with XP, by and large, but the "one size fits all" default setup can't possibly create the best performance for each individual machine.

I don't know who told you that XP was fully monolithic, but they were wrong. It's highly modular and dynamic. It will tweak and configure itself for the hardware it is installed on.

When using older video cards like a TNT's, I'm not sure any tweak of the OS is going to help once the card's own processing power is maxed out.

Firstly, I deliberately ran at lowest possible settings so as to not max the card out. Secondly, you can see that the card wasn't maxed out by observing the performance hit when all the RAM was used up.

One thing you failed to test was startup times

If there is enough interest I can return and time them. However, I boot my rig once per day at most (if I'm getting something big on bittorrent I will leave it on for a few days to a week without rebooting), so I didn't think it would be that important to people.

 

[fbrdphreak]

Awesome :thumbsup:

 

[JackMDS]

Originally posted by: Fresh DaemonAnd mean time there is a PSU issue that has a sticky in the General forum, and much of the content there is conceptually similar to the BV claims.

Agreed, enthusiasts pick PSU's way in excess of their needs. Now I have a 535W Enermax, however, DFI doesn't give any support unless you have at least 480W and I found the 535W on sale for less than a ~480W, so I plead not guilty.[/quote]LOL I am more guilty than you I use 650W PSU.

Why 650W? Coz it is the same size, same cables., and power supplies do not take more electricity if they are rated high, since they deliver according to need.

So I got a 650W power supply with two Fans 20/24pins mobo plug and SATA, for $29.

It works very well for few month very stable with a DFI NF4, overclocked Venice 3200+, x64 with fast PCIeX.

Similar to the services issue that you researched so well, a $100 (and more) power supply does not provide any more than my $28.

:sun:

 

[n0cmonkey]

Originally posted by: JackMDS

Similar to the services issue that you researched so well, a $100 (and more) power supply does not provide any more than my $28.

Except quality. While you can probably diagnose PSU issues, it's an ongoing struggle here on Anandtech forums. I didn't realize how many stability issues can be caused by crappy powersupplies until I followed mechBgon around the tech support and GH forums.

 

[Sunner]

Originally posted by: n0cmonkey

Originally posted by: JackMDS

Similar to the services issue that you researched so well, a $100 (and more) power supply does not provide any more than my $28.

Except quality. While you can probably diagnose PSU issues, it's an ongoing struggle here on Anandtech forums. I didn't realize how many stability issues can be caused by crappy powersupplies until I followed mechBgon around the tech support and GH forums.

This one I agree on.

 

[JackMDS]

Originally posted by: Sunner

Originally posted by: n0cmonkey

Originally posted by: JackMDS

Similar to the services issue that you researched so well, a $100 (and more) power supply does not provide any more than my $28.

Except quality. While you can probably diagnose PSU issues, it's an ongoing struggle here on Anandtech forums. I didn't realize how many stability issues can be caused by crappy powersupplies until I followed mechBgon around the tech support and GH forums.

This one I agree on.

I use to design (many years ago) Power supplies for Aviation communication equipment (a much more complicated issue than Gaming Computers). I know the stuff, at that time it was much more complicated than today PSUs since the regulation Integrated Circuit were more primitive than today chips.

I do not want to get here to a big analysis of Loads, Ripple, stability, Noise, etc., and what part of it can be compensated by using a higher Wattage power supply that cost just few $ more.

My statement is related to the general issue of the thread. I.e. doing things like switching off Services, or buying very expensive PSUs without actual technological proof that it makes a difference.

I am not saying that there are No crappy PSUs. I just do not think that the inflated prices of the popular units are justified. It base on stories and induced fear into people that do not really understand the technology.

:sun:

 

[Fresh Daemon]

C'mon guys, off-topic or what!

Anyway, there should be some updates soon. I'm going to put the screenshots back, find a better way to benchmark a swapfile, and run a wider variety of benches. System II is out of commission, it had an IBM Deathstar that did what they usually do. Now it has a 15GB Seagate, which isn't enough room for a separate install (and there's no way I'm going to subject my primary install to BV's horse puckey), but System I still works.

Anyone want to donate me a small hard drive in the name of science?

 

[xtknight]

Originally posted by: JackMDS
I am not saying that there are No crappy PSUs. I just do not think that the inflated prices of the popular units are justified. It base on stories and induced fear into people that do not really understand the technology.

