Unleash the full strength of HD 6970M CrossfireX - Hybrid VBIOS to unlock the virtual cores - a new milestone?

That's my guess as well. So much for being excited about this.

 

Keep in mind that I looked at this posted "A01" hybrid not the A03 hybrid that's claimed to have enabled HT-like behavior, so don't count it out entirely.

BTW, re: Voltage..

These are supposedly "virtual shader cores" you are enabling, and it is compared to HT, so in that context, lets assume FurMark (or whatever high-heat-output GPU testing program you use) represents heat and power draw at or near true 100% GPU utilization.

So then, when you play games, the heat/power draw doesn't get as high as it does in FurMark. With this "HT-like" stuff in place, you could reasonably expect all 3d apps/games to now reach Furmark levels of heat/power draw, as the GPU would be near/at true 100% utilization.

 

Haha. I hadn't even looked at the batch file talked about here yet...
http://forum.notebookreview.com/ali...irtual-cores-new-milestone-9.html#post7875759

Surely some chinaman is having a good laugh The batch file repeatedly pings localhost and changes directory to C:\ while echoing stuff to the screen to make it look like it's doing something.. then it does this:
atiwinflash -unlockrom 0
which unlocks card 0 for flashing.. then
atiwinflash -unlcokrom 1
yes, that's a typo, meaning this command doesn't work, lol.
then it flashes the "modified" 6990 bios to both cards.. which i imagine card 1 doesn't really get flashed since there's a typo.

Someone should let him know how to implement "sleep" in a batch file so it can look a little more legitimate vs. pinging yourself repeatedly varying numbers of times for varying pauses, lol

 

Lol. we just need the clevers that say " I told you " thanks DA_G

 

Lol speedy, since you asked for it:

"I told you so!"

Now lets get hyperthreading straight shall we? (I remember doing this way back when I first joined)

Hyperthreading (HT) is useful in a general purpose processor like a CPU, it deals with lots of different types of data and each core will have many kinds of resources for each thread and when a core is designed for HT it will generally be widened to ensure it has a few extras to help along (a lot of work goes into the start of the pipeline).

Now when a thread executes it wont be using all the resources, therefore if you are clever you can weave in another thread to use those unused units so more of your core is at work.

Now let's jump over to the GPU. This is a different beast, it gets fed the same specialised data over and over which means it can be tuned only to have the units it needs, no extras are put in as they would be wasted. So we have lots and lots of "simple" narrow units designed for one task. There is no way to hyperthread those units, in the simplest terms there is no "room" to squeeze two "threads" in.

So as soon as they started throwing a term like HT around I knew it was BS.

 

Yeah, that was pretty much my first blush, too.

But, I figured, if I looked at the BIOS and it had been entirely re-compiled, with a different config.h and the modified source, sure, anything's possible.

Then I saw the BIOS and sadfaced

In my Windows Mobile hacking days I ran into that too, Chinese guys would always fake new firmware revisions, I never understood what the demand was for that in China, but apparently there's enough demand for them to do it to vBIOSen too..

 

The only unlocking would be some sort of magic compiler that could fill all VLIW units going at all times. That would be impressive.

 

I didn't read your posts carefully and my initial guess would be you've downloaded some wrong (and probably out-dated?) files.

You'll need to have replied to the post of the first link in my OP to be able to see the download link for the VBIOS for 6970M CF.

By curiosity I downloaded the correct file, with filename "Alienware_M18x_AMD_Radeon_HD6970M_虚拟核心VBIOS测试版本VirualKernel_Beta_v1.0.rar" (File size: 12.59 MB). It's encrypted with a password. I'm not sure about the length of the password (whether it's applicable to crack it with some GPUs).

These download links would be a pain for those who doesn't speak Chinese. I've uploaded the file to MU, just for those who's interested:

http://www.megaupload.com/?d=8YFT5EP9

I don't have M18x to test it. I'm just watching. It looks like if you wish to test it, you'll have to ask for a serial number from the author. Also, according to the author, in order to have the Hybrid VBIOS work, you'll need to satisfy 3 conditions:

1) The VBIOS
2) The modified driver
3) A modification to the Windows Registry

These have been stated in the FAQ.

 

I grabbed the link from here: http://forum.notebookreview.com/ali...irtual-cores-new-milestone-9.html#post7875783

Which I presume was downloaded by someone who replied to the OP?

I'll check those links now..

Edit: as you said, passworded.

