Yep, Fermi is noticeably better than the 5000 series when it comes to minimum framerates; especially in multigpu.
Well 6870 at 250$ is a 460 counter imo, they did make noise about it as the 460 was coming out so ...
I think previously both companies used to come out with stuff around the same time so it seemed like it was a counter but really one is always trying to keep the competitors newer products in check with new ones themselves, marketing strat I guess ( dont take too long to risk making something super and lose market share in the meantime and if it doesnt come out right ..... fermi ? :p:)
Right now nvidia are trying to make something that will be a performance leap so they can sell their big chips and make decent money forcing amd to undercut them.
Amd are probably focussing on keeping their good start and looks like they have a big bad beast chip of their own incoming as well so really right now its looking good for amd on many fronts they just have to stay on the ball.
nvidia have to get close and stay competetive and see this out and be on time and faster than amd on the next node.
This is just what I think is going on , ofcourse with the entire market situation last year this year we havent seen the 2 getting into a price war. But I dont think there was some agreement to that effect just common sense amd was happy with how they were doing they didnt want to push their luck.
Thats why the whole 6870 thing , its not a spiritual succesor to the 5870 which was a good jump on the 4870 and good enough to make a x2 card out of , 6870 is being named aggresively to steal 460 sales.
Its going to have the performance to do so , its going to be named such , and its a small chip amd can price it aggressively.
Anyhow in a week it should be on the egg and we will see how this chapter unfolds , lets see if it takes the market by storm.
That's exactly my point, they didn't get into a direct competition, and we didn't see any price war either.
I'm didn't say, and don't think they have made an agreement, at least not a direct one, but both has been reluctant to counter new GPUs in the right time in the whole of last year.
The result was very high GPU-pries (in both camp) last year. I guess they both like it, and therefore I'm wondering if they are going to continue this " peace" with 6 months delays to counter each other. It will be interesting to see how they counter each other in this round.
The first part it is unproven. The bold part, no they were not.
http://www.theinquirer.net/inquirer/...i-price-fixing
http://www.tomshardware.com/news/Nvi...rust,6421.html
No it is not, see above.
The suit was settled for 1.7 Million. Much less than dragging it through the courts would have cost. Quote from Toms HW "A settlement does not admit guilt, as a quick settlement could be beneficial even for an innocent party."
The only difference was that back then they were trying to release the 400 series (a lot later than what was said). As for the 500 series (as some already posted) they doubt it will be release any time soon. Or believe it would be nothing more than a refresh of the current arch.
I want to see "something" from nVidia soon after the release of HD68xx.
Rather a new series an GPUs, but a refresh or a big price-cut would do too. Then we know they are trying to fight back, with what they can. Otherwise, I guess we can expect that AMD will do the same for nVidia to keep the "peace" when they finally do something in 6 months or so.
Roadmap from 4Gamer.net: http://www.4gamer.net/games/099/G009...013/SS/001.png
They say GF110 products might have 576 CC and there's the possibility of a dual GF112. What's up with the 320-bit memory interface for GF112, anyway?
http://translate.googleusercontent.c...lEaUtiG803lpYQ
nVidia has to bend the law of physic & time-space continuum if they plan on building a single chip with 4 GPCs 768 SPs 128 TMUs 64 ROPs 512 bit based on GF 104 design, using the same TSMC 40 nm process node. Reduce it to 3/4 of it, and we have a bussiness, but i think it would arrive at late Q1 2011 at the earliest. They should prove it first that they can build & sell in a good quantity, full fledged GF 100 and GF 104, before talking so much nonsense about slideware products.
This guy also did a similar roadmap for Ati, and look how it turned out:
http://www.4gamer.net/games/110/G011...093/TN/002.jpg
The best sign when you know they are going to be late is when you start reading a lot of cuda/physx again. No different when they were late with the 400 series. Oh yeah, they suppose to be releasing a complete re-write of physx in 2011 (1st half) were they suppose to improve CPU functionality. We'll, see how well that goes.
This thread can be closed. Not news, not even rumors, just pure fantasy. Other forums replicate the "rumor". Please , this is not april fools.
Via is preparing a gpu to obliterate HD6990. You read it here first.
http://vga.zol.com.cn/199/1997670.html
First "review", though only vantage scores given.
I welcome any attempt by Nvidia to bring us faster cards. But I'm kind of amused at the concept that the problem with Fermi is that it isn't big enough.
Not impressive IMHO
http://img413.imageshack.us/img413/4516/captureqz.jpg
Its less impressive that he likely pulled the numbers from somewhere for the older cards and the scores for the 5870 and 5850 are substantially higher.
Newish scores(they are even higher now).
http://www.guru3d.com/article/asus-m...inum-review/19
1000-3000 points higher.
on 3dmark06, the 5870 should score much higher than that.
