Flashrom To Support Flashing ATI Graphics Cards

[phoronix]

Phoronix: Flashrom To Support Flashing ATI Graphics Cards

We have learned that Flashrom, an open-source program for flashing the BIOSes on many different motherboards / chipsets, is soon going to be picking up support for flashing the video BIOS image on ATI graphics cards. Specifically, it should be possible to flash the BIOS of the ATI Radeon X1000 (R500) series and potentially the Radeon HD 2000/3000 (R600) series too...

http://www.phoronix.com/vr.php?view=NzI2Mw

[n0nsense]

I tried to use it for my Nvidia 9400 based Gygabite MB (MCP79) and it failed.
I tried 2 more M/B and had no success ...

[bugmenot]

Nice!
Free flashing also for graphics cards! :)

[yoshi314]

i tried flashrom on my k8nf9u gigabyte mainboard. bios chip was detected. i was able to dump the bios. but...

all i can say - dual bios solution is a life saver.

[Adarion]

Interesting thing, but it needs of course some content to flash on these VGA-BIOS Chips. And of course one should have a backup card in store. Especially with the lot of possibly different implementations by card vendors you may need a lot of flags/command line options to set as it is with the mainboards.

[yoshi314]

but it needs of course some content to flash on these VGA-BIOS Chipsif you report bugs on freedesktop.org frequrently you probably would learn the typical linux videocard bios dumping mantra; at least as far as ati cards are concerned.

it would not be too difficult to get a hold of a e.g. bios file from a different (preferably newer revision of the same model) card this way.

[Adarion]

if you report bugs on freedesktop.org frequrently you probably would learn the typical linux videocard bios dumping mantra; at least as far as ati cards are concerned.
it would not be too difficult to get a hold of a e.g. bios file from a different (preferably newer revision of the same model) card this way.

Aww, don't tempt me. Sounds interesting but I might end up bricking my cards.
Is there a link for a procedure howto and neccessary tools how to do a dump of a (ATI) VGA BIOS?

[Louise]

Why does a graphics card have a BIOS???

What goes on when a graphics card is powered on?

[bridgman]

The main point is that the System BIOS doesn't know anything about graphics hardware, so all code and data specific to the display has to be provided by the Video BIOS. That way if you take one graphics card out and put another one in you get new BIOS information to match the new hardware... and you need that in order to light up the screen during the boot process, before the OS and drivers have loaded. Everything soldered onto the motherboard is handled by the System BIOS, everything soldered onto the graphics card is covered by Video BIOS.

When you have a motherboard with integrated graphics, there is still a separate VBIOS but the VBIOS image is usually stored in the same ROM as the SBIOS (and the ROM is bigger as a result).

As to what happens (ie what the BIOS does) :

The SBIOS calls VBIOS to initialize the card, and then makes subsequent calls (typically using INT10) to display information on the screen. BIOS code implements the standard INT10 calls and VESA BIOS Extensions (VBE) which are used by SBIOS and also by a number of operating systems. One example of this is the standard "vesa" DDX driver, which makes VBE calls to control the graphics card.

http://en.wikipedia.org/wiki/INT_10

http://en.wikipedia.org/wiki/VESA_BIOS_Extensions

The VBIOS also provides information which is used by drivers to understand the details of the card, eg what kind of connectors, how they are wired, memory type/size/speed etc. This can either be in the form of data tables or code or both.

[Louise]

Interesting!

Makes one wonder, what the motivation is to flash it :)

I can understand the purpose of coreBoot, if you are Google, but flashing a graphics card BIOS? :)

[bridgman]

Interesting!

Makes one wonder, what the motivation is to flash it :)

I can understand the purpose of coreBoot, if you are Google, but flashing a graphics card BIOS? :)

Back in the R300 days (and maybe earlier) we shipped some inexpensive cards which used the same GPU chip as more expensive cards but with some of the graphics pipes disabled. The GPUs on those boards were usually ones which failed functional test with all pipes enabled but which passed all the tests with one or more pipes disabled.

Flashing the BIOS allowed you to fool the driver into thinking it was running on a more expensive card, which enabled all the pipes and increased performance, but which often resulted in some corruption and occasional hangs as well. Ever since then, the dream of "flashing the BIOS to make my card run twice as fast" has persisted.

More recently, there were some cases where board manufacturers shipped VBIOS images which had less-than-optimal temp/speed profiles for the fan controllers. In these cases the manufacturers posted BIOS images with improved fan controller profile settings, along with a utility which allowed the BIOS to be re-flashed with the improved profile.

A number of users took advantage of tools which tweaked the clock settings in the BIOS images, allowing them to over-clock the boards further than the manufacturers allowed.

When the 4870 came out a number of owners decided to re-flash their boards without manufacturer support and used an existing flash utility. The result in almost every case was a non-functional board since the 4870 boards used a larger (128KB) ROM. This, in turn, resulted in a rash of reports that new 4870 boards were "failing for no reason" after a few hours use. After a new flash utility was released which programmed the *entire* ROM image the reports of "sudden 4870 death" stopped just as suddenly.

The main point is that flash tools are great for allowing in-the-field updates provided by the manufacturer (ie putting in an image designed for your board), but flashing with the image from even a slightly different board is generally a Bad Thing, and nearly impossible for the driver to code around since the BIOS contents are the driver's only source of information about the hardware on which it is running.

[bugmenot]

I think it is a good thing and if amd does not provide a bios flasher, it's great that other people do. I had problems with an OEM mainboard and flashed a different image, a BIOS from phillips, although i brought the pc in mediamarkt and there war actually a microstar BIOS in it. But this contained many bugs, you cannot imagine how happy I was, if I could flash another image! I hate it to depend on the manufacturer, because they do usually not help in such cases.

