so nvidia was not lying! amazing!
Printable View
so nvidia was not lying! amazing!
So after reading a good dozen or so reviews I think I have a decent summary.
GTX480
+Price / Performance level very near 5870
+Price is fair in regards to price based on the available alternatives (again point 1)
+Indisputably faster than the 5870, often with noticeable better minimum framerates
+Respectable clock headroom ( although I didn't see enough info on the increases in power / noise at overclocked settings )
+Very nice SLI scaling with 2 cards for a new release
- 100watts more power draw at load over a 5870 ( aka not power efficient at all ) Do note that load power draw of *2* 5870 is often similar to the 480 according to many of these reviews*
- 25watts more idle draw ( not terrible but not impressive either by any stretch )
- Hot as the lava in metroid ( pure pwnage reference ;) )
- Loudest reference card
-Underwhelming performance at 2560x1600 ( although there are multiple mentions of this being driver related... which strikes me as odd )
So assuming they don't combust down the road ( of which Nvidia attest to not happening... ) if you have the power supply, cooling (ie adequete case) and money and don't mind the noise, the 480 is a decent card @ its intended MSRP. Now there is still the question of availability and potential price gouging.
All that said, a 512sp part on a smaller node would probaley be epic. Now wether we can expect to see this before ATI brings out a new line... doubtful. And as far as a dual gpu part, all I can say is good luck. Given Nvidia's claimed TDPs are easily reached ( quite interesting ), they'd have to pull some serious voodoo magic to manage a dual gpu card that works AND makes sense at its price and performance level ( a very difficult prospect I'd say )
Not too bad, over a 10% OC is nice on a stock cooler considering how hot the card runs. I want to see some sub zero results.
So not great but not bad. Pretty much what most thought > 5870 but < 5970. But 480 sli is easily king. Drivers would not doubt improve performance further and the oc (770 or 800 on evga) would prob bring it within 10% of 5970. Stll think a shrink would make this scream. Best thing is that this whole mess is over!
Texture fillrate is way down on a 5870... maybe that is what is gimping 2560 x 1600 performance?
GTX 480:
http://i263.photobucket.com/albums/i...mm/480gpuz.jpg
HD 5870:
http://i263.photobucket.com/albums/i...5870-gpu-z.jpg
Nowadays acoustics is a huge proposition. Using the good old 'headphones' excuse doesn't count - for one to cancel out the 50 dB fan one must turn up the sound to achieve 50 dB or higher constant noise levels, not including transients which may very well spike to the loudness of a real gunshot.
I will simply never buy it based upon the acoustics.
The hardocp videos are eye opening, the cards are fricken loud! And it was an open air test, I can only imagine how loud/hot they get inside a case, lol.
complete and total failure by nvidia in my eyes.... who the hell would buy one when they arent much better than the current line of cards,,, and i am not even talking about ATI cards...
:( Fermi is a huge disappointment. Here I was hoping NV's cards would be a bit more powerful to maybe spark some minor price drops on ATI's cards but I don't see that happening any time soon. The 470 is a complete flop. It only beats the 5850 by a few percent but will cost considerably more, generates a lot more heat, uses a lot more power, and runs a good bit louder. That's before you consider overclocking both which would give the 5850 the performance lead. NV specific extras (CUDA, PhysX, etc) aren't worth the price premium since they wont have much impact on the average gamer. The 480 is a bit better, it's performance lead over the 5870 is more apparent than the 470's lead over its competition. Still though, like the 470 it generates a lot more heat, uses a lot more power, runs a lot louder, and will cost more. Maybe NV can sort out some of these issues with their B1 rev and give us some good competition. Though by that time ATI will be close to launching their next series of cards so who knows what will happen.
GTX480 SLI sounds like a damn jet engine. :shakes:
Sadly, I fully agree. This 5870 is as loud as I'll cope with. My 4870 x2 was too loud and my short experiance with a 295 was only marginally better noise wise.
