brain_stew
Member
DennisK4 said:
No chance in hell. ATI already had to sabotage the 5890's clocks to get it under the 300w PCIe limit, and no one is seriously expecting Fermi to consume less power than a 5870.
-
Hey Guest. Check out your NeoGAF Wrapped 2025 results here!
Fermi (Nvidia Next Gen) GCPU Architecture: Thread of promises, waiting and 2010
- Thread starter artist
- Start date
No chance in hell. ATI already had to sabotage the 5890's clocks to get it under the 300w PCIe limit, and no one is seriously expecting Fermi to consume less power than a 5870.
godhandiscen
There are millions of whiny 5-year olds on Earth, and I AM THEIR KING.
I love the fact they use my motherboard (ASUS Rampage Extreme II) along with the Corsair cooling system I have yet to install.bee said:
Now, I am thinking about skiping this gen altogether, or wait for the refreshes. Unless I can't play Bad Company 2 at 60fps, I won't think I will purchase a new card this year.
Nvidia might also reduce the clocks and promote it as an ultra overclockable part like the PR spin ATI gave to the 5890. However, I doubt Nvidia would want to release the 395 so soon, unless they need it badly because they are getting plumeted in the benchmarks against the 5xxx line, and they need to at least to have the top dog to make them look good.brain_stew said:
bee
Member
fermi obviously isn't gonna be cheap, nvidia cards never have been since the TNT days and that's when they didn't have yield and heat problems. 5870's currently cost around £340 in the UK, day one with extremely limited availability and price gouging i can't see how they could possibly be below £450, both 8800gtx and gtx280 were above that. however in the following weeks both those could be had for around the £370 mark, which is where i'll bite at and i should get £80-100 for my gtx 260
Moar Dual-Fermi rumor: it's gonna be hot!
http://www.fudzilla.com/content/view/17198/1/
If the rumor that Fermi is going to be only 20% faster than a 5870 and clearly slower than the dual GPU 5970 is true, nVidia may feel pressured to make a (very hot) dual GPU card to win back the performance crown.
http://www.fudzilla.com/content/view/17198/1/
If the rumor that Fermi is going to be only 20% faster than a 5870 and clearly slower than the dual GPU 5970 is true, nVidia may feel pressured to make a (very hot) dual GPU card to win back the performance crown.
Rumor: Nvidia Fermi production in 3rd week February
http://vr-zone.com/articles/-rumour-nvidia-fermi-production-in-3rd-week-february/8273.html
http://vr-zone.com/articles/-rumour-nvidia-fermi-production-in-3rd-week-february/8273.html
They need the 395 because the 5970 on pure transistor count is going to always beat Fermi single chip. You can't magic away that much of a physical hardware advantage by trying to be clever. 5970 = 4.3bn transistors vs. Fermi Single = 3.0-3.2bn. Come on.
395 is coming for sure. How they get it below 300w, I would love to see that.
395 is coming for sure. How they get it below 300w, I would love to see that.
I love my 295 and the more powah the better....
BUT
I would prefer the GPU twins enforce a tick-tock cycle so that on every tick they make some goddamn efficiency improvements! The current rate of progression although great is absolutely dangerous from the perspective that it is not sustainable from NV.
Can not wait till Intel get their shit in gear and enter. Very happy to see AMD use the half-nodes. Nvidia need a fucking slap for their neolithic approach. To just thow even more trannies into the mix is not really being clever. With Chipzilla coming eventually, or by buying NV, they'll finally start getting a grip on performance per watt. If you max that, the high end will be the absolute shit.
BUT
I would prefer the GPU twins enforce a tick-tock cycle so that on every tick they make some goddamn efficiency improvements! The current rate of progression although great is absolutely dangerous from the perspective that it is not sustainable from NV.
Can not wait till Intel get their shit in gear and enter. Very happy to see AMD use the half-nodes. Nvidia need a fucking slap for their neolithic approach. To just thow even more trannies into the mix is not really being clever. With Chipzilla coming eventually, or by buying NV, they'll finally start getting a grip on performance per watt. If you max that, the high end will be the absolute shit.
Dr. Zoidberg
Member
avaya said:
My friend learned this the hard way...
Anyway, I want to see something from Nvidia this quarter that can beat the 5850 AND come in for less than <$270. Otherwise I'll be disappointed in them.
avaya said:
How is Intel gonna help the problem? Larrabee was rumored to be around 300 watts and the performance sucked. That's why it got canned for 2010.