If a graphics card manufacturer is too stupid to implement good fan control or power saving, why not flash another image to be happy? Of course it is dangerous, but that's the only way.

[energyman]

well, back in my nvidia days a lot of people with problems solved them by flashing the card's bios...

[Louise]

Back in the R300 days (and maybe earlier) we shipped some inexpensive cards which used the same GPU chip as more expensive cards but with some of the graphics pipes disabled. The GPUs on those boards were usually ones which failed functional test with all pipes enabled but which passed all the tests with one or more pipes disabled.

Flashing the BIOS allowed you to fool the driver into thinking it was running on a more expensive card, which enabled all the pipes and increased performance, but which often resulted in some corruption and occasional hangs as well. Ever since then, the dream of "flashing the BIOS to make my card run twice as fast" has persisted.

More recently, there were some cases where board manufacturers shipped VBIOS images which had less-than-optimal temp/speed profiles for the fan controllers. In these cases the manufacturers posted BIOS images with improved fan controller profile settings, along with a utility which allowed the BIOS to be re-flashed with the improved profile.

A number of users took advantage of tools which tweaked the clock settings in the BIOS images, allowing them to over-clock the boards further than the manufacturers allowed.


So the VBIOS on my Asus graphics card is programmed by Asus, and not ATi?

Does IGP's also have a VBIOS? (Sounds expensive to put a flash inside the Northbridge?)


When the 4870 came out a number of owners decided to re-flash their boards without manufacturer support and used an existing flash utility. The result in almost every case was a non-functional board since the 4870 boards used a larger (128KB) ROM. This, in turn, resulted in a rash of reports that new 4870 boards were "failing for no reason" after a few hours use. After a new flash utility was released which programmed the *entire* ROM image the reports of "sudden 4870 death" stopped just as suddenly.


haha :D


The main point is that flash tools are great for allowing in-the-field updates provided by the manufacturer (ie putting in an image designed for your board), but flashing with the image from even a slightly different board is generally a Bad Thing, and nearly impossible for the driver to code around since the BIOS contents are the driver's only source of information about the hardware on which it is running.

I will never understand, why someone (today) would want to try and enable extra features on GPU's/CPU's. They are so cheap today, and even thought AMD have planned to release a dual core Phenom, which actually is a quad core Phenom, I would never want to trade a 100% working CPU for a maybe running CPU.

Then there's those that say that it is so easy to enable the extra cores on e.g. a triple core to become a quad core, and AMD wants users to do that. But still, a AM2+ quad core costs almost nothing now a days.

And with the AMD/Intel case yesterday, we might even see even cheaper quad cores due to increased sells :D

[bridgman]

So the VBIOS on my Asus graphics card is programmed by Asus, and not ATi?

Does IGP's also have a VBIOS? (Sounds expensive to put a flash inside the Northbridge?)

We supply the BIOS code and data tables but the board mfg changes specific data tables to match their board design & preferences.

IGPs also have a VBIOS, but it's stored in the same ROM chip on the motherboard which holds the system BIOS (SBIOS). It's not actually stored inside the Northbridge chip.

[Melcar]

Back in the R300 days (and maybe earlier) we shipped some inexpensive cards which used the same GPU chip as more expensive cards but with some of the graphics pipes disabled. The GPUs on those boards were usually ones which failed functional test with all pipes enabled but which passed all the tests with one or more pipes disabled.

Flashing the BIOS allowed you to fool the driver into thinking it was running on a more expensive card, which enabled all the pipes and increased performance, but which often resulted in some corruption and occasional hangs as well. Ever since then, the dream of "flashing the BIOS to make my card run twice as fast" has persisted.

...


Those where the good ol' days. x800gto -> x850xt. So much awesomeness.

[bridgman]

Yeah, that's what I'm talking about ;)

There was a 4-pipe R300 board which could be reflashed to 8 pipes as well.

[Adarion]

The result in almost every case was a non-functional board since the 4870 boards used a larger (128KB) ROM. This, in turn, resulted in a rash of reports that new 4870 boards were "failing for no reason" after a few hours use.
Heh. Don't tell me there have been people flashing their cards with a more or less obviously inappropriate VBIOS image and then complainig to ATI-AMD their cards would fail.
LOL. I mean, how fresh can people just be?

If I bricked my HW due to my own incompetence I wouldn't dare to complain, I'd rather ask for an advice what could be done to revive it.


How different are VBIOS implementations by the vendors? Is there a generic image/code from ATI/AMD that would work everywhere (like a driver does) or do they really contain a lot of "unique" code? Because then one would have a hard time to try updating some card when the specific vendor won't support in in terms of VBIOS updates. (I'm glad to have my passive Sapphire 3870 and my passive IGPs so I won't run into fan noise problems...)

[energyman]

Adarion you never have been on an overclocker board. Those imbeciles RMA perfectly fine boards because they do not overclock as good as one from another random guy (thus making boards more expensive for everybody). Those idiots flash bios files from different boards&cards to get some mythical bios option or gain a 0.05% increase in superpi or 20 more points in 3dmark. And if they brick it - they complain. Very, very loudly. To the manufacturer. Hardware would be cheaper and the world a better place without this dumb f***s.