Its kind of funny actually, AMD took the rough performance of their previous flagship and offered a solution at roughly half the power consumption. Nvidia put out a card slightly better than their flagship ( more often than not it seems ) but at a similar power envelope. The only conclusion that would be fair to reach in my mind is that GF100 has great potential but it isn't an ideal solution at 40nm ( although how much performance per watt could be gained at a smaller node is beyond my mathimatical abilities )
im not sure that its indisputably faster then the 5870. it (480) looses in a few games, its not as good in vantage, and it has a lower OC potential (atleast until there is liquid and if there is voltage control it may change). most sites also used 10.2 isntead of 10.1 or 10.3 that are 5-10% faster.
its an exact repeat of the last gen just with a 6 month delay from NV, NV is hotter larger and louder, ati is smaller, less wattage and clocks better.
they dont need to go lower than 40nm they need a fab that can use single voltage trenches, ati had the same problem the 5870 (and the other 40nm and the yields with the 4770) was supposed to be 130-150W and clocked higher, but when u have a larger die it has a much large effect to change the power delivery to make it have some redundancy.
but the 5850 clocks way higher than the 470 will, the 5850 will clock even higher than the 5870 so its within a couple percent points, if u have voltage control. and then there is the power draw. without ocing its in a perfect spot though and should for the 5850 to $250-275 again
Bit-tech
http://www.bit-tech.net/hardware/201...1-5gb-review/1
TechReport and Xbitlabs is still MIA :D
I believe the saying goes... don't let the facts get in the way of the truth!
Most reviews have the 5850 and 470 trading blows, with the 470 occasionally creeping towards the 5870 range, but usually on average just above the 5850. However, the 5850 overtakes it at higher resolutions.
Oh and it also uses 80-100W more than the 5850 (while being louder!)
I'd say the 5850 will be quite relevant, and if it gets back to MSRP, will really hurt Nvidia there
:rofl::rofl:
You have to admit though it does make a nice grill addition to any case! Always wanted to make pizza at a LAN party and game at the same time. Those stands charge too much!
Seriously though.....The heat is a little much! That to me is a real deal breaker. Even if I put it on it's own loop, I mean if the pump stops something is going to catch fire or something. Fermi is the perfect name for this card. Let's hope it does not adopt the name Chernobyl.
So this summarize fermi heat efficiency.
http://img.photobucket.com/albums/v2...en/ardilla.gif
i might be bias to both the 5870 and gtx 470 since i do tend to spend $400ish for a card, and coincedentally, i feel those are the 2 best cards atm with all things considered.
One fact is the wide range of models on the Cypress family. For example the whisper-silent PowerColor HD5870 PCS+ both in idle and load which is impressive.
its funny how nearly all benchmarks in different websites show differing results..........
especially in BC 2 and Dirt2, MOST especially in BC2... at one site 480 is significantly faster, at others they are equal, at others 5870 leads by a good margin. just look at all the reviews and tell me that they add up.
Wow the damn card is LOUD, anyone without 3rd party cooling can throw their speaker sound system out the window. :down:
I like that the 480 gtx is faster but sound + 100W over the 5870 is a definite no buy. :shrug:
Any reviewers taking apart the card to see how big the coolers are and compare the two?
Perhaps to say it is indisputably faster is a bit of a stretch but it is faster more often than not none the less, which for most would be enough if they can look past the obvious cons.
Differences in system config ( cpu, motherboard, ram, hard drive ), software ( drivers, windows used / optimization level ) make it very easy to show these differences. Since BC2 has no benchmark utility, its understandable that results will vary ( ie did they use a realistic server load, did they use a single player map, did they use an empty multipayer game ect ) Most sites don't specify how and where they benched a game. Obviously with ingame benchs, youd expect results to be somewhat consistent but the afforementioned varriables can still sway the balance none the less. Thats why its best to consult as many reviews as possible to get a balanced opinion overall.
http://www.guru3d.com/article/geforc...-480-review/21
Quote:
Now, for Dirt 2 we unfortunately had to ditch all results we had made in the past and I thing there will be a number of websites out there that made this istake. The previous time demo test was based on the DX11 demo of the game. The demo code however will only work at DX9 for the GeForce GTX 400.
yeah, that with Dirt 2, but what about BC2 ?