As for the future, I'm not sure the desktop dedicated GPU has much of a future left anyways. Pretty sure that's why Nvidia went with the radical approach they did with fermi (Why spend billions in R&D on something with limited future. Hence, the dual purpose of fermi.)
If you want to look at the real future of Nvidia, I suggest you pay attention to what they're doing with Tegra. They're radicalizing what's expected in the performance/per watt area.
1-D_FTW said:
Intel have been very aggressive at coming down the nodes and maximising performance. Nvidia haven't. Intel will start offering the whole package (a competitive version of it!), the OEMs will bite. They can only offer the whole package if they maximise performance per watt since it will inevitably lead to cheaper chips. The others will be forced down the same route.
Desktop GPU's are still NV's bread and butter. People predicting their death are way short of the mark.
Not surprised by Tegra when you compare it to IMG's products. That market has been stagnant for so long.
Projectjustice
Banned
Can someone break down the overall outlook of Nvidias new card is going to be? Dont feel like going back and re-reading everything. Ive been waiting for Nvidia's answer to ATi's DX11 cards. Im using a 9800 GTX and its time to upgrade.
http://www.rage3d.com/board/showpost.php?p=1336134574&postcount=1573
Finally some good news, on tessellation atleast.
http://www.xtremesystems.org/forums/showpost.php?p=4193724&postcount=734
Beats the 5870 hands down in many games.
Beats the 5870 hands down in many games.
xemumanic
Member
irfan said:
So finally, its known to be HW tessellation.
Speaking of which, if ATI hardware has had a tessellator function for some time, including the 360 GPU, is there any hope of at very least the 360 using it? Or does putting it to use kill performance?
This is one area I know next to nothing about.
Minsc
Gold Member
xemumanic said:
Yea, I just wanted to grab the quote about Fermi having hardware tessellation, and realized grabbing it from you might have cause confusion, sorry.
There's been a little back and forth discussion about how nVidia's tessellation which was said to be done via software could be better/faster than ATI's hardware tessellation.
Of course, I could have gotten entirely confused again, but now it looks like any earlier judgments about hardware tessellation being a waste are somewhat void.
xemumanic
Member
Minsc said:
Yeah, already forgotten.
But a 'software' solution, given the basis of the rumor, (that it was done via the CUDA cores) may or may not have been better. We'll never know, because that doesn't seem like it was ever the plan; but I think the claims of it possibly being faster stemmed for a backslash to the initial response to the rumor, where ATI fans were already pointing and laughing. And who's to say it would've been better or a total fucktard way to accomplish it?
But I guess in hindsight, I guess we should have known it'd be hardware-based, isn't that a part of the DX11 requirements?
Take the following statements witha grain of salt, I still dunno wtf tessellation IS. :lol
Speaking of which, in plain English, wtf is tessellation?
Minsc
Gold Member
xemumanic said:Yeah, already forgotten.
But a 'software' solution, given the basis of the rumor, (that it was done via the CUDA cores) may or may not have been better. We'll never know, because that doesn't seem like it was ever the plan; but I think the claims of it possibly being faster stemmed for a backslash to the initial response to the rumor, where ATI fans were already pointing and laughing. And who's to say it would've been better or a total fucktard way to accomplish it?
But I guess in hindsight, I guess we should have known it'd be hardware-based, isn't that a part of the DX11 requirements?
Take the following statements witha grain of salt, I still dunno wtf tessellation IS. :lol
Speaking of which, in plain English, wtf is tessellation?
Put simply, it divides up the verticies of polygons in to more polygons, and you can use the additional detail in conjunction with displacement to create neat effects like cloth blowing in the wind, and water ripplings when a car drives over a puddle.
It's a bit of a performance hit, which was unexpected since (to me anyway) it originally was made to sound like it was minimal, and the tessellation from the Unigine Heaven benchmark demo brought framerates from ATI's cards down a ton.
In the image above, the right half is the mesh / wireframe with tessellation, and it's pretty apparent how much extra detail you can create to work with, and extrude the polygons to create more spikes on the dragon.
