Massman

Found her:

Performance leak by OBR (HyperFormance tech) http://pctuning.tyden.cz/component/content/article/1-aktualni-zpravy/23394-exklusivne-na-pct-lucidlogix-virtu-mvp-v-akci

Hello all, As you might have read on the Gigabyte Tech Daily blog, there's a new software technology by LucidLogix to be launched shortly. We all know Lucid because of the crazy "let's have nvidia and ati cards work together"-technology, but this stuff is more software than it is hardware. For now. The HWBOT staff is currently discussing its policies regarding this new technology. Given the size of impact it might have on competitive benching here at HWBOT, we also like to hear what the community has to say on it. Anyway, in short, what is Lucid MVP about? Well, here's a bit of lecture for ya! For more details on the specific approach Lucid took with this software, please refer to the whitepaper. In my opinion, a must-read to really understand the main concept of the technology. Whitepaper: http://www.lucidlogix.com/download/WP-Eliminating%20Graphics%20Pipeline%20Redundancies%20181011.pdf To summarize a lot of text, here's what's going on: 1) Lucid combines any IGP with a discrete graphics card 2) The IGP's task is to track down redundant frames 3) When there's a redundant frame, the IGP tells the discrete graphics card not to render it. This means that instead of rendering the entire frame (=100%), the discrete graphics card might render a part of the (~25%) or not at all (0%). 4) Because some frames are not rendered completely, more frames can be processed (20% rendering instead of 100% = 5x faster). 5) The FPS counter goes up super-dooper fast. So it boosts FPS, what's the big deal? As the PR states, the Virtu MVP is very likely to boost FPS rates by 30-70%. From tests, however, some reports state over 400% boost in FPS. A miracle? Or just cheating the counter? Neither. As Lucid indicates in interviews, the boost in FPS is not a real increase in FPS. Here are some quotes: Cheat! Cheat! Cheat! Hehe, I knew you'd react in this manner (since I wrote this post). Yes, on first sight it seems to be a cheat of the FPS counter. But, well, it's not. The 'false' FPS reading is not so much the fault of the Virtu MVP, but in fact ... the fault of the FPS counter itself. Virtu MVP does not directly address the FPS counter to make the performance look better. Actually, the Virtu MVP software doesn't 'care' about the FPS when doing its tasks, it cares about render resource efficiency. As you understand fromt the process summary, this new technology makes it possible to either partially render frames or skip them entirely. The cool thing about this is that the image quality does not go down and Lucid claims the gaming responsiveness actually improves due to a better balance! Now, this technology is in fact very interesting because it resolves efficiency related issues: because the IGP figures out which frames have already been rendered, the discrete graphics card doesn't have to render it again. Instead of doing double (or triple work), the IGP makes sure it's only doing the necessary work. In my humble opinion ... it's the future . Software ... or hardware? Now, here comes another tricky part! In the whitepaper written by Lucid, they mention different implementations of this technology. One of the ideas is to have this technology embedded on-chip. Practically, put it on the IGP silicon. That would make it a hardware-based solution, not software. Just imagine this technology being implemented on a next-generation Intel IGP, with the software being adopted as one of the instruction sets. That would mean you wouldn't have to install any additional software to have the FPS boost. It would be a complete hardware-based solution. Allow or disallow? There are a lot of arguments that can be made either against the software or in favor of it. I'd like to hear your opinion on this. Please, if you want to give your opinion, also use some sort of argumentation. Posts such as "this is total shit" will not be read. Instead, you might want to post "I think this is shit, because ...".

fehlgeschlagen ...

What I said is the give-aways we do are on HWBOT. I don't want companies to think they can boost their own Facebook pages through HWBOT by simply giving away a random graphics card. If you want to give away a graphics card at HWBOT, no problem. Put the card aside and we'll figure out a nice competition here at HWBOT. Just like we did when AMD offered us an FX-8150 directly. None of that "please like my page and you might win something" BS please. Thank you! (moved to offtopic)

We need development time. As usual.

Are you running the highest value of LLC ?

Hope you send the link to the original article. If you linked to here, obviously hwbot will be 'against asus' again

This is the layout of the 7950 DCU2 pcb: This is the layout of the 7950 ref pcb: Seems like the PCB is a bit different (which is normal for the DCU version). I think the problem is related to the weight of the cooler and the mounting mechanism. At least, that's what I get from the translation of the article. I think the author fixed the temperature issue by tightning up the screws holding the mounting and got normal results. It happens, it's not out of the ordinary. What would be bad, however, is these cards making it to retail with this design error. I think that's what the author is hinting at and why he published the article: he complained to Asus about this issue, got ignored and found cards being pushed to retail anyway. I wonder if it's present on the 7970 too. I assume it would be, given the similar mounting method?

Whoah, so close! Nice!

I still have this GTX275 Lightning that needs a fresh PWM. Shall I bring it to CeBIT ?

Holy shit: link Apparently there would be an issue when using the card horizontally (like in a normal system) and the contact between GPU and cooler. A French site tested the DCU2 cooler with a 7950 and, well, it's not doing too well. Slamms, can you check if you have similar issues with the 7970?

... when you start using a coaster to put your cold beverage on the coffee table because you know the dangers of condensation.

fixed.

How good PSC. Are we talking DDR3-2200C6-9-6 ?

Hm, looking better indeed.

Hm, valid point there. We do save league positions and league type together, so the we can update it retroactively.

Oh, right, missed that! Jups, makes sense now

... when a cardboard box is a worthy alternative for any computer case

Physics are quite low for 5.5G, I think. Even for v1.0.3 and only DDR3-2133. Is it possible one of the channels wasn't really enabled? (like on X58, only showing 12GB addressed in bootup screen and taskmanager)

I prefer good PSC over good Hyper any day of the week (except for sunday)

That's not a good moment to inform your friends you're a "masterbencher"

That's fine. Worse would be you actually going into the bedroom when she replies "but the graphics card is in here!"

Hm. CPU bottleneck?

... when you no longer care about the strange looks you're getting when buying a 1L jar of vaseline.