Wait, so Charlie was right about the DiRT2 demo? /shocked
Like I said, I don't think I've ever quite seen a card release that had such ridiculous performance deltas between different games, but within games themselves!
I knew the card was contentious/controversial, but it looks like all the blood is being spilled now
BC2 has the same thing u have to force it, harware canuks would be the most reliable IMO since they go through all of the configs and set them up properly even if they dont ship that way (like ut3 games) but they did not test dirt2 right as the 5870 should edge it out.
750/1500/3800 on the Hydro FTW. With that in mind I wouldn't expect any OC models to come with insane clocks, as the past would dictate that the Hydro models always had the highest clock in their line. I have a feeling vendors will be conservative to keep potential RMAs down, at least until a non reference card / cooling solution is possible ( which might not be any time soon )
I want a GTX480 for Folding@Home like NAOW! Unfortunately my room will get too damn hot and my power bill will shoot up way too high. I think it is prudent to wait this out a little to see a refresh from Nvidia, AMD's next gen GPU due out, and Stanford's OpenCL F@H client.
We also better see a mother :banana::banana::banana::banana:ing price drop on 5870s RIGHT NOW!
Yep. They would be doing more nicely with a price drop, though :D
Damn right!
Also I took another look at GTX480's price. I don't think I'll have one of those in my system for a while. This feels like when G80 came out it was fast as hell but way too power hungry and priced into the stratosphere. The refresh was just what the doctor ordered though.
EDIT: The 5970 coming down to $599 would be very nice :).
the way it looks from the tpu review we have a new x1 core contender for 3d03 :D
Minus being awesome :rofl: Can't neglect the fact that the G80 came out with no competition and absolutely owned the previous generations.
I still fondly remember my 8800GTX. It was the first Nvidia card I bought right at launch and it impressed me then and still would now all considered.
Guru3D says the card is 43dBa full load... they can't hear it over the rest of their PC. Not sure what Kyle is doing different.
And the GTX470 is 20% better than a 5850 at 16x10 and 19x10 with 4xAA (as per the HWC review, and they tested with 10.3). It costs more ($50) but you get better FPS/$ but just worse FPS/watt.
So it is not a true "green" card :P
The power consumption really is HOLY :banana::banana::banana::banana:. Honestly, im impressed. Furthermore, ATI cards seem to have lower in game power consumption, and much greater in Furmark. Fermi's power consumption doesnt differ much in game and in furmark though. And really, a single GPU card to consume more power than a dual GPU card is astounding. For less performance
Wow
Our small contribution with review of gtx480
http://www.hwgurus.com/forum/showthr...id=3#post14555
Sound measurement is as much an art as it is a science, so much depends on how you take the measurement, what distance, and in what external/internal environment.
As such, it is probably a good idea to look up a Guru3D review of a card you have and use that to level yourself against their scale/technique.
For example, I have a 295 GTX, and a crude sound meter (it's in one of my DVMs), it peaks at 68 dB at 12 inches from the back of my case (a CM ATSC 840), with a baseline background of about 45 dB.
http://www.guru3d.com/article/asus-g...iew-engtx295/5
This does not make him wrong, it just means he measures differently or is in a different environment.
I also have a 4870 X2, in a TT Armor, at load it measures about 58 or so dB (it is much quieter than my 295 but still noticable when loaded)...
http://www.guru3d.com/article/his-ra...048mb-review/5
He again gets 43 dB...
At 250W TDP, GPU temps of 95 C ... it is unreasonable to think this card is not going to be loud as it will definitely need to push a large volume of air.
The TomsHardware article also quotes it as being very loud
I think Guru3D's is the anomaly on the noise part
I've been reading for 5 hours non-stop & I need a breather to let everything sink in... but how is it that we can have 20 reviews of Bad Company 2 and have such varied results?
