Post your french link on whatever french site you frequent.

you had to say something right? well done!

yes, fuck off with your irrelevant link

/ignore

from now on, you may say whatever you want, hf

twat much

It seems that you hold a grudge for pointing out your not quite rational thought process.

As easy as hitting "Translate" in Google Chrome.

This is an english-speaking forum.

Just hit the translate button.

I don't have a translate button.

It's next to the brain button , oh wait

Try holding the power button for a few seconds then.

Crytek used the 2M € nvidia gave them, and coded tesselation in a way that it uses shaders on amd cards instead of the special tesselation units that the latest amd cards have, here ya go.

But dont worry nvidia are burning themselves, by always trying to build monstrous chips that consume a ton of power, have bad yields produce a lot of heat etc, and even on the latest gtx590 they dont have hardware over current protection, so you see people doing small overclocks and the cards are burning which is a shame for a card of that price tag

http://www.techpowerup.com/148722/Foundry-Del...-2012.html

http://www.hardocp.com/article/2011/07/07/e3_..._hardware/

I used to like nvidia, but when they started waving fake cards to the crowd and that G92 renaming rampage, blocking the use of an nvidia card for physc and amd card has main gpu etc and all these cheesy stuff they do, i started to pull away from them.

Talking about crysis 2, it was ruined because of consoles, they had to change the style of their maps to suit console players and the very little hardware resources that consoles have.
Crysis 2 Multiplayer is shit, no power struggle mode killed it, it was a very enjoyable mode to play with friends on a small lan.

well, when ATI was nowhere in the '90 and when they only released utterly slow and shit cards, I really wanted them to be successful (I remember they gave me an early sample of rage3d on a GDC, it was usless, no joke) But since they are on top now, I hoping for nvidia to get better.. but this might change my mind.. this is just too much showbiz :)

well nvidia cards are usually overpriced at release date, look at the gtx590 100 euros more expensive than the 6990, gtx580 200 euros more than the 6970, then you have the sweet spot 2*HD6870 150€ each card and rapes the gtx580 everyday. even at that point AMD/ATI prices are better, although nvidia cant go down and wont go down, they just need to rethink their strategies, and they have the potential there because their user base is very strong and still better than AMD, but they are loosing market.

I really like CUDA tbh, even if it's not the best I can think of, it's still something I would miss with an ATI card.

well ati has something similar that uses open CL, u can used that to do FEA etc and it is called AMD Accelerated Parallel Processing (APP) technology. This might be interesting:

http://semiaccurate.com/2010/07/07/nvidia-pur...physx-cpu/

also this is the main article:

http://www.realworldtech.com/page.cfm?Article...43&p=1

yea, I know about them, but thanks for the articles:)

ps.: where did you read they got 2Mil, did I miss it?

i read about that on techpowerup.com

Edit: those articles are about physx

yea, i know it's about PhysX. I never liked it anyway, I always looked PhysX as some kind og a gimmick marketing thingy:)
I think even CUDA is "not so fast" anymore since we have Sandy Bridge, but it's a complete system, and a well polished one.

Cuda is still really fast! Today's gpus are extremely powerfull well beyond any cpu, just a prime example, 1.5 years ago a japanese company i believe, ordered 1.5K AMD gpus to build a super computer. I believe that in a future we can go in two directions, for the low power devices like tablets, notebooks and even some of the cheaper desktops will use something similar to today's AMD Fusion, and the High performance computing machines will have a solution like nvidia tesla, and in this last case the cpu will be like a northbridge for these cards.

Sadly i don't see much future for Desktop PC's :( let's hope I'm wrong.

Yea CUDA is awesome, but sandy-Bridge and the followers are all very powerful processors, dunno how TESLA and CUDA will cope with them in the long run.

The usage fields for GPUs and x64 processors are quite different. GPUs are number crunching beasts, but are terrible at handling strings etc. while the x64 CPUs are general purpose computing units that have a wide usage field. I would be surprised if any x64 processor in the near future comes close to GPUs in terms handling number calculations.

Toms hardware did a bit of testing on this for password cracking. In the case of "Compression: RAR Normal Encryption: AES-128" i5-2500k (Sandy bridge) did 213 passwords per second while 2xgtx 570 in SLI did 14 605 passwords per second, which is quite a big difference.

CPU
GPU

yes, I not meant they are faster now, I just said that I'm not sure about the future CPUs (or APUs).Think about the speed of Quick Sync for example. Nobody had any idea it would be that fast.