Well, the modified driver and registry ain't gonna do jack if the vBIOS is bull like the one I looked at, unless the modification is nearly entirely in the driver (which seems doubtful)

Again, i'm only commenting on what I got a chance to look at thus far (which is not what's in that passworded archive) - so my judgement remains reserved, but my skepticism level is somewhere in geosynchronous orbit..

 

I don't think so. The correct file is supposed to be 12MB large, which is supposed to include the VBIOS, the modified driver and the registry file.

 

if so...then who uploaded the first round of bogus links then or that was just part of the puzzle?

edit:
now that is looking far more like it will do something.

 

you guys do realize I could have built a password protected rar file with a bunch of files that looked like they do something right? This is one of those I'll believe it when I see it things.

 

im trying to use some recovery password software on this .. but no luck lolz.
i guess he used chinese characters ^^

from the look of the content inside, looked like this one have potential

 

"benyouhui" is the notebookreview-equivalent in China. The author of this Hybrid VBIOS thing is the super moderator of the forum. I don't think it does more good than harm if all this is just kidding. That site doesn't need to fish traffic by such tricks.

The author and those who's invited to test it have mentioned that it works great in 3DMark11 but can cause lag spikes / stuttering in games.

 

huh,....you seem to know quite allot. you a long time member over there?
and if so...why no vantage bench?

side note: with that many pipes running...i would guess it's because it's now staved of voltage. speculation of course.

 

Yes when I started to play with laptops I was in that forum, much earlier than my join date on notebookreview Now I play a lot with desktop computers so I no longer visit laptop forums that often.

It looks like the author has no intention to invite more people to test it.

 

nothing wrong with desktops...even though none of my boards are in cases... just the servers

oh well...easy come easy go.

 

well I guess we'll just wait and see... meanwhile, i flashed my 6970Ms to 730/1100.

 

Which software did you try? I've tried igrargpu on my GTX580 but it doesn't run; then I've tried igrargpu on my HD 5850 but it seems to take forever.

 

well it doesn't matter ...
the problem lies in the password that contains Chinese character not Alphabet!

yea all we can do now is wait and see...

 

The author just declined to invite me to do the alpha test. Says a company is offering a good price for it, and it may need confidentiality.

 

He's either in denial because it doesnt work at all or hes in denial because it doesnt work at all. I'll believe this once we have some real proof not some screenshot. This is just like that marketing fail of the 6990M where the chart says its 25% faster.

 

I do believe in the software and if I was the developer I would not let anybody I don't REALLY know test it.

I would focus on getting the rights to the idea and selling it to ATI/Nvidia.

 

Well it was a theoretical 23% faster....so half that and you will have real world expectations haha

 

Yep, I think that's the case, if the whole thing is true. If they made a vBIOS modification that forces the GPU to work 100% during benches/games (similar to furmark's level of stress), they need to make sure their idea is protected and only then maybe offer the code to beta testers.

 

here is what I see:

1: the screen shoots always show 6990m on gpuz and he say is because the drivers , I don't believe that he has real 6990's

2:he say the 6990m is supported however he don't warantee performance increase or degradation , I don't believe that he has 6990's

3:when I download the bios and I open with RBE it say the bios has been pre modified with RBE already and that's what DA_G saw too so that is just a 6990m bios with upped clocks

4:Like John say they don't show a vantage run because you can't tweak the settings on CCC for that benchmark so ?...

5:I don't know why people like to waste other people's time with BS like this

 

Agree , this is how I got my GPU fried :S trying to roll back everything , but don't have time to do it, laptop is not with me.

 

discuss this with expert these days
conclusion: simulate muti thread like this, "hardware timing" invoves a lot(just like ddr3 cl7 cl9 cl11), if ati/nvidia were able to do they should have done since intel ht was published years ago

btw im trying to unlock true stream processors...........

 

i ran the program like phantom, but i dialed down the clock/shaders before i did it. all it did was overclock my cards, no big deal. would love to get a better look at that 12mb file.

 

so when is this going to be released?

 

all these video cards out and this is just now coming to light?

 

Are you going to solder a 580 desktop chip to your 580m ?

once you finish your 580m is going to be a Frankestain 580 lol.