Very true. The 40kW (IIRC) power supply my brother built from scratch for an electric car didn't cost very much. It was only, say, 8x bigger too. I also hate people who pass off a 500W PSU as "just enough" for a system that in reality would use maybe 250W and no more. While I'm not as keen on electricity as my brother, I do understand the 500W doesn't do 500W all the time. Still, it would do like 380W continually maybe. My old PC (Video Rig) used 170W (P4 2.6 GHz, Radeon 9500 PRO) under full load, and it ran on an Antec 300W. Unsure of the draw of my current one (Gaming Rig).

 

[dclive]

I recently looked at a Celeron 450 I built years ago. 125W PS.

The Compaq AMD 3000+ XP I have (S6000Z) is something like 185W, and houses an ATI Radeon 9700 Pro, 2 optical drives, and 2 hard drives without issue.

A clone box with AMD 3200+ XP (well, an overclocked Athlon-M 2500+ that's running at 2300 actual mhz), a SCSI RAID PCI card, an IDE PCI card, 6 hard drives including 2 10,000 RPM SCSI drives, 2 optical drives, an ATI Radeon X800 XT PE, and 2 512M sticks of RAM get along wonderfully on a plain, generic 300W power supply.

I think the entire power supply business is absurdly overblown.

 

[mechBgon]

Originally posted by: dclive
I recently looked at a Celeron 450 I built years ago. 125W PS.

The Compaq AMD 3000+ XP I have (S6000Z) is something like 185W, and houses an ATI Radeon 9700 Pro, 2 optical drives, and 2 hard drives without issue.

A clone box with AMD 3200+ XP (well, an overclocked Athlon-M 2500+ that's running at 2300 actual mhz), a SCSI RAID PCI card, an IDE PCI card, 6 hard drives including 2 10,000 RPM SCSI drives, 2 optical drives, an ATI Radeon X800 XT PE, and 2 512M sticks of RAM get along wonderfully on a plain, generic 300W power supply.

I think the entire power supply business is absurdly overblown.

The times, they are a-changin' link to DFI Street forums

DFI no longer will support customers who do not adhere to the minimum power supply requirements.

if the board has a 24-pin power connector, then you must have a true 24-pin power supply.

not a 20-to-24 pin adapter

not a 5,000,323watt 20-pin with 80000 amps on the 12v rail

a true 24-pin native power supply.

period.

there is no argument anymore on this subject.

if the board has a 24-pin power connector, and we state (as we have) that you need a true 24-pin power supply and that a 20-to-24 pin adapter does not work and is not supported, then you need a true native 24-pin power supply.

no arguments

no trying to tell us we dont know what we are talking about.

period

you only have a 20-pin psu? a modded 20-pin psu? a 20-pin psu with a 24-pin adapter?

you are not adhering to the minimim specs required by these boards to run. this means that any problems you have while not adhering to the minimum requirements will result in you being told 'you need a 24-pin native power supply of at least 480w' and that is the only answer you will receive on your problem until you meet the minimum requirements.

sounds harsh?

we don't tell you to put a 24-pin power supply in just to make you angry and see you complain about how you have to upgrade your power supply.

we tell you that you are required to have a 24-pin 480w power supply because that is what is necessary.

if you want to argue about it, by all means continue arguing about it.

but you are not meeting minimum requirements and will not get any answer other than 'you MUST have a 24-pin native 480w power supply to run these motherboards and that is your problem so please upgrade to a minimum 480w 24-pin power supply'


not much else to say on this subject.

i'm tired of some of you who think you know more than our engineers continuing to argue that we have no clue what we are talking about and you can skirt the minimum requirements.

follow the recommended guidelines.

After the first 10,000 tech support posts or so, I saw a pattern develop... there is a higher incidence of cheap or generic or low-wattage PSUs in the computers that the owners are asking for help with. No, I don't hear about the ones that don't have problems. But when I do, if it looks like a case of cheapitus maximus and troubleshooting doesn't lead to any other conclusions, I suggest a reputable brand with ample wattage, and what do you know... yeah, that is often the answer. Some brands are worse than others, of course.

/ 2¢ worth

 

[dclive]

A frustrated dfi tech can post whatever he wants to a forum post as a random user. That doesn't make it an official DFI recommendation, nor does it mean it's even necessarily correct.

The real answer is still that you need as much power as the sum of your parts' power requirements. And clearly, a 300W PS is plenty for 6 drives, a SCSI RAID card, a very high end ATI X800 XT PE graphics card that sucks significant power, and other peripherals in an overclocked AMD XP 3200+ system.