Also in the last 4-5 years. I see TH a little bit like how I see Crytek in this thread. Yes they still one of the best and they do great reviews and tests indeed. But I mostly feel they are doing things when the industrie ask them to do it. If the sale of the hard-drives. cases or whatever are falling for two weeks, you can be pretty sure there is gonna be an article about them on TH:)

I just took the links from tomshardware cause it was a good example. A better example would probably be "folding at home" but is harder to get accurate numbers.

GPUs are dependent on parallelization to perform tasks at extreme speeds. Something like video encoding doesn't lend itself well to parallelization as the frames depend a lot on each other. Number crunching where you can split the tasks into hundreds of tasks that can be performed in parallel is where you get a major benefit. A single threaded program would run faster on a 4ghz dual core i5 than a 2.8ghz quad core i5, it is all about using the right tool for the right type of program.

The main advantages of GPUs are extreme memory bandwith, parallelization that CPUs will only be able to dream of for the next 4-5 years, number crunching.

The main drawbacks are: slow when task can not be parallelized, terrible at handling strings (unless that has changed in the last 2 years), not that useful for things that do not involve number crunching.

I believe it would require a major architectural change to todays desktop CPUs for them to reach the number crunching performance of a GTX580 within the next 2-3 years, on a single die. As long as CPUs are meant to be able to do everything in your computer, then they will not be able to do everything great. You will see some more functionality, that is commonly used in a lot of desktop software, be moved into hardware. The difference here between putting something like quick sync into hardware and using a software program to perform the video encoding is that the hardware is fixed and takes up die size while the software can be changed and is free in terms of transistors added to the die. There will not be a huge amount of specialized functions added to the CPUs. CUDA has always been best at solving mathematical problems, that can be run in parallel, and it is up to the programmer what kind of problem he wants to solve, not the chip maker. This is why CPUs will not touch GPUs for a long time in the GPUs specialized field.

I dunno what is quickstep, If you meant Quick-Sync that one is indeed a specialized thing, but also uses the CPUs multimedia (SSE+AVX) parts a lot.

While as I already mentioned I really like CUDA and agree with you for today case, still, your example is a little flawed, since you need a complettely different codepath for a GPU vs a CPU.
If you would run a code made for the Sandy on the Fermi, the Fermi would struggle like hell, while the Sandy would also cry on his knees if you would try to run the 500 thread code optimized for Fermi on it. (little weird example since they can't run each other's code:D)
You can't compare this kind of things like that, you need the same application optimized for each, to be able to reach a conclusion. Future CPUs from Intel and AMD will all have powerful GPU's and new instructions (AVX2) inside them, imo don't underestimate the possibilities there:)

Folding at home is a good example, as it has different clients for CPU, AMD GPUs and Nvidia GPUs that all perform the same task. Same applies to the tomshardware tests as the programs were made for their respective platforms and perform the same tasks.

While you can put lower end GPUs on the CPU die, you can't put high end ones due to the transistor count, power requirements, TDP and cost. A high and graphics card has 2-3 times the transistor count of a sandy bridge i5 CPU, which is what is needed to have so many parallel processors.

That's the point, it's surprised me as well, just look it up.
In folding@home the gtx580 is doing less than 20k ppd, while 2600k sandy can do more

edit: the first answer for the question of this guy might give you some insight

i have had a terrible experience with ati drivers. those ppl seriously cannot program proper drivers.

they are getting better, they changed the entire driver team, but still not as good as nvidia.

ATM, they got worse drivers but much faster Rasterisers than Nvidia. The only thing Nvidia is better besides drivers is tesstelation. That's why they processing the whole sea under the city for nothing, or that's why the street is built from triangles smaller than a pixel, for no apparent reason .. the whole thing is happening to overcome and beat ATI. If they would use PC hardware as optimaly as they do on consoles, a gts450 would put out crysis2 fullHD with 60+fps for sure. But if ppl only buy hardware in every 3-4 years, the PC gaming industry with all the dedicated hardware market would collapse in no time, so I can understand that, it's "just" business, they need continuous compsumption.