I thought only Nvidia was capable of this naming stupidity, but now i see AMD is too.
They think we need bigger numbers to keep buying. Not just a bigger series number, but further along in the series.
I can just imagine the marketing-critters in the Nvidia marketing department when they were discussing going from 9xxx series to a 3 digit number... i bet they were losing sleep over that, fearing people wouldn't buy as many because of the lower number.
Lying to your customers indirectly by making a 6870 inferior or about the same compared to a 5870 just pisses me off, but oh well.
Can we concentrate discussion regarding AMD's incoming new gen cards in the appropiate threads ? There are several already and this one isn't included.
somebody please close this stupid thread
Oh i see, now everyone begins to agree that AMD is doing something really stupid and the fanbabies want the thread closed rather than we bash them for being :banana::banana::banana::banana:s. Heh really funny if 6870 is barely faster than 5850 which means with the crapalyst factor and failfire and ofcourse now they come with amd on them so that just screams MOAR FAIL 460 1gb is still gonna be the card to get.
:D
hd 6870 is the gtx 460 amd counter-part not the gtx 480 counter part. thats the hd 6970 and there is still no word on that yet.
Release date of the gtx 580: 21/12/2012 ( wait you really thought the world destroy by itself this date , nah man ,the gtx 580 will destroy it)
Nah dunno I hope AMD get a good beastly chip out 2nd gen DX11 they have had time , let it be hot , let it be big , let it be a little pricey but if it beats 480 handsomely like as fast as 470 sli OH BOY !!!
yeah NVIDIA have new core:up:
GF110
GF112
GF119
http://img219.imageshack.us/img219/1853/001af.png
^^ If that slide is true, it should be good. Getting sick of my GTX 480, and I'm also curious what ATI's 69xx brings as well.
512 bit 768 core @40 nm?
nvidia going for a one chip per wafer strategy?
Nvidia better pulloff AMD like magic on this 40nm node if this 768Core monster is true, amd did it on 45nm with cpus, but they have more control over the process than nvidia...
But for the moment it looks like theyre going with a move from too big core to a terribly too big core.
ROFL .... best line ive heard in this thread
btw that slide looks as fake as it can be and 2nd ... GF119 ??? LOLL
ps: it just look like more bullcrap made up by nvidia to go agains the launch of the 6k series .. get people to wait more .. its an announcement of an announcement ... good job green goblins ... good job
all those charts that pops up all appear at a time that is obvious ... its fake .. move on ... lets stop believing those fud post please ... there is no way nvidia is so stupid to try and make something even more complex then a full 512 shader gpu ...
I got bored and poked photoshop.. :shrug:
LOL nice :P funniest version of this image as of late.
LOL are we there yet ... LOLl why cant ice cube hold that green screen
from that slide the gf110 looks like a gx2 of a gf112
and GF119 doesnt make sense at all ...
its the "119" that doesnt makes sense.
yes it does make sense ... but half the memory controler memory will it not hinder the perf so much that the 5450 would be close in performance .... and price wise??? will they pull the same price as a 430minus a few $ or cut in half ????
still 96 shaders .. just like gt430 ... so why refresh it with half the memory controler ????? ... that's why i said it doesnt make sense
bigger numbers always means better .... you must know that by now dont you???
so amd is reducing cost and power, and nvidias reply to that is to offer a hotter, more expensive part? :confused2
too bad, i had hoped gf110/gf102 would be more than 512sps at least...
I hope nvidia gets the 512 cuda core 485 GTX or 580 or whatever its going to be named lined up for launch soon. Been happy with my 465 but I'm hitting video ram with my modded games and SLI won't help me. Would be a nice xmas upgrade for myself.
DID I JUST HEAR FULL GF104?????
http://img683.imageshack.us/img683/6379/readyt.jpg
L]
Man it seems like new video cards every month now
nah the die size is going to be as pictured above.... larger than life...LITERALLY..
If you read the 4Gamer article, seems like GF119 is the next ION chip.
Based on the discussions here I got myself a 5970 last fryday to replace my 2 x 260 gtx. was going to go for the 480 at first but hear too much negative.
The card is fantastic. runs quite and stays cool. (moderate load)
Whats the holdup on the full GF104 GTX 475 and the dual GF104 495?
I don't get why AMD has reduced the performance with HD68xx. :shrug: They have given nVidia a real chance to take on now.
nVidia doesn't need a GTX 580 now. All they need is a full 384 SP GF104, and a double-GPU with the same chip, with a competitive price. They should also drop the price of GTX 460. Then thay can make the life realy hard for AMD.
The low yields, simple as that.
I think nvidia is still playing the "we need to maintin our marketshare" (they lost marketshare in reality)
WTF! Barts is midrange, Cayman is the product a GTX580 would have to contend with.