I game @ 1920 x 1200.. I had thought Fermi would tear Battlefield up, it doesn't... now I am pissed for waiting 6 months!
Summmmeeeeer
http://www.legitreviews.com/article/1258/16/
Only a true nvidia fanboy would brave his sanity for this. I sure as hell won't. Part of the reason why I considered the GTX260 was for the superior thermals, given equal performance to the 4870.Quote:
In my personal system (Corsair 800D Chassis) with two monitors the GeForce GTX 480 graphics card would idle at 90C and if it was a sunny day and my office was warmer it would idle at 92. I fired up the new DX11 game title Aliens Versus Predator and with GPU-Z in the background I saw the temperature reach 99C while gaming for around 30 minutes. At this temperature the fan is spinning at 70dB and it honestly was not an enjoyable gaming experience. I asked NVIDIA if the card was built to run at temperatures this high and they claim that the GeForce GTX 400 series was built to operate at high temperatures and reminded me that 105C was the peak temperature for the GeForce GTX 480 video card. While benchmarking the GeForce GTX 480 graphics card on the open test bench I found the outside of the heatsink to reach 50C on the fan side and 59C on the exhaust side, so this card without a doubt will put out some heat.
I think most discrepancies in the reviews are due to older drivers.
Isn't the BC2 multiplayer engine different from the single player engine? That might be the cause of the discrepancies also
Wow this is getting a lot of attention. Hundreds of posts on XS, the review sites are still having problems. I'm having trouble reading the Anandtech review atm. Didn't know there were so many computer geeks in the world lol. :D
So... besides the fact that is late, extremely power hungry and hot for a single gpu card, having a small edge over 5870, and the possibility that these card are going to be more rare than a yeti,
The biggest concern nvidia must be having is how on earth are they going to be selling these as HPC cards, when they are clearly power hogs, and potentially too hot for a rack?
Hearing those HardOCP videos seals the deal for me, there is no way in hell I'm buying these cards, earlier indications from reviews suggested the fan profile wasn't aggressive enough topping out at 64%. Based on what I saw in those videos there is no way that can be 64% of max RPM and if that really is HOLY :banana::banana::banana::banana:.
Knowing how hot it gets down here in FL I just cannot see me being able to use these in SLI without my room turning into a sauna.
Northern Islands it is.....:shakes:
The irony is that nVidia focused a lot of attention to make it more attractive to this market, and I am with you ... at least their top end card anyway, i don't see many in the HPC circles jumping up and snagging these, perhaps a clocked down version that runs much cooler.
What the hell? All the reviews say ~50C idle (most say sub-50 but I'll give you the benefit of the doubt). What the.... what... wha.....Quote:
GTX 480 graphics card would idle at 90C
6 months for this ? Nvidia epic fail :(
It's too bad about the thermals and power usage. Would make a damn killer F@H card. Just for comparison, I light my whole upstairs of my house on 42 watts (14 GU10 LEDs 3 watts each). This card would probably cost me more to run 24/7 in 2 months then what it would take to light my one house floor for a year! :eek:
The card isn't loud. Neither one is actually. You can't equate noise from a crappy mic on a camera to what a card would sound like in your case with the side panel on. The videos seriously mislead people regarding the noise output.
The problem is most reviewers bench on open systems which is why the card needs to be installed into a case before SUBJECTIVE conclusions regarding noise can be made.
Want loud? Install a HD 4870 X2, GTX 295 or a HD 4890 and you'll get loud.
It does run ULTRA hot though.
if you ran your lights for 6 hours per day, 7 days a week (so 42 hours a week) and you wanted to compare 24/7 SLI 480gtx (using anand's number of ~850watts for the total system at peak), you'd basically spend 8kwh on your lights versus 632kwh on your computer. In terms of money, that'd be 1$ vs 95$ (assuming about 14.5cents per kwh). If you are in Europe, you can double those numbers to accommodate for the ridiculous electricity prices. Using your initial example, that would mean you could run your lights for something like 8 years before you spend the money you would've spent on crunching.