[Kano]

Using tools like atitool (win) you can get a vga bios backup for ati cards. nvclock makes a backup for nvidia ones. In case you want to mod it ;) Which i did for a Radeon 9700 back then, you could raise the refresh rates up to 85 hz for vesa modes (good for crt) and oc it by default - could be done with rovclock too, but i wanted more speed before that tool was developed *g*

[Melcar]

Adarion you never have been on an overclocker board. Those imbeciles RMA perfectly fine boards because they do not overclock as good as one from another random guy (thus making boards more expensive for everybody). Those idiots flash bios files from different boards&cards to get some mythical bios option or gain a 0.05% increase in superpi or 20 more points in 3dmark. And if they brick it - they complain. Very, very loudly. To the manufacturer. Hardware would be cheaper and the world a better place without this dumb f***s.


The new crop of overclockers is more like that. Back in my day things like that happened rarely. I think it's because back then one really had to know the hardware, whereas nowadays any fool can download utilities (and sometimes even direct from the manufacturer) and start pushing sliders.

[Kano]

Gfx cards die much faster than the really old ones, specially nvidia 6 series and ATI cards R300-R400 seem to fail currently very often - but usally after some years of use. I think most of em die because of overheat due to failing fans. When you find a stable oc without artefacts you can usually use that without problems, as the chips are only clocked to a well known default that works everywhere. But when you oc too fast and don't check for artefacts then you can get lockups and the cause is hard to identify.

[energyman]

Using tools like atitool (win) you can get a vga bios backup for ati cards. nvclock makes a backup for nvidia ones. In case you want to mod it ;) Which i did for a Radeon 9700 back then, you could raise the refresh rates up to 85 hz for vesa modes (good for crt) and oc it by default - could be done with rovclock too, but i wanted more speed before that tool was developed *g*

and the backup helps you how much when the card is completly unresponsive after a flash?

[Kano]

Well i had that issue too, when i tried to flash another bios instead of modding the original one. But i made a bootfloppy that restored the backup automatically and that recoverd it. Then i used a mod tool to just change the clockspeeds + refresh rates and flashed the modded back, that worked fine. When you try different vga bios versions it is more likely that you break something, but it is not lost in most cases, with right way you can recover it.

[Adarion]

The new crop of overclockers is more like that. Back in my day things like that happened rarely. I think it's because back then one really had to know the hardware, whereas nowadays any fool can download utilities (and sometimes even direct from the manufacturer) and start pushing sliders.

I guess that's true. But back in the times of old we all had to know at least a little bit about computers, you had to be somehow dedicated to them, be a little bit of a freak; someone who had no problems reading a manual and a book about it.
Nowadays everybody can go shop for a colorful box and make a mess with it (ever noticed that a lot of people buying notebooks care a shit for the inside, they just go for the optical look). Furthermore OV today is nearly a mass sport, indeed. I often noticed that all kinds of e.g. mainboard manufractureres offer jumpers, settings, potentiometers and so on for OC purposes and also market their stuff as OCable.
But who cares about the warning signs that this behaviour is unsupported and would void warranty?

But the more people use computers the more idiots you will find. There are several funny pages about DAUs (German for luser, dumbest possible user) on the net and they show also people returning items to the shop and complaining but having used a hammer to force a CPU in a completely wrong socket. You know, even some of these failsafe to connect connectors (eg. Molex power for PATA-HDDs) inserted 180° wrong. With force. Breaking everything. But then people come complain or some of them hope the seller won't notice and refund the item. Or they come with "it was sold to me like that". Yeah.

I think a computer is far more complicated and able to do more things than e.g. a car. But when you want to drive a car you have to make a license and show that you can follow basic rules of traffic. I think something like that should be done for computers, too. Nobody needs to understand all internals (unless you want to mess with them) but some basic knowledge...
*sigh* Dreaming is still allowed, right?

[nanonyme]

Well i had that issue too, when i tried to flash another bios instead of modding the original one. But i made a bootfloppy that restored the backup automatically and that recoverd it. Then i used a mod tool to just change the clockspeeds + refresh rates and flashed the modded back, that worked fine. When you try different vga bios versions it is more likely that you break something, but it is not lost in most cases, with right way you can recover it.Works great in a laptop without a floppy drive? ^^
Edit: Actually seems desktops are giving up floppy drive connectors too. I doubt the bypassing that skips directly to floppy boot so you can even fix motherboard BIOS problems (I don't really know how it works but apparently some good motherboards could do it with the right pin setup) would work that well with an USB floppy drive... Although graphic card BIOS is simpler to fix, laptops are a place imo where *no one* but vendors should *ever* flash anything. (so much for the fan speed and powersaving? :) )

[curaga]

With a desktop, no need for floppies - just insert a PCI card to see something, reflash the agp/pci-e card.

[yesterday]

I will never understand, why someone (today) would want to try and enable extra features on GPU's/CPU's. They are so cheap today, and even thought AMD have planned to release a dual core Phenom, which actually is a quad core Phenom, I would never want to trade a 100% working CPU for a maybe running CPU.

Then there's those that say that it is so easy to enable the extra cores on e.g. a triple core to become a quad core, and AMD wants users to do that. But still, a AM2+ quad core costs almost nothing now a days.

And with the AMD/Intel case yesterday, we might even see even cheaper quad cores due to increased sells :D

Not all people who flash the BIOS do so coz we wanna be l33t hax0rs.

In the Utopian fantasy that bridgman describes, there would be no reason to flash the video bios. Then again, there would be very little reason to flash the system bios either.

However that world is just that, fantasy. People DO come across cards that have problems that can ONLY be solved flashing the BIOS. I have two examples from the same NVIDIA card. On a Galaxy 7900gt that was factory OC'd (ie not by me), the fan controller for a bizarre reason turned the fan OFF after 50 deg C. This means when playing a games in , the card fan would turn off just when it was needed most. This problem was in the BIOS. The controller wouldn't respond to changes in the driver. Unfortunately, I could never fix this problem, because no-one was really sure how the thermal management was done in the BIOS, and Nvidia (and Galaxy) by then had moved on to the 8000 series and forgot about us.