 

I hope not lol

The desktop chip is like half the board size of a MXM card. Those things are HUGE

 

that's just the cover check the pics on post #128

 

AMD already has multithreaded rendering in the works:

Microsoft DirectX® 11

And I think there was one game that already took advantage of it. What this guy is doing is complete BS for several reasons (which I don't have the time atm to explain)

 

The multithreaded rendering shown in the screenshot there is implemented already and is for the CPU portion of the renderpath, but the "hyperthreading-like' stuff discussed here is for the GPU part, rather than the CPU part. And yeah, agreed on the several reasons for calling BS

 

I havent followed this closely so I could be way off base. But when I saw the clocks 800/1000 and the benches...I just said "these are the scores I get when I OC to that basic OC." I don't see what the hype is? It just looks like overclocked videocard bios that uses some other things to make it appear it's doing things it's not. That is my Nub layman terms of what I think is happening.

 

i say...run vantage.
and furmark pretty much said..."look, you are having issues at 23 fps, what makes you think your going to do better" lol

 

Haven't been following this thread. When will it be released?

 

never. im betting its just fraud

 

Guys this has been 100% confirmed fake.

 

Shocker!

Can this be closed now? lol

 

Got any proof? (LOL I know it sounds funny) So what lead you to this conclusion?

 

When I first read the title I laughed my butts off. HT or anything like HT first needs hardware level design support. This means if AMD had any such logic on the chip they did it to have the feature from the get go. Not for some person in a forum to come up with the idea for a chip he/she never even designed.

Whatever DX 11 related multithreading that maybe going on with supported apps, you can only do so much with driver tuning.

Like most have point have pointed out on this thread already, if it was in the drivers then it could be possible to expect some sort of performance increase in certain benchmarks while suffering elsewhere. Pretty arrogant of this person who made these claims of having HT like modifications for his cards, when AMD who have a bunch of engineers working on this never made such moves or claims. They would be the first to think of all radical ideas to check for performance increases. Its always a trade off, you need to have the best possible performance across the board not just in 1 or 2 applications.

I don't specialize in driver side optimizations, I work on the layouts and actual asic design. So the moment the claims of Vbios modifications and some sort of HT references were made, I knew this was snail oil stuff. It was fun reading the whole thread though just to see how the grand claim was put forward.

A software/driver claim would have got people going maybe for a little bit longer before getting caught bluffing on that one as well.

If he can make AMD's drivers alone work 25% better than what their current driver development team can do on every title, every benchmark out there, then he will be given the top job right away. But being an engineer myself I can categorically dismiss the notion that AMD's folks couldn't get their driver model thought out to get the best out of their own architecture. lol I mean come on who does this guy think he is?

Besides this isn't a one man show, no matter how brilliant one might be this sort of work requires a team of highly capable engineers.

 

Yup, but the idea of utilizing GPU 100% is interesting. If Furmark can stress it to the max why can't the developers make the necessary adjustments for games?

 

Because it's easy to have lots of repeating instructions that stress the card as much as possible, but hard to actually use that in a game.

AMD use MAD instructions to give the power of their chips (In terraflops) but no game can be made out of those.

scook, the bios was examined and only a couple of values have been shifted, about enough to incorrectly report the number of shaders

 

or
it could have been manipulated to not force certain textures to be rendered in the quality that they should..thus giving the illusion of more fps. like with the 4870's and AA...(how you get higher fps when enabling AA/AF)

which vantage and furmark seem to not be really effected by this.

 

Depends a lot on the scene and the logic units required to render that scene. GPUs are smart enough to truncate data that wont be rastered, so we do see often some units are at rest therefor using less power.

Furmark is designed to have most units that are on, in a constant state of operation which leads to higher than usual power draw. But even Furmark might not be able to use every single logic the GPU has, this is because its not using all kinds of logic that a regular game does.

Another way to put it, we might never be able to use regular software or even stuff like Furmark to reach the absolute peak power draw. Its either all of it but with gaps of idle time in between OR most of it with constant operation.

 

sorry to say..but furmark is the "in" software for gpu testing. it's one of the main reasons nvidia invoked the ocp/ovp on the desktop models. for fear of it burning the vrm's.

 

Load on VRMs has no direct relation to whether all elements within the GPUs are being used all at the same time.

Cost is one of the factor that leads to choices made on how many phases and what components are used to make the VRM circuit (inductor, capacitor, resistance etc). So its natural to expect board partners to not design the VRMs with power hogs like Furmark in mind. But Furmark or other programs don't really use all the logic on the GPU at the same time. If it where ever possible to use everything at the same time the load will be much higher than what furmark can really draw.