I sure hope HPC people are looking at this and wondering if there's a way they could potentially get fission reactors ready to power a cluster of these...
So anyway, this card is so hot it'll take it's own dedicated water loop, it's so bad in price/perf that it won't be attractive to anyone making less than 100000$/year and it'll get re-released in 6 months with all the bits unlocked when they finally figure out how to make it work properly.
I'm going to pass on that crapola.
I think Fermi has taken the title of "space heater" from the old Pentium D's.
I want to see what availability looks like, but honestly its fairly well priced considering the recent past of the company that has launched it.
I don't want to make the unneeded PR for nvidia, because I think it doesn't desrve it, but I can say that I will be getting a couple of 480s for sli purposes...)) But I have my own argumentation for it - I just got used to the nv image quality too much. It's not that ati's iq is inferior, they are just different... So even if it were slightly faster than285 but with dx11 support I would've got it anyway. All in all though fermi looks like fail in general, but still, it seems to have a strong potential, in its future revisions probably . Kind of. heh
The you and I must have different versions or something. Mine is ULTRA loud...as is one of my HD 4890 cards. My GTX 295 does beat them all out though.... ;)
If the mic isn't set properly it will take directional sound and not ambient. My Canon S2IS has the same problem and it picks up EVERY sound from a nail scratch to a pen writing on paper if the wrong setting is selected.
The GTX 470 isn't quiet by any stretch but it doesn't make as much racket as that video would lead you to believe. I'm just stating this from my personal experience with the card.
The HD 5870 is VERY quiet though. There is no doubt about that.
Agreed... it isn't loud like the HARDOCP videos make it seem in my experience. I was trying to compare the video volumes and I think the microphone's frequency response might have a lot to do with the perceived volume.. the mic might respond more to the pitch of one card than another? :shrug:
If I can equate noise levels with some of the cards I have gone through in recent months, I'd put noise output at something like this from loudest to quietest.
Palit 9600 GSO (my mistake for ever putting it in my system)
GTX 295
HD 4870 (with the revised fan speed profile)
HD 4890
HD 4870 X2
GTX 480
GTX 280
GTX 470
GTX 285 / GTX 275
HD 5970
HD 5870
HD 5850
Then again, we all have to remember that certain noises bother some people more than others.
same model cards can often be louder than another due to a bad mount\thermal paste application which is unfortunately fairly common
gpu stays the same temp but the fan revs harder to keep it there
They performed as expected, but I was hoping for a better pricing just to push ATI's cards back down to at least MSRP. The one thing that I think Nvidia did right compared to ATI was backporting the Nvidia surround vision (or whatever it is called) to the previous generation, meaning that if you have current GT200 cards you can get a similar experience to eyefinity without buying new cards. Granted, it isn't for everyone, but it makes the current generation last that much longer. The heat output on the cards doesn't bother me or the noise since I would water cool them anyway, but the price and availability is the big issue here. We will probably not see a good supply for several months. At least this should push the prices of the current cards down on Nvidia's side. Why buy a GTX295 or a GTX285 when this is "available".
If you're 'used to' the nvidia image quality, it means the very best nvidia IQ you've seen would be the 2 series.
In which case, the 5870 has BETTER IQ than what you're used to.
Fail justification is fail.
Someone with a fermi post up the AF check tube benchmark now.
Well, I read the Guru3D report and it doesn't seem that bad, They got a good boost in performance at higher resolutions when overclocking (Looked bottlenecked at lower resolutions). In some cases the GTX 480 was beating the 5970 if people read it. But it was mostly beating the 5870.
I think I will wait till GTX 485, where they unlock that final SM. Plus EVGA doesn't have a SSC or FTW Air on the website yet.
Looks like a good card once all the hiccups are ironed out. But technically they are right. The fastest single GPU card. But it will give you radiation poison :p:.