My second problem with the card was due to the factory OC settings. Cards with a factory OC didn't play nice with the NVIDIA drivers on Windows, resulting in the dread scrambled screen when starting any 3D app. This problem still persists for the 7900gt. The only permanent solution was to the flash the BIOS and change the core or memory clock.

Flashing the Video BIOS is often the last resort, but how stupid is it not to have the tools to do it on your OS of choice? Just imagine if the only way you could flash your system BIOS was by using Windows...

[yesterday]

Adarion you never have been on an overclocker board. Those imbeciles RMA perfectly fine boards because they do not overclock as good as one from another random guy (thus making boards more expensive for everybody). Those idiots flash bios files from different boards&cards to get some mythical bios option or gain a 0.05% increase in superpi or 20 more points in 3dmark. And if they brick it - they complain. Very, very loudly. To the manufacturer. Hardware would be cheaper and the world a better place without this dumb f***s.

I don't know why you have to spew vitriol at overclocking. I used to overclock, because I found it fun and it's fun to find the limits of your hardware. Having an 1.8GHz Dual Core A64 running at 2.7Ghz is pretty nice, and it DOES make a difference. None of the overclocking communities I was a part of would ever condone RMAing parts you broke overclocking, so please don't tar people with same brush based on your few experiences.

And please, the overclocking communities are important to hardware companies. They help establish performance leaders, and foster brand awareness. Intel for example, leaked a number of benchmarks for the Core series on xtremesytems before the launch to create buzz.

Additionally, companies actively sell expensive "OC" products to tap into this market, so I have little sympathy for them when products DO get returned after being busted. I'm sure you can agree that selling a product based on its overclocking potential, then fully trying to disclaim responsibility when said product is overclocked, is just plain stupid.

Oh, and enthusiasts and overclockers can really help pinpoint the future viability of current products. The lack of OCing headrom in Phenom I gave the impression that the processor speeds wouldn't scale well, and this was right. Similarly, the OCing performance of the next gen Phenom gave people hope that AMD would come out with a more competitive product sooner rather than later, and this proved true as well. And the popularity and performance of OC'd Pentium M Desktop boxes made Intel look at that "old" technology. The result was the dropping of Netburst and the birth of the Core architecture.

[energyman]

no, Intel dropped Netburst because it was slaughtered by Amd's A64 and did not scale half as well as hoped ( you might remember when Intel released the P4 and talked about reaching 5Ghz and more). Also everything I wrote is true - just visit the different oc forums like ocworkbench. 'I only reached XY so I RMA'd the board' is a common theme. And last but not least - most people do not overclock. But the companies cater for this very small and stupid market. Why? Because magazines are stupid. Way too much weight on useless overclocking, not enough on noise, heat price/performance. The points most important for the vast majority.

[yesterday]

no, Intel dropped Netburst because it was slaughtered by Amd's A64 and did not scale half as well as hoped ( you might remember when Intel released the P4 and talked about reaching 5Ghz and more). Also everything I wrote is true - just visit the different oc forums like ocworkbench. 'I only reached XY so I RMA'd the board' is a common theme. And last but not least - most people do not overclock. But the companies cater for this very small and stupid market. Why? Because magazines are stupid. Way too much weight on useless overclocking, not enough on noise, heat price/performance. The points most important for the vast majority.

Everything you said must be true because you say so eh? All overclockers are idiots and ruining IT because you believe it huh?

You can point to oc forums you don't like, while I can name xtremesystems, HardOCP, anand, OCAU, the DFI forums etc., that have good reputations. I was an overclocker and never RMA'd a single piece of hardware. In the same vein, I could find forums full of Linux zealots. Does this mean all Linux users are total arse-holes, and that Linux is totally useless?

And yes, most people don't overclock, just as in the same way, most people don't buy the highest performing graphics card. But for many of these companies the performance and enthusiast market is important for their branding and image. And it obviously impacts their sales otherwise they wouldn't build such products. I fail to see how this is the fault of the "media". People looking for overclocking performance will look for sites that focus on this, in the same way that I go to silentpcreview and phoronix, based on my interests. All the major general sites are quite balanced. Why rag on overclockers here anyways? Why not rag on people who get SLI or high end cards that draw 80W at idle that are hugely expensive? These setups get way more focus in mainstream hardware circles than overclocking.

The idea that overclocking is useless is just stupid. Just because you aren't into doesn't mean other people don't find value. Overclocking started as a way for getting more bang for your buck. Buying a $50 product and getting the same performance as a $200 product is pretty useful for some people.

But please, feel free to ignore everything I say and tell me that "everything you say is true"

[lordmozilla]

The idea that overclocking is useless is just stupid. Just because you aren't into doesn't mean other people don't find value. Overclocking started as a way for getting more bang for your buck. Buying a $50 product and getting the same performance as a $200 product is pretty useful for some people.


Usually the time spent on getting your extra money's worth doesnt make it worth it.

I know. I have a Q6600 running at 3.0ghz. I can't notice the difference. It probably means i get a few more FPS in some games and can run some stuff faster. I dont care. I can get it to 3.6ghz if I want to. I dont cause it gets all the fans noisy but thats not the point.

I also had a 6800XE flashed as a 6800Ultra and a 5900XT flashed as a 5950Ultra, an AMD Athlon XP 2500M running at 2.6ghz.