Why you wouldn't just create the spikes to begin with, I don't understand, it seems a bit overkill to bump the polygon count up that much when a simpler method would suffice. But the tessellation effects used in DiRT 2 are more simplistic and effective.
bee said:
Yes that's when the NDA lifts. This article says information and results will come in on january 17th
http://www.sweclockers.com/nyhet/10370-nvidia_visar_geforce_gf100_bakom_stangda_dorrar/
I am definitely getting a Fermi when it comes out
Minsc
Gold Member
darkwings said:Yes that's when the NDA lifts. This article says information and results will come in on january 17th
http://www.sweclockers.com/nyhet/10370-nvidia_visar_geforce_gf100_bakom_stangda_dorrar/
I am definitely getting a Fermi when it comes out
Even if they're over $500? I hope they come out for less, and force ATI's prices down and do well, but I'm curious to what people would pay for Fermi, are people willing to pay $400, $500, $600, or even more (perhaps due to limited supply and excess demand driving prices higher)?
Stallion Free
Cock Encumbered
I am so goddamn hyped for Fermi since they are now wiping away all my fears. Ideally the reviews will be out, the cards will be widely available and the prices will be lowering some by the summer time when I will be ready to upgrade.
TouchMyBox
Member
Smash88 said:
I must be blind because I fail to see where he says how much better fermi is. Everybody is expecting fermi to be at least 10% faster, hearing that it is faster than the 5870 is not surprising.
Edit: I just missed out on a joke, didn't I? :lol
Stallion Free
Cock Encumbered
dionysus said:
I never said it would be good, I said I am hyped... because I am hoping it's good. Apparently there is going to be hardware tesselation so I won't be losing anything by sticking with Nvidia and I stick with Nvidia because Nhancer is critical to my PC gaming experience.
Schrade
Member
nVIDIA drivers are usually much better in general than ATI both installation and its control panel.darkwings said:
ATI's drivers have always been shoddy and when they went to using a bloated piece of shit .NET control panel with the Catalyst Control Center they just fell apart. I hate dealing with ATI cards because of that. A shame that the hardware is so good and the software is so bad.
secretanchitman
Member
Schrade said:
effing exactly. dont get me wrong, i like ati's cards, they are great. its just that the drivers are completely terrible.
nextgeneration
Member
Stallion Free said:
Ditto. I really, really want a new pc this year and the prices on the high end radeon cards are way too high. More competition is good and hopefully, prices fall across the board once Nvidia announces Fermi.
fizzelopeguss
Banned
I'd rather stick with nvidia, nhancer is a killer app.
Schrade said:
This.
Back in the day... I used to think that if ATI had better drivers... Their cards would woop Nvidia's up and down, because on paper... They seemed to be beasts.
That and their usability is pretty crap. Don't even get me started with ATI with their mobile drivers. With the latest drivers installed on my laptop (modded with mobility modder), it essentially won't scale lower resolutions up to fit the laptop scree. It forces 1:1 mapping. The only way to enable scaling is to switch the CCC to Basic and go through a little wizard... And that only works when you're NOT at your native resolution. Returning to your native resolution turns it off *again*. It's annoying that I have to switch to a lower resolution, turn on scaling, then play whatever game. Once I'm done I return to native so I have to repeat the former process every time I want to play say.. MW2.
Another peeve is that I can't set the 3D settings on a game-by-game basis. I heard somewhere that "profiles" are the solution, but I think it's quite a pain in the ass to have to go through that. In my desktop, I went from dual 8800 GTs in SLI to a 4890 (I couldn't stand the microstutter with the 8800 GTs) and that's pretty much the biggest Nvidia feature I miss. Among other little things here and there.
On the flipside, one thing that I want Nvidia to borrow from ATI is their "Overdrive" tab... Basically allow me to change my card's clocks without having to download another 100+MBs for extra software. And have a simpler way to check your temps without having to load that obnoxious and sometimes glitchy 3D "system monitor". Basically copy+paste that tab somewhere in your drivers.
Other than that, Nvidia's drivers are far superior to ATI's.
Minsc
Gold Member
secretanchitman said:
See, this just sounds to me like blind hate. If the drivers were "completely terrible" how is it out of the 250+ games I have installed right now, about all but two of them run with all settings on max and with no problems?
Sure, there's examples where ATI's had problems (and that's not necessarily ATI's fault all the time, sometimes it's just lack of testing entirely on the developers side), but I do literally have between 200-300 games installed, and 99.5% all run perfectly with every option maxed out. I don't know what more I could have hoped for! To call that completely terrible sounds very misleading.
The drivers work well, and I don't know what part people are calling terrible, the CCC loads up much much faster than it used to, remembers the last tab so you can always just leave it on the overview, and 99.9% of the time, I don't have to go in to it for anything at all anyway, so it's like it's not there, except for one or two rare occasions when I want to manually adjust something, which takes all of a few seconds.