Is Nvidia cheating at Crysis???

The details are at the link but here is my summary of how interpret what Elite Bastards observed.

169.04 drivers scoring about 7% higher in the Crysis time demo by rendering reflections about 1 in 10 frames rather than the normal 1 in 3.

The question is: Driver bug or Nvidia purposely diminishing IQ to score higher in the most anticipated game/benchmark. I say benchmark because we know Crysis will be included in almost every GPU review going forward.
21 answers Last reply
More about nvidia cheating crysis
  1. It's not cheating to me until there is a notible difference in quality. If we the same quality for better frames what does it matter?
  2. Cant beleive they would bother for 2-3 FPS if it was a more startling differance then it would look a bit dodgy but i see it as a driver bug or just a porley implemented optimisation that isnt working properly like the review said in the conclusion.
  3. itotallybelieveyou said:
    It's not cheating to me until there is a notible difference in quality. If we the same quality for better frames what does it matter?

    You did read and look at the review didnt you ? Those mountains sloping of sure look like a noticable differance in quality to me.
  4. I heard they cheated for the benchmarks in summer or before that...and it was only for 3D mark not games...But it didn't matter I dont know why they did that because they're cards own anyways this year.
  5. I have not read the article but will tonight. But My very first reaction to the Crysis demo release day beta drivers was to question if performance came at IQ expense. I stayed with the WHQL's just to test them first.

    edit: wow nice read and very interesting results. Pretty bold statements made in their conclusion too about crysis demo performance and the inaccurate inflated results with these drivers.
  6. I wonder why nVidia would do such a thing when they are in the performance lead? Maybe they know something about 3800 series we dont? Or just overzealos nVidia programers? I doubt this change was directed from above, they were caught green-handed (nice Inq-expression ;)) before to risk it so obviously while the gain is too small. Look for example to nVidia recent cheating in HD decoding, the gain was 50+% at the expense of quality. THAT was deliberate, while I doubt its the case on this one. I may be wrong, those firms have no decency whatoever anyway ;)
  7. Crysis is the most Nvidia optimized game out there as far as I can tell. Nvidia hardware is going to win most benchmark I believe. I can say this because when viewing the SLI load balancing graph when running just the pre release demo, they are almost pegged out the whole time. This indicates a high level of game optimization to use both cards. This is higher than any other SLI game I have run yet.
  8. i wouldn't be too surprised... Nvidia did the same "optimizing" (Cough cheating) with the whole Geforce Fx5xxxx fiasco was mainly with 3dMark but also in major game titles to make certain things not even render at all.

    you can google it and everything hell i think toms hardware did a whole article on it if you can hunt down and find it.
  9. pauldh said:
    I have not read the article but will tonight.

    Same here, but sofar from what you guys are saying (and I trust EB cause I know a few of the guys) it sounds like the Floptimization debate will start again, and we can call this one FartCry2 in the name of the first Crytek/FartCry floptimization;

    So wat's the new name for Crysis? CrySissy?

    Yes it's a terrible joke, but here's a laughing baby to make you smile.
  10. That's just... creepy.

    I have the feeling it's just a bug or some trick that they are still working out, for now it's just a beta so I'm not too worried. Unlike the FX series, right now nvidia has the performance crown, so I'm guessing they are just trying to optimize for the game as much as they can. I remain hopeful, because it's a great game.
  11. Sure, it could be a bug they didn't know about, but it's possible the few fps were worth it to them and they did know about it or do it on purpose. We question why the performance leader would do this knowingly. For one thing, very few people will know and understand about these driver inflated scores compared to those who will look at and link to the benchies without any knowledge they are inflated. Plus NV can just say "give us a break, it's a beta driver" and cover this as a bug. Dumb move to risk IQ being discovered for a few fps, or smart move as the 1/4" longer bar graphs will make many ooh and ahh for some time to come? As TGGA said, it could very well be a hot topic/debate/arguement for a while now that's it's been brought to the surface.
  12. Ah Grape, not too bad a joke, and the baby is hilarious. :)
  13. Gald you like the baby, she seems happy, I think she just found out she's been given a Wii and Nintendo stock for her birthday. :sol:

    I think the thing that's funny about this is that the driver that is SPECIFICALLY MADE FOR CRYSIS is the one with this issue, and the previous driver didn't have this bug. I'm gonna leave the office shortly and then I'm now going to finally, yes FINALLY install Crysis and look at it up close, then read the article (to get the perspective you guys got having played it).

    The thing is nV could easily undo this by simply releasing a different version of their drivers, but I wonder how many revisions this would've existed in if it hadn't been shown here if it was only reported as a bug by users?

    They still have plausible deniability regardless there of because it could easily be that tweaking one thing broke another. But this type of change points to something else going on that broke what appears to be an otherwise ok previous driver. If it's a legitimate optimization it would be fixed and then rolled into a future update like Quack was, but if it's not you may never see the return of that same performance boost (at least not without that floptimization).

    I agree it's a good thing they found it before the launch, because afterwards would pretty much mean none of those initial reviews would be updated, and like you say everyone would point to them long after the fact.

    I wonder what driver Anand is going to use for their review?
  14. I should quote the whole messege as you made all very good points. The timing of this discover was not only good for ATI, it was critical for all of us for reasons you mentioned. Will there be a new beta driver release for the retail game that fixes this and drops performance again? Can they keep the performance if they fix it? Otherwise like you said the next few driver releases may have had it and all initial crysis benchies wouldn't be trustworthy. At least I'd have to assume EB and some other sites will follow up on this. Not sure about Anand though. .04's for GF8's and leave the HDxxxx cards out of 0xaa tests. ;)
  15. Oh, and about time with Crysis. Enjoy!!!!!!!!!!!!!
  16. Maybe they did this in Stalker too.
  17. Both companies have been doing this forever.....


    Not even worth the performance difference(not for the strechyness)....
  18. I wondered why the mountains were stretched, it looked kinda funny when I was running the benchmarks earlier.

    PS their page is taking forever to load page 2 so i'll go back to reading tomorrow night. (must've been digged or something like how the tweakguides server got too much traffic, this crysis stuff is popular lol).
  19. Quick update before heading home, Hanners got a reply from CrisRay (IIRC he was a mod for nVNews before going to nZone) in the B3D forums, saying they will fix this in the next batch of Beta drivers (always with the betas :sarcasm:);

    No mention of when nV was originally planning on releasing these (if at all), but it looks like now that the IQ has hit the fan that the fix will arrive in time for the game does.

    Speaking of such things, you should see Kyle showing his stripes by attacking Hanners, instead of commenting on the problem. He seems to blame EB for this, and says it's ok for nV because they were Beta drivers, not WHQL. Although it won't stop him from running his tests with Betas, look at his GF8800GT review, only WHQL drivers are for the HD2900. :heink:

    Look at Kyle attack the messenger in this thread;

    Funny thing is even after Hanners said he wasn't contacted by ATi, Kyle simply sloughs it off as 'well you've maligned me before, so payback' type of reply.

    My question is, even if Kyle was contacted by ATi about this, when was he going to tell the end-user / gamers? After the Crysis review had already been circulated?
    I also love that the article he uses to questioning Hanners and defending nV has a nice BFG GF8800GT ad right beside it. :pfff:
  20. TheGreatGrapeApe said:
    I also love that the article he uses to questioning Hanners and defending nV has a nice BFG GF8800GT ad right beside it. :pfff:

    Thats funny, right below that ad I see an ad for ATI/Sapphire.
  21. Yes interesting, since before that spot was occupied by MSI ads not the Sapphire one.
Ask a new question

Read More

Graphics Cards Benchmark Crysis Nvidia Graphics