It's always more trouble than its worth. But people love the idea of getting that extra performance for free and love the feeling of being elite.

True a 3ghz quad core would have set me back alot at the time, but I probably spent 50hrs overall fixing small problems, messing with memory timings etc... 50x5 = £250, so at minimum wage I could probably have afforded the better CPU.

End of story, overclocking makes you feel good about your money spending, it doesnt give you anything. Even the best overclock usually doesnt mean you can change from medium to high on any game setting!

But by no means believe it is usefull.

[suokko]

But by no means believe it is usefull.

It is hobby :) To find good parts that over clock well and doing that overclocking can be fun. Same idea as tuning a car :)

[yesterday]

Usually the time spent on getting your extra money's worth doesnt make it worth it.

I know. I have a Q6600 running at 3.0ghz. I can't notice the difference. It probably means i get a few more FPS in some games and can run some stuff faster. I dont care. I can get it to 3.6ghz if I want to. I dont cause it gets all the fans noisy but thats not the point.

I also had a 6800XE flashed as a 6800Ultra and a 5900XT flashed as a 5950Ultra, an AMD Athlon XP 2500M running at 2.6ghz.

It's always more trouble than its worth. But people love the idea of getting that extra performance for free and love the feeling of being elite.

True a 3ghz quad core would have set me back alot at the time, but I probably spent 50hrs overall fixing small problems, messing with memory timings etc... 50x5 = £250, so at minimum wage I could probably have afforded the better CPU.

End of story, overclocking makes you feel good about your money spending, it doesnt give you anything. Even the best overclock usually doesnt mean you can change from medium to high on any game setting!

But by no means believe it is usefull.

Wait,
are you actually telling us that increased clockspeed doesn't make a processor faster? Because, that sort of is the definition of clockspeed.

The last time I looked, chip companies sold processors with various clock speeds. People generally choose the fastest
they can afford of a certain architecture. Intel sell Quads of the same family from 2.67 to 3Ghz. They sell Duals of the same family ranging from 2.67 to 3.33Ghz. These are retail chips. What does overclocking give me? If it works, it gives me a chip at the same speed than the more expensive retail version.

Now, arguing whether the extra clockspeed actually translates to improved user experience is pointless. Firstly, it, depends on what you are doing. Secondly, you may as well ask this of chip makers, instead of overclockers. If it's pointless OCing a chip from 2.66 to 3 Ghz, it's equally pointless to buy a 3Ghz chip instead of the cheaper one.

The reality is that clockspeed DOES improve performance. Show me a benchmark where a lower clocked processor performs better than a higher clocked processor of the same architecture. Whether you notice the improvement now, next year, or never really isn't the point.

And seriously, taking wage calculations? Do you skip work to fiddle with your computer? I don't. The amount of pain comes down to the amount of research you do in products, the amount of time I spent researching what to buy when OCing pretty much equals the amount of time I spend now looking for cool/quiet parts and linux compatible stuff.

[energyman]

you need at least 10% improvement to be able to tell the difference. On almost all systems not the CPU is the bottleneck, but harddisks&co. So even if you increase the mhz of the cpu by 30%, it will translate in a 'visible' improvement of maybe 3% - if you are lucky. Way below the threshold.

Also, every feature costs money. Boards could be a lot cheaper without useless overclocking options. There are some options that are valid, even good, like memory voltage, memory speed. But the rest? Southbridge voltage? why? Now, less noise, less energy consumption, less heat, that benefits everybody not only the 1-5% who overclock.

At the end overclocking is a waste of time. It is just a penis lenght comparism. Most people don't benefit from it. The vast majority is hurt by higher prices. This makes overclocking stupid.

[Ant P.]

I wonder if messing with my R300's BIOS to change some things would get it into a useful state again.

I got the card for free because its RAM would get corrupt after 1-2 hours and then it stays that way even over a cold boot. It works fine when it's not been used for a few days. Would be nice to recycle it at a lower clock instead of just binning it.

[yesterday]

you need at least 10% improvement to be able to tell the difference. On almost all systems not the CPU is the bottleneck, but harddisks&co. So even if you increase the mhz of the cpu by 30%, it will translate in a 'visible' improvement of maybe 3% - if you are lucky. Way below the threshold.

As I have already said, this has nothing to do with overclocking. Intel and AMD release numerous chips under different clockspeeds. Why don't you write them a letter letting them know they are ruining IT?

If I OC my 8200 to 2.8 Ghz (less than 10% increase), how is this any different to buying a 8400 instead?

AMD latest Phenom IIs come in speeds of 3.2 and 3.0 Ghz. Less than a 10% difference.

Arguing that "overclocking is pointless" is just pretty much arguing that clockspeed increases are pointless. You should take that up with the processor companies before you go after the overclockers. Oh, and remember to leave out people who actually have compute-bound tasks from the discussion.

Also, every feature costs money. Boards could be a lot cheaper without useless overclocking options. There are some options that are valid, even good, like memory voltage, memory speed. But the rest? Southbridge voltage? why? Now, less noise, less energy consumption, less heat, that benefits everybody not only the 1-5% who overclock.

Wow, there is alot of misinformation here that I don't know where to start. Firstly, only a handful of boards are designed with OCing in mind. A look through all the major mobo websites shows that the majority of boards are not OC focused. Add in OEM boards, and you'll realise that the vast majority of boards aren't designed for any sort of hardware tweaking. Boards that are OC focused cost WAY more than non-OC boards. With this the case, how can you argue that overclocking is making parts more expensive? It's also a bit silly to act as if boards are hugely expensive. You can find heaps of boards here in Australia for around $100, which is about the cost of a video game. I consider this cheap. Also, do not confuse BIOS options with an OC focused design. A board may offer many BIOS options, but not guarantee to run out of spec. Having a BIOS menu isn't the same as designing for tweakability.