I'd be more comfortable with an EE commenting on this, but I suspect that the real requirement for most of these boards is stable amperage, and that's why I view with extreme skepticism most of these "Buy a power supply of XXYYZZ watts". Clearly, Compaq, Dell, and a host of other vendors ship with far smaller power supplies and they have no problems. Clearly, most of us use far smaller power supplies and we have no problems; a 480W PS is absurd.

Sorry; one frustrated DFI tech doesn't make policy, and a 480W PS for every DFI motherboard is stupid. If that's their official policy, it's stupid. But I'm sure it isn't. To me it looks like a clean way that DFI can wash their hands of problems and force their users to spend money when they don't need to, and that says a lot about how poor their support model is if that's their policy. Again, I can't believe it is.

All HP boards use 400W or smaller PSs here: http://www.shopping.hp.com/webapp/shopp...tegory=desktops/hp_pavilion&catLevel=1

All the Compaq boards (see left side of page) use 300W PSs or smaller on that same link.

Here http://www.intel.com/support/platform/pentium4/cases.htm Intel says 250W is good for a "typical" P4 system.

nVidia says 350W, and that's with their most power-hungry GPU in the system: http://www.xbitlabs.com/news/video/display/20040514055549.html (most users wouldn't need anywhere near that amount of wattage)

Tom's Hardware says about 300W is good for most folks building their own system: http://www.tomshardware.com/howto/200507111/stresstest-01.html -- and he comments on how power supplies are a ripoff business taking advantage of many people's percieved need for a higher-wattage PS. That alone should send up warning flags to most buyers that they're typically buying things they don't need.

Even DFI's own spec sheets don't agree with that tech's rant. http://www.dfi.com.tw/Upload/Manual/lputnf4%20847505102.pdf (page 62) says anything from 350W to 500W, depending on CPU and parts. Not "always 480W" at all! --- and that's with their high-end SLI board, assumes dual graphics cards, etc. etc. - most of their boards would need far less power.

Again, it depends on what's in the machine for what power supply is required. For most people, 480W isn't even close to being needed - 300W will typically easily suffice, 350W if you want to spend a little more and get a little more headroom.

 

[Fresh Daemon]

This is what I said before:

I think the misconception is that when nVidia says "you need 550W for SLI", they mean 550W from a crappy PSU, which will fluctuate and not deliver good amps. They could say "you need a 400W PSU for SLI from a good manufacturer" but then they would have to start defining a good manufacturer, and for nVidia to start endorsing or condemning various third parties is going to be a legal nightmare.

Replace "nVidia" with "DFI". The problem is that if DFI says "go get a 300W PSU", some people will go buy a $10 special, and it will be unstable, and it will crash, and they will complain. So then DFI can either say, "we meant get a Seasonic/Enermax/PCP&C/OCZ 300W PSU", then they might get sued by anyone they leave out, and other legal problems. Or they can say, "just get a 480W PSU", and that way, even if the customer gets an el-cheapo 480W PSU, odds are that the 480W no-name would at least have equal amps to a 300W PSU of good brand.

That's my estimation. I know that people happily run DFI rigs on good-name 300W PSU's with plenty of amps. I also know that DFI rigs exhibit a lot of problems with 400W no-name PSU's. Personally, I think it would be best if the industry could move away from defining power requirements in terms of watts and move to defining it in amps.

 

[mechBgon]

The real answer is still that you need as much power as the sum of your parts' power requirements.

I don't think you're taking into account the possibilities of overhead and transient conditions. For example, peak amperage draw by a modern fluid-bearing HDD is high at startup. Some people who have lots of drives for storing their media collection, and don't have the benefit of staggered-start SCSI goodness, may need "too much" power just to get the system rolling at boot-up. Seen that before around here too.

I agree that a well-made PSU is going to more closely approach your ideal that the PSU only needs as much wattage as the (true, overhead+transient) maximum load of the parts. Reality being what it is, and people tending to upgrade/add stuff as time goes by, and the cost of an "overkill" wattage PSU being not much higher than the just-enough one, I'm sticking with what I know works most consistently. People spend an extra $20 on much more worthless things than upgrading their PSU to "overkill" wattages. LED fans come to mind...

 

[dclive]

Originally posted by: mechBgon

The real answer is still that you need as much power as the sum of your parts' power requirements.