You are thinking that 68xx is high end part? :confused:
Barts(HD6800) is performance part that will replace 5800 series,having similar performance. Cayman is the high end part,featuring 1900+ SP (unknown exact number as of now) and this thing will destroy any GF100 based card out there,even the 512 CUDA cores parts if they ever launch. Then,there comes the dual Cayman card,the ultimate crusher,waiting to destroy any potential dual GF100 card ,if that thing ever sees a light of the day.
Oh I see, they have messed up the naming schema?. I thought they had tow single-GPU and a double GPU, the same as last rounds. But it's good to hear they are releasing a high-end card too. It's good for competition and prices.
When this Cayman is going to show up?
GTX480 is ~15% faster than 5870 today.Stream processor efficiency has been upped by quite a bit with NI generation(1120SP in BartsXT parts will almost be equal to 1600SP in Cypress-so lets say that "IPC" will go up by a decent 20%,not massive 42% as the SP count difference first shows).We will have ~1920SP with Cayman,each of which will be faster per clock than Evergreen class SP.Clock speeds are unknown for Cayman,but let's assume similar clock speed as 5870. On the conservative side,Cayman based 69xx will be ~45%(1.2 for the IPC x 1.2 for the SP count increase) faster than 5870.This alone is enough for it to crush any GF100 based single GPU card.
45% faster than a 5870 is horribly close to 5970
sounds like one beastly improvement on the same node. AMD/ATI really do know how to do magic things with their 45 and 40nm parts
Why would you have anything against such a optimism? We all should be happy if nVidia can do "something" to counter these GPUs, and push the prices down.
Why don't you like to keep the optimism for a good competition? Do you like to pay extra high prices in this round too?
I am a bit pessimistic based only on the fact that Nvidia's CEO is a BIT of a douche
lol this Monster (GTX 580) Coming in Q4 ..... this Monster use Only 250w.
I think this card GTX 580 will be Better than GTX 480 around 20-25% or more :D
this card will have 2GB/512bit/40nm/GF110
release date: Q4 2010 :up:
http://img205.imageshack.us/img205/3...vzmmi9v1ud.jpg
^lol
series is 400
model is 500
no images or references.. only trash talk
It looks like the card is real and we might get a performance preview in november.
It sounds like the gtx 580 is real, if charlie is saying it is(if the most pessimistic hating guy is saying), I think at the very least the card has to exist.
http://www.semiaccurate.com/2010/10/...ch-next-month/
The plan is simple, in November when DAAMIT launches Cayman, the top single chip Northern Islands family, Nvidia will 'launch' the GTX580 as a 'spoiler'. Tame sites will tout the wonderfullness of this paper part, and a few loyal Nvidia 'partner sites' may even get a card.
Cards will actually be available in late January.
Atleast we are getting performance early this round rather than waiting another six months for that.
How much does this go down if barts xt proves just a little bit faster than a 5850? The only thing I can see bart xt outperforming a 5870 is in directx 11. It impressive enough if they can get 5850 performance out of a die that is 33% smaller.
How many times do you need to read that 512bit + GDDR5 is the most idiotic way to make a gpu?
GDDR5 was made to save space and reduce costs. Only dual GT104 makes sense, and that's more or less what the "rumor" point, not a single-gpu(that's pure nonsense and fantasy).
They can sell two GT104 in a dual card, for a chip with 512bit they need to remake the GT104, that's worthless.
Charlie is in LOL-mode.
OOOOOOOMG you have this card HD 6970 :DQuote:
Cayman based 69xx will be ~45%(1.2 for the IPC x 1.2 for the SP count increase) faster than 5870.This alone is enough for it to crush any GF100 based single GPU card.
you say 45% ...... ok GTX 480 faster than 5870 around 26%
45%-26%=19%. that Mean Cayman XT (6970) faster than GTX 480 Only around 19%. so if the GTX 580 (look here boy) come faster than GTX 480 around 20-25% ........ that Mean GTX 580 will be faster than Cayman XT (6970) around 6% :D
Of course, all this talking expectations
I find it funny so many people are putting stock in this 580 when Nvidia hasn't even produced a fully enabled 40nm chip that's much smaller than what this is rumored to be
A dual GF104 makes a lot more sense, and its combined #'s actually mirror the rumors
Is there any info on cayman die size ? It should be big too, i wonder if GF100 size or even more.