Secondly, you speak as if power efficiency coupled with high performance is a trivial problem solve. Well, it's not. It has nothing to do with the overclocking community. The laptop/netbook and server/workstation spaces have a HUGE emphasis on power/performance, but in those spaces lower powered devices with high performance are very expensive, because they are difficult to design and build. In fact, in consumer land, look at AMDs latest TOL graphics card, the 4890. The thing idles at a handy 71W. That's what's needed to compete on a performance level. You pay a price or performance premium for power efficiency, because it's just easier to design inefficient systems. Optimization of ANYTHING is hard. Blaming overclockers for the thermodynamic realities of computation is silly and disingenuous.

At the end overclocking is a waste of time. It is just a penis lenght comparism. Most people don't benefit from it. The vast majority is hurt by higher prices. This makes overclocking stupid.

I love how you just ignore previous points and just repeat the same thing, albeit with more pejorative language.

The only thing you've proven is that you don't like overclocking, and that you are somewhat pretentious in assuming that things you don't like are worthless. In fact, you are willing to trot out misguided and specious arguments just to justify your distaste.

[yesterday]

I wonder if messing with my R300's BIOS to change some things would get it into a useful state again.

I got the card for free because its RAM would get corrupt after 1-2 hours and then it stays that way even over a cold boot. It works fine when it's not been used for a few days. Would be nice to recycle it at a lower clock instead of just binning it.

Have you run the card with a lower mem clock though software? If it's still busted, it's probably beyond your ability to fix.

Keep in mind the video BIOS is not like the system BIOS. You don't get a nice menu with lot's of options. You basically just flash in a binary file. It's not something you typically "mess with". There is generally a specific reason to change it and specific part you want to update. Just flipping bits is not going to be very successful.

In your case, to get the lower clockspeed you'd need to either figure out the BIOS binary yourself, or find someone or some document that shows you what to change to change the mem clock.

I would try to lower it in software (I guess this means Windows?) first, then look at a BIOS flash.

[Kano]

Well i think it is not problematic when you can reach higher clockspeeds without raising the vcore at all, because then the vcore is still reduced with speedstep/powernow. Many boards do not reduce it when you select a fixed voltage, some provide extra voltage + x %, thats maybe better for saving energy. I had a E6600 which ran at 3 GHz (fsb 333 instead of 266) with standard vcore (or a little raised due to automatic oc function) and run a Q9300 at 3 GHz (fsb 400 instead of 333) without those extra voltage. For testing i pushed it to the limits, but thats not really usefull for everydays work. I would call oc without raising vcore the smart way, because it provides better performance - and when you have to compile lots of things then +600 or +500 mhz are noticeable with still no extreme powerconsumption - EIST still on...

[lordmozilla]

you can probably work out that with the increased heat output of an overclocked processor, you could have saved on your electricity bill and therefore used that to fund a slightly more expensive processor.

I'm not saying overclocking doesnt improve performance. It just doesnt do it in any actual viable and useful way. Sure your benchmarks are slightly higher. Thats the same as Gentoo guyrs that recompile everything with -O3 cause that'll make everything so much faster.

Hell it's a hobby, nothing more.

[energyman]

when you want a faster cpu, buy a faster cpu. No need to 'overclock'. And even 'low end' mobos come with oc feautres today because the idiotic online hardware testing sites are all about overclocking. And a lot less about the more important heat/energy consumption. Overclocking hurts for zero benefits except to keep a very small part of the population busy.

[curaga]

Without overclocking features you wouldn't be able to underclock either, and that is (one way) how one can get lower power bill/energy consumption. If a certain cpu is sold at 3Ghz/1.35V, and you can lower it to 1.25V while not dropping speed, that can have a nice effect on the power usage without affecting performance.

[monraaf]

Well I have one of those energy efficient processors from AMD. It's supposed to have a clock speed from 1GHZ to 2.5GHZ. I've upped the bus speed in the bios, and it now runs from 1.1GHZ to 2.8GHZ, stable as a rock. Most of the time the CPU runs at 1.1GHZ and the power usage of my complete pc is between 52 and 62 watt. With CPU intensive tasks when the CPU is running at 2.8 GHZ I do really gain about 10% as to when it was 2.5GHZ, so why shouldn't I?

[energyman]

there is no need of oc features to be able to underclock.

[energyman]

Well I have one of those energy efficient processors from AMD. It's supposed to have a clock speed from 1GHZ to 2.5GHZ. I've upped the bus speed in the bios, and it now runs from 1.1GHZ to 2.8GHZ, stable as a rock. Most of the time the CPU runs at 1.1GHZ and the power usage of my complete pc is between 52 and 62 watt. With CPU intensive tasks when the CPU is running at 2.8 GHZ I do really gain about 10% as to when it was 2.5GHZ, so why shouldn't I?


because overclocking turns a deterministic machine into a random number generator.

[monraaf]

because overclocking turns a deterministic machine into a random number generator.

Well, if you push it to the extreme probably yes. Moderate overclocking never had a negative impact for me.

[energyman]

why should a manufacturer sell a chip at lower speed, making less money?
Well, because it shows problems with the higher speeds. Just because you don't see the problems doesn't mean that they are not there. Do you constantly check your dmesg? Do you know what is caught by internal checks? Do you really know that everything on your harddisks is error free? A single bit flip in a picture, a video stream, game.. who sees that?