I don't think you're taking into account the possibilities of overhead and transient conditions. For example, peak amperage draw by a modern fluid-bearing HDD is high at startup. Some people who have lots of drives for storing their media collection, and don't have the benefit of staggered-start SCSI goodness, may need "too much" power just to get the system rolling at boot-up. Seen that before around here too.

I agree that a well-made PSU is going to more closely approach your ideal that the PSU only needs as much wattage as the (true, overhead+transient) maximum load of the parts. Reality being what it is, and people tending to upgrade/add stuff as time goes by, and the cost of an "overkill" wattage PSU being not much higher than the just-enough one, I'm sticking with what I know works most consistently. People spend an extra $20 on much more worthless things than upgrading their PSU to "overkill" wattages. LED fans come to mind...

It's not up to me - Intel, AMD, Tom's Hardware, and even DFI (in official documentation) agree with what I've written (clearly, since I'm just quoting them).

And while the hard drives may have higher overhead when starting up, other parts of the system will have lower overhead. In the end, it tends to even out.

I don't disagree that some people overload their system. I do, however, disagree that everyone needs a 480W PS. Even a 300W PS is massive overkill for many of the systems out there today. Dell systems in their business (optiplex) lineup typically go from 210W in SFF to 305W in the minitower. That suggests that Dell knows something about power requirements - and it also suggests Dell's partner, Intel, is telling them that 305W is fine on the high end and 210W is fine on the low end.

That's a far cry from that tech's 480W diatribe. I hope DFI will comment on why he wrote that (or censure his statement), because it's very different from what other vendors have written.

 

[mechBgon]

I don't know if you realized this, dclive, but

1) some of the DFI's are SLI boards. You can plop two 75W video cards into them. The enthusiast will be piling two opticals and three or four HDDs into it for their media collection.

2) Dell has used 350W PSUs in mid-line systems that could accept a decent gaming card, and last I checked, their XPS line uses 460W. Even Shuttle has some SFF systems with 350W units.

 

[mechBgon]

Update: Dell XPS now uses a 650W power supply. XPS 600

It's NEW! The most advanced Dell XPS ever built. Includes a newly engineered chassis with Dual PCI Express graphics and your choice of Genuine Windows® XP operating systems. The XPS 600 also offers a variety of processor options including Intel® Pentium® D or Extreme Edition Dual Core Processors with HT Technology, or Pentium 4 Processor with HT technology up to 670 (3.80GHz, 2MB L2 cache, 800MHz FSB). Additional air venting provides improved chassis air-flow and there are 7 exciting back lighting options. It's breaking boundaries for the total multimedia experience.

With the possibility of a dual-core P4 Extreme Edition, plus two 7800GTX's or whatever comes next, and next after that, and next after that, I'll give the 650W PSU a "just right in the long haul" rating for that system. Which of course means I'm crazy as usual but whatever.

Addendum: It might be worth noting that the peak output of a typical ATX power supply goes down significantly as temperature goes up. Using an Enermax 430W dual-fan unit as an example, it's rated for:

• full output at at 0-25°C
• 80% of that at 40°C
• zero output at 70°C

In a modern system that might have two 3D accelerators and a CPU putting out a combined total of ~250W of heat at full throttle, right below the PSU, plus the usual other hardware, will the upper-rear interior of the case necessarily be 25°C? Closer to 40°C, maybe? This is a reality factor that gets overlooked by some folks.

edit: it's clear that neither of us is going to change our minds, so let's call it a draw and let the thread resume its regularly-scheduled programming

 

[dclive]

Originally posted by: mechBgon
I don't know if you realized this, dclive, but

1) some of the DFI's are SLI boards. You can plop two 75W video cards into them. The enthusiast will be piling two opticals and three or four HDDs into it for their media collection.

2) Dell has used 350W PSUs in mid-line systems that could accept a decent gaming card, and last I checked, their XPS line uses 460W. Even Shuttle has some SFF systems with 350W units.

All of that is true, but the DFI tech's post that you quoted didn't mention any of that - it simply said that for support you'd better have a 480W PS in there. That's absurd. And all of those numbers you've posted, you'll note, are under 480W. Again, if you have the box loaded to the top with hardware and SLI, you might need a slightly larger PS than the 250W-350W generally seen and used today, but that isn't the same as 'use 480W because we say so', which contradicts DFI's posted manuals.

Dell's AGP-equipped and PCI-E equipped systems also have far less than 350W - some at 305W, and some less. It certainly isn't a de-facto requirement.