Looks like people have a reason to start them nuclear meltdown jokes all over again.
alright then... sounds like this 580 is a real killer and i like these performance numbers youre giving, they sound much more accurate and real than anything on next months cayman. i think you should forget cayman and wait for this 580. should be out maybe before cayman even - nvidia never hypes.:rolleyes:
Maybe the new card will use roofing nails instead of wood screws. :shrug: Let's see if we get such a preview! :ROTF:
:DQuote:
alright then... sounds like this 580 is a real killer and i like these performance numbers youre giving, they sound much more accurate and real than anything on next months cayman. i think you should forget cayman and wait for this 580. should be out maybe before cayman even - nvidia never hypes.
maybe cayman XT will kick GTX 580 ass :D Who knowsQuote:
Of course, all this talking expectations
This is Nvidia talking trying not to lose more marketshare... We really have no reason to believe them until we have a working card in our hands.
nvidia never hypes?!
I'm an optimist, i really like good competition beetween IHVs, but things must be put into reality perspective. A response from nVidia is very much expected, they have good engineers there & plenty of patents+R&D money, but their decision & execution in this process noce has been lacking, so much focus was put into GPGPU & overdone the geometry performance, the cards are unbalanced & not competitive sizewise + power consumption wise.
As i predicted, early Q1 2011 at the earliest, but like nVidia's fashion, when you have no product, spin your story and prepare woodscrewed cards for the masses. This has happened in the past, so if anyone is blind to this reality, what can i say ? Even an a$$ (donkey) doesn't fall into the same pothole twice.
512 bit buswidth DOESN'T COMPUTE with 48 ROP partitions, so the specs & release date are as good as gold in the closet.
Oh, so you now LIKE the card, right ? Getting a bit optimist if nVidia is the subject of the matter, are we ? :D
Seriously, i also expect some sort of reveal or exposure of their expected response for this new gen AMD chips family, but if they (nVidia) pull another woodscrew debacle, they will simply kill any last goodwill that many reasonable enthusiasts still have for them.
BTW, i like your "optimism" regarding Bart cards performance. :clap:
RV 770 (55 nm vs RV 670) and GT200 (65 nm vs G92) say hello. Not saying it WILL happen this time around with Cayman vs Cypress, but considering AMD does have some wiggle room (new/revised mArch, diesize & power envelope budget), being optimist in this issue is certainly within reason and not unfounded/fanbois talking. :welcome:
i just love selected optimism by certain forum members, really amusing. :ROTF:
Expecting a 768 SP 512 bit single chip out of TSMC 40 nm process that is feasible to make, is simply madness or people talking out of their rear.
Rumor says less than 400 mm^2, most likely 380-390 mm^2. Considering Bart diesize (230 mm^2), we can expect double the SPs & TMUs if the ROPs stay the same number, or atleast 50% more SPs & TMUs if the ROPs & buswidth grow proportionally. Ofcourse, that's assuming the underlying mArch of BOTH chips is the same. :)
Anybody who claims GTX 480 is 26% faster than HD 5870 ON AVERAGE is certainly a person who love talking out of their rear & has certain allegiance toward "certain" IHV. :shakes:
lol .....yo boy Look here before you judge meQuote:
Anybody who claims GTX 480 is 26% faster than HD 5870 ON AVERAGE is certainly a person who love talking out of their rear & has certain allegiance toward "certain" IHV.
http://img196.imageshack.us/img196/1549/28599227.jpg
Some people do not know what they talking about
I thought Nvidia was increasing numbers on a yearly basis, June 2008 GTX 280, June 2009 supposed to be GTX 380, June 2010 GTX 480, January 2011 GTX 580?, June 2011 GTX 680?, looks like by the end of 2011 the GTX 780 will be around competing against the HD 7xxx series :ROTF: and maybe another zero or x will be included from somewhere, maybe GTX 780x :rofl:
that makes the 5870 20% slower and the 480 23.8% faster, then u dont even give the link to were u got it.
and those stats on the total depend alot on what u benced and how u did it, like witb BC2, if u take the intro (were most sites bench since u cannot demo) the 480 smokes with all the water and geometry but if u do the truck chase or a jungle run the 5870 and 480 are even if not the 5870 on top since the 480 is ROP starved.
What nvidia should do for q4 as many said is getting a dual 460 out for agressive prices which should put some competition against the caymans. especially with some added tmus and faster gddr it would really scale well as with all 460s and if priced well - 349-369 us$ it should bring nvidia some good volumes.
:ROTF:
135/109 = 1.238 aka GTX 480 is ~23.8% faster than HD 5870 OR
109/135 = 0.807 aka HD 5870 is ~20% slower than GTX 480
You're really GREAT at math, perhaps should consider miniaturizing yourself to replace the stream processors in the upcoming Northern Islands chip themselves ?? :clap:
Agreed, most quite fair sites that i've read stated that GTX 480 is around 15-20% faster than HD 5870 on AVERAGE, depending on the setting & games/scenarios selected. I really don't think Cypress based cards would still be selling so strongly with pricing structure intact even when it almost reaches the end of its lifetime, if GTX 480 is really 26% faster than it was on AVERAGE.