[Kano]

Thats very simple: the cheaper chips sell better. So in case you need more cheaper chips then just relabel better ones if your yield is too high for the top chips. Thats not much differnet to the X3 cpus from AMD, lots of em work as X4 too. It is a big amount of marketing involved what you really write on the box. Of couse you can be unlucky too that your specific cpu has got no good oc range, so when you don't want to play and run checktools like Prime95 for serveral hours to check stability better buy a higher rated chip directly. But i think nothing is wrong when you oc your cpu - just verify the stability and correctness, as thats the major point when you need correct results. When you only play games and it does not hurt you when the system crashes then maybe skip that step, but i never would use a system thats not 100% working fine.

[lordmozilla]

Without overclocking features you wouldn't be able to underclock either, and that is (one way) how one can get lower power bill/energy consumption. If a certain cpu is sold at 3Ghz/1.35V, and you can lower it to 1.25V while not dropping speed, that can have a nice effect on the power usage without affecting performance.

The reason you can underclock is because it's great for laptops. The tech was brought in for desktops that simple.

[curaga]

The options to tweak voltage and speed, I consider them OC features and they were first brought for OC. If they weren't there, one could not tune them downwards either.

No, I don't mean speedstep or the other auto-things initially from laptop cpus.

[energyman]

options to 'tweak' voltage and speed always have been there because in former times both had been set manually. OC brought nothing.

[yesterday]

when you want a faster cpu, buy a faster cpu. No need to 'overclock'. And even 'low end' mobos come with oc feautres today because the idiotic online hardware testing sites are all about overclocking. And a lot less about the more important heat/energy consumption. Overclocking hurts for zero benefits except to keep a very small part of the population busy.

Please explain the difference between overclocking and buying a more expensive chip, like I have already asked about three times.

Once again you have said overclocking hurts the world, but have provided no explanation. You CLAIM have the manufacturers are wasting money on overclocking features, but you have provided no plausible reason to believe this assertion, since the vast majority of boards aren't promoted for their overclocking ability or are OEM.

You seem to believe that providing a menu in the BIOS is the same as designing for overclocking. Well, I already answered this, but once again I will say -- providing BIOS options doesn't mean a board has been DESIGNED for overclocking. A board has been designed for overclocking if the manufacturer spends time making sure that the changeable BIOS options actually work. Providing a menu to options is trivial. Making a board that can run with different settings than expected is much harder.

I have a number of low-end boards with all sorts of options, that don't overclock at all. Or look at the numerous reviews of low and midrange boards that don't overclock at all. What is the basis for your claim? And how do OEM boards fit into this claim? Are you claiming HP and Dell are responding to the evil review sites that barely feature their products?



because overclocking turns a deterministic machine into a random number generator.

why should a manufacturer sell a chip at lower speed, making less money?
Well, because it shows problems with the higher speeds. Just because you don't see the problems doesn't mean that they are not there. Do you constantly check your dmesg? Do you know what is caught by internal checks? Do you really know that everything on your harddisks is error free? A single bit flip in a picture, a video stream, game.. who sees that?

This is either plain sophistry or ignorance.

Manufacturers bin high performing parts as lower clocked chips because they adapt to the needs of the market. Intel can't just sell $1000 chips. Selling each and every chip at the highest speed it can run is a great way to go out of business. This is just basic business sense.

It all depends on what a manufacturer has planned to do with a particular stepping. It could be marked as top of the line, or if the demand is for lower priced chips, binned lower. Which is why your success in overclocking, while obviously dependent on each chip, is hugely related to the stepping of your chip. Later steppings can often give high OCs due to improvements in the fab process, for chips that have already been designated as the lower-end.

As such, it is perfectly plausible to get a chip that is perfectly stable at higher clockspeeds. You CAN test for stability, and many overclockers contribute to distributed computing projects which verify the accuracy of their data.

People have replied to every single point you have raised against overclocking, yet you can only reply with the same assertions. Repeating invalidated arguments doesn't actually strengthen your arguments at all. In fact, it only makes YOU look idiotic.

[yesterday]

options to 'tweak' voltage and speed always have been there because in former times both had been set manually. OC brought nothing.

Firstly, as I've already said, who cares what options are in the menu? That has nothing to do with how the board was designed.

Secondly, why don't you tell us which features BIOS makers are including that are costing the industry so much?

[energyman]

if you buy a faster chip it was tested and is guaranteed to run at the higher speed without errors. If you overclock you got a random number generator.

I posted that before. But hey, maybe your overclocking efforts make text disappear.

And about that 'bios menu' - the bios 'menu' only can show crap that is implemented. So to make voltages tweakable over a wide range you need more expensive hardware - and yes, the 10 cent/1000 units do make a difference. To be able to accomodate a wide range of voltages/clocks even out of spec you need a more complicated board layout - with more layers, which does increase the costs a lot. Do I really need to go on?

[deanjo]

If you overclock you got a random number generator.



That is a load of crap. As manufacturing improves, yields of chips are often artificially limited in speed to meet market demand for cheap lower end processors. This is why your lower end chips in a family are traditionally your best overclockers. Even the processor companies promote overclocking with there black editions, overclocking utils and reference BIOS. Pretty much every processor out there today has overclocking headroom which allows glitch free computation at higher then specified speeds. If anything overclockers are responsible for the higher quality of parts we see being used in boards across the full range. We aren't dealing with Cyrix 6x86's and K6-3's anymore these days.

[energyman]

*yawn* and who tells you that a chip is good? Also, as prices show you - better yield usually results in lower prices and even scrapping slower, cheaper chips. Why sell a good chip a lot cheaper, if you can sell it a bit cheaper?