The ugly thing is what Crytek did there.. The fact is that they deliberately, and intentionally slowing down performance by 40-50% (according the data obtained by PrefStuido) is something what should be punished real hard imho.

nice links

that's the mainreason i'm still using nvidia cards more than 10years.
You will always have problems with ati-drivers here and there.

same here:) The lower power consumption is very tempting but I'm not really into playing BSOD in high-res;)

for games, yes.

for anything that uses just the stream processors to do massive amount of basic operations in parallel, the ati cards are already twice as fast as nvidias. plus they tend to be cheapers than nvidia cards aimed at the same market segment. so for hash cracking etc ati is the obvious choice.

edit: i am not sure how the brands fare as rendering farms. would actually like to know whether people mass rendering photoshop or 3d movies use ati or nvidia

"or anything that uses just the stream processors to do massive amount of basic operations in parallel"

such as?

mostly encryption related stuff, but i guess folding@home would be the case as well

so basically useless shit that nobody cares about.

i see what you did there

Bitcoin mining,

hd 6850 (150 €) is as fast as 2*GTX580 (~1000 €)

http://www.esreality.com/?a=post&id=2116231#pid2114337

kinda like yourself then

Is it profitable after you pay for electricity (let's assume that bitcoins are really worth what people say they are)?

well bitcoins price fluctuates a lot, so i wouldn't risk it. Well a few months ago it was profitable, but since mining is becoming geometrically harder i don't know if it is profitable anymore.

So this argument is without merit.

but i see people doing mining clusters with 9*HD6990 and these cards suck a lot of power, so i guess it should be profitable, but im not sure.

have you tried Creative?

firefox has "my translate" from google if you search addons for translate

Crytek s'étant focalisé principalement sur le côté multi-plateforme de Crysis 2, et donc sur le support des consoles, la version PC a été quelque peu délaissée par rapport à ce dont le développeur nous avait habitué avec Far Cry puis avec Crysis premier du nom. Il faut dire que les consoles commencent à accuser un retard technologique de près 5 ans sur les cartes graphiques modernes. Certes elles peuvent être exploitées plus efficacement sans la lourdeur des APIs PC et peuvent se contenter de résolutions inférieures (720p), mais cela ne permet plus de compenser le gouffre qui s'est creusé, d'autant plus qu'elles ne supportent pas les dernières fonctionnalités telles que la tessellation programmable (son utilisation est limitée sur la Xbox 360).





Oh ok, now I know.

1click and you get http://translate.googleusercontent.com/transl...Nn_IDqzxpg

yes, and they even told us that they are doing it on the PC and will port the game to the consoles...pff, that's why we had to disable auto-aim in the PC version;)

this topic needs more french links

I thought it was important gaming news, just wanted to share. Btw, (like most of the ppl), I was able to read it, even if I don't speak french.
It's ok if you don't care, sorry for wasting your time:)

You should have posted a google translate link as well for those few who still don't know how to use google translate.

Anyways, thanks for the link. I do believe crytek did the DX11 patch in the way that they felt was the most rewarding for them in terms of cost and time. Everyone that bought AMD cards after the Unigine heaven benchmark became standard in reviews, knew of AMDs weak tesselation performance.

I didn't know you can link to translated pages, I had no idea google made that available:( sorry, ofc I learn something every day, and thanks for the info.:)

Rendering massive amount of things (again it's 40% of the frame-time in average!) what the user will never ever see, and doing it just to lick nvidia's ass for 2 millions (which is pocket-money for real).
It's everything but not "most rewarding" for sure. They cheated on their costumers, this can't be justified with anything imo.


edit: what you said about ATI is weaker at tesselations, is true.... , but it would still compensate for that at the rasterization phase of the pipeline..
The nasty thing is that they process all those shit to make the ATI card struggle, then they just run a geometry shader on the whole scene before rasterisation which looks up and removes this fake things. If they would let the geometry there for rasterisation, the Gforce would be still slower... this is the "bad" part of the story

Tbh I could care less if people believe Crytek purposely tried to screw AMD or not. I doubt there are two separate render paths for Nvidia and AMD, so as long as both cards perform the same things, then I do not see a huge issue. Few games I know of come close to Crysis 2 with 1.9, dx11 and texture patch in terms of graphics and performance in Crysis 2 is quite good compared to the graphics shown.

It is primarily the tessellation on bricks and a few other things that kills the framerate on AMD cards. Walking along a hallway full of bricks, on a single GPU AMD card, will bring the framerate to unplayable levels with everything on max in 1080p. Currently the only way to get around it is to reduce maximum tesselation to 8 or 16 in the catalyst control panel. This is similar to what benchmarks like unigine heaven have shown.