Yup.
http://www.youtube.com/watch?v=UrWrnFTO5Xg
See how much trouble I caused w/ this video of a 4890? :p:
Well, yeah the supersampling of the image displayd and true multisampling fsaa really make the image look crystal clear, but um... you know nvidia's anisotropy filtering is better (and I mean of higher quality here), and the mix of csaa/msaa + supersampling transparent aa add that nvidia-specific image look. The image is blurrier on nvidia cards than on ati's. I got used to that blurriness, the ati iq is just too clear for my taste. It probably may be considrered of higher quality, though.
From what I see from these reviews is confusing. Some reviews love it some are not too excited. But I have to say with spending this much time on these cards I would have thought power would be lower... I'm not surprised it runs hot but still....
It's late I'm tired I'll read the rest tomorrow but right now the 5850 is lookin pretty good. Not like I can afford 2x 480 lol I wouldn't even want to pay for the power bill for them.
No, it's not.
Actually run the AF tester from the 5870 and 285. Look at the results.
http://img21.imageshack.us/img21/4306/afanim.gif
thats one of the things that i hate, its like the consoles u cant tell that are lacking AA most of the time since the edges are blurred a little. with the leap ati made by changing the AF method i would hope that NV would then take initiative to match atis AA methods now that ati matched the AF of NV.
It is the af tester doesn't do justice. There are some reviews on the net that proove that. I also beleive my eyes from the games experience. fsaa is better on ati cards and the af is better on nv cards... this is how it works these days.
here is some read for ya:
http://alienbabeltech.com/main/?p=12648&page=3
Prove it.
I just proved you wrong. Counter argue. (use the balance of arguments and meet at an equal level or lose)
http://img339.imageshack.us/img339/6...efae7b7c55.jpg
Counter argue or above. If not = wrong.
Quote:
Originally Posted by Kyle
http://hardforum.com/showpost.php?p=...&postcount=238
The AF tester shows exactly how the 5xxx series implementation is theoretically superior. You either refute (theoretically superior) or not. Subjective experience doesn't factor into this. Actually, the 285 on the text makes the M's look very jagged, so now we enter in the foggy territory of personal opinion.
Now the codename Fermi makes sense to me: Fermi = NuclearFail, the only thing impressive is the SLI scaling everything else is sad for being 6 months late, I hope nvidia finally learns from this and drops the "Big chip is the way" bull:banana::banana::banana::banana: theory and their ego. By the way some are claiming "wait on drivers" I remember hearing this before (R600), don't expect miracles either, nvidias driver team is far superior than ATI's team but they had 6 months+ and they miserably failed to impress, they did it with SLI only I mention. Fermi is a fail period, been a nvidia fan myself im truly dissapointed and I must admit ATI did it right this time. In my conclusion the only reason to buy Fermi is if you're planning to build a new system with multi-gpu setup and you do not care about consumption, thermals and price.
^ I think your sig says it all.
I'm getting a 5870.
http://upload.wikimedia.org/wikipedi...47/Lindos1.svg
well, humans tend to even perceive different frequencies at different loudness even if they have the same sound pressure, as shown in the graph
an 75dB 20hz sound is likely to be percieved as loud as -5dB 3000hz sound in the most extreme case
"It's okay, CUDA will fix everything"
"Fermi is the second coming" point of view: Epic Fail.
"GT200 sucessor / RV870 competitor after 6 months" point of view: it's OK. No more, no less than realistically expected.
Power consuption, heat and noise are ridiculous, though. Apart from those going SLI, running outrageous res. and/or multi-monitor I fail to see any good reason to grab a gtx480, imho. From the reviews I read (anand, HWC, Elite Bastards, ixbt and graphs from another sites) the gtx470, if it's really priced at $450, Nvidia should bury them and pretend it was a April's fool joke.
Given all the questions about noise levels, and the uncertainty of [H] videos, it would be interesting if you would do a test comparing all the GPUs you mentioned, and actually test actual Decibels from, say 2 locations, across a couple games. Or maybe just do it for a few of the cards (the 4xx and the 5xxx)?
Would you be willing to do that? I'm sure people would like to see the data for actual games, not a synthetic test like Furmark (especially if AMD throttles this application, which alters results).