You overclock, you risk your box (also, remember the electro-migration caused mass dying of P4 CPUs a couple of years ago? Yes, overclocking shortens the life span and increases the risk of failure, who would have thought of that?). But because of oc idiots everywhere prices INCREASE because 'bad overclocking boards' are RMAed and idiotic 'oc features' are built in everywhere.

[deanjo]

*yawn* and who tells you that a chip is good? Also, as prices show you - better yield usually results in lower prices and even scrapping slower, cheaper chips. Why sell a good chip a lot cheaper, if you can sell it a bit cheaper?

It's called market demand. This is what the real world goes by. Where a OEM may extend a models life 6 to 8 months before introducing it's replacement they still need those chips at lower specifications. It cost's no more to make a highend chip then a low end one. The lowend processors are what they base their projected profits from, not the extreme editions that they charge a dear premium for because it is a small market.


You overclock, you risk your box (also, remember the electro-migration caused mass dying of P4 CPUs a couple of years ago? Yes, overclocking shortens the life span and increases the risk of failure, who would have thought of that?). But because of oc idiots everywhere prices INCREASE because 'bad overclocking boards' are RMAed and idiotic 'oc features' are built in everywhere.Again myth, the silicon on a chip in a given family of processor is built no different despite the speed. Yes, early on in a development cycle the better chips are binned for the "perfomance" editions and they charge premium price for those capabilities. As a product goes through it's product life cycle yields get better and the process refined but yet demand still exists for those lowerend chips for budget machines and thermal consideration. A 125 watt chip vs 65 watt chip are often pulled from the same piece of silicon. Underclock a performance chip to the 65 watt chip and you will see the same power consumption. Chip designs are done with maximum voltage in mind. I will bet you there are more boards RMA'd on any given day due to a PEBKAC situation vs a board that has been RMA'd because it has not overclocked well. Usually the overclockers are more tech savy then your off the street guy. They tend to know exponentially more then the average joe consumer or the avg joe tech support for that matter.

[yesterday]

you can probably work out that with the increased heat output of an overclocked processor, you could have saved on your electricity bill and therefore used that to fund a slightly more expensive processor.

Not quite. Retail stock voltages differ not only for chips with different speeds, but also in the same speed chips and chips from different revisions. Manufacturers typically have a range of acceptable voltages. For example, IIRC Intel specifies that the 8200 has a range of 0.8 to 1.35 V, and my stock voltage is 1.1V. I've seen others with a stock of 1.2 - 1.3, or 0.9. So, buying a retail box is no guarantee for lower power, just power draw within the threshold.

In terms of money saved, I have some simple calculations

On average, Australian electricity costs about $3.5 Kwd.The difference between an 8200 and 8600 in Australia in about $160. Of course, a voltage bump doesn't translate to 1KW of increased draw.

To save $160 dollars a year, I would have to draw around 125W extra from overclocking the chip : 3.5 * 125/1000 * 365 =~ 160.

But 110W is no where near the about of extra Watts drawn from a voltage bump. A voltage bump from 1.1 to 1.2, translates to about 20W. So, it would take me about 6 years to save the money for the faster chip. This is assuming 24/7 for 365 days.

If I really wanted to save money, I would take a high end graphics card like the 4890 (around 70W idle) and replace it with a lower end but still decent card like the 4670 (3W idle).


I'm not saying overclocking doesnt improve performance. It just doesnt do it in any actual viable and useful way. Sure your benchmarks are slightly higher. Thats the same as Gentoo guyrs that recompile everything with -O3 cause that'll make everything so much faster.

Well, this is purely subjective, and depends entirely on your task. In that regard, we could ask you why you bothered to by a Quad Core processor. There are plenty of people out there who think we should all be happy with PIIIs.

Of course, if you are saying that a clockspeed increase is the same as a compiler option, you have been grossly misinformed, and obviously your tasks are not compute bound. Once again, what's the difference between overclocking my 8200 and buying an e8600?

Hell it's a hobby, nothing more.

Not sure what this is supposed to mean. Playing games is a hobby. For many people, so is kernel hacking. So what?

[yesterday]

if you buy a faster chip it was tested and is guaranteed to run at the higher speed without errors. If you overclock you got a random number generator.

I posted that before. But hey, maybe your overclocking efforts make text disappear.

This has been now answered by three people. Maybe you have selective blindness.

Yes a faster retail chip is almost guaranteed to run at speed X. But that doesn't imply that a slower overclocked chip CANNOT run at the same speed with the same stability.

This isn't magic. It's just the realities of the fab process.

And about that 'bios menu' - the bios 'menu' only can show crap that is implemented. So to make voltages tweakable over a wide range you need more expensive hardware - and yes, the 10 cent/1000 units do make a difference. To be able to accomodate a wide range of voltages/clocks even out of spec you need a more complicated board layout - with more layers, which does increase the costs a lot. Do I really need to go on?

Boards NEED to be tweakable by requirement. Designers MUST implement a number of different voltage, timings, clock, and bus configurations to support a wide range of processors and memory modules. How else can my board function when I use different memory modules or processors with different speeds and bus speeds, if a variety of configuration options cannot be changed in some way? Exposing these configuration options via the BIOS menu doesn't imply that the board will work out of spec or with all possible configurations. It only implies that I can TRY to run it run it, with unsupported consequences.

Or are you saying that the time spent in menu design is what is driving up costs?

Do you own or have ever owned an OEM PC, or a low-end or non-overclocking board? They don't overclock well because they aren't designed for it. You can play with menus till the end of time, but on some boards changing the clock by 1Mhz won't get you past POST.