[edit]:
Google translate link below:
http://translate.googleusercontent.com/transl...LhP1BQKgGg

We see this thing differently:) Halving my fps under dx11 (whatever card i have) just because Nvidia paid for it, is not OK for me, and it's something the PC games must be freed of.

Just do what I did, set tessellation to 8 in catalyst control panel. It will give you quite a bit of the tessellation and give you good frame rates. At 1080p with everything at max the frame rates went from 30 to 40 in a room full of bricks by doing this and the graphics quality impact is only minor. Running a 6950@900mhz with all shaders and texture units unlocked (essentially around reference 6970 performance).

I use nvidia cards, and I wont play with anything below 90-100 fps:) You could say that I should lower details than,. but crysis is a shit game imho, and the only reason I would play it is the visual entertainment.

i guess you think it is shit because maybe you dont play it the fun way :>

http://www.youtube.com/user/NanosuitNinja

This guy play crysis the proper way, and it is a lot of fun to watch!

a small video he made

http://www.youtube.com/watch?v=6GfnwmZXhpE&am...ideo_title

nah, the only way I play any game is the fun way xD

thx for the links btw;)

what kind of temps you get on that card? idle and load, my 6970 is like hell on hearth no need for heater on winter.

It comes with a factory mounted custom cooler and I run with a well ventilated case so temps will most likely not be comparable. Old case (mediocre ventilation) 870 core 1250 memory and 6950 level of shaders it ran at aprox 65 degrees (42 over room temp). In my new case it was 58 degrees at the same card settings. With OC to 900 core, 1325 memory, powertune +20 (because more shaders draw more power) and all shaders unlocked it will hit about 63 degrees. Temps are after being at full load for 15+ minutes. Measured by putting the temp monitor on second screen while gaming and check it after playing for 15 minutes or more without leaving the game or going to menu.

On Idle it is around 44 degrees, due to running dual monitors. Cards use quite a bit power more when running dual monitors than single monitor. With dual monitors it runs at 450 core 1325 memory at desktop, single monitor it runs at 250 core 150 memory 36 degrees.

Room temp for all figures is between 23 and 25 degrees.

With a stock cooler I would most likely hit temps somewhere between 75 and 85 degrees Celsius.

not bad at all! i need to run with the side panel off and depending on room temperature, usually between 26 and 30 celcius i get 62 - 67 ºC idle temperature and on load it get as hot as 92 celcius under tessmark 0.3 test set1 in the lowest settings.

A curious thing i noticed: under tessmark 0.3 if i use the map set 1 and make it run at 32X tesselation and 8XAA i get the gpu stressed to 100% and about 85ºC max temp with the fan at 37%, gpu always at 880MHz with a max VDDC current of 63.8A. I get about 57Fps on this preset.

The interesting thing is when i put the tesselation on the lowest setting the temperatures start to sky rocketing and the fan starts screaming with a 99% gpu load and max VDDC current of 127.8A and constant 100A. i get about 435Fps on this preset.
Can it be the tesselation units bottlenecking the shaders?
Other thing i noticed is that the ati drivers are shit for openGL4.0 tesselation, and tessmark uses openGL4.0

Edit: i run my monitor at 120Hz, so the gpu idles at 500Mhz and mems stay at 1375MHz

Why would you want to run with side panel off? Most good cases give equal or better temps with case closed than case open due to airflow. Just make sure you have good intake fan(s) in front and exhaust in the back/top and you should have ok airflow. Having a side fan helps a lot on GFX temps though. Have a 200mm side fan that blows cool air over the GFX and MB, which probaly lowers GFX temp by 3-4 degrees at full load as the GFX cooler always has cool air at the intakes.

AMD is notorious for bad OpenGL performance and I doubt it has changed since almost all games nowdays run on direct3d. The 8xAA will probably bottleneck your GPU though along with the tessellation. Try without AA and only tessellation. I do suspect AMD to be bad at tessellation once it passes 16x though.

The 120hz shouldn't matter for the idle state of the GPU. Mine runs 120hz and the GPU still goes to 250/150 if I only have 1 monitor attached. Anyways I wouldn't worry too much about the temps unless you have overclocked. It should be able to hit close to 100 degrees and still operate OK so 92 degrees should not be a problem.

my case is pure crap! well mine goes 500/1375 @ 120Hz and 250/150 @ 60Hz. @ 250/150 it idles @ 43ºC