NVIDIA GeForce GTX 480M is out

Price is ridiculous.

Like I said in the other thread. It doesn't cost that much to downclock a $279 GTX 465. It's not like they underwent any heavy R&D costs to slap 480M over the 465 badge.

Of course, I'm simplifying things to make a point.

Terrible business decision, -1 customer, and so on..

 

Let it come, i dont get all the QQing, no one is forcing you to buy one.

Overall i welcome back nvidia for trying out, even if its a monster in heat and power needs and expensive as hell, still think this will push ATI to release a new card with their 6000 series. Overall if one companies completely dominates, their research and innovation simply just settles down to where we as consumers lose.

Im very interested into seeing the real first GTX480m on users, and see how they really are.

 

At this point they might as well replace the battery bays with fans/heatsinks and be done with it - current clevos with gt260/280's are already only managing ~50 mins of battery time. It's a shame from a consumer perspective too, since ATI now has an unofficial monopoly over mid/high end mobile gpu's for a while, and I really like nvidia's driver support over ATI.

And yeah, cost ratios are another thing they seem to have thrown out the window with efficiency and cooling... I doubt we'll see a fermi on anything under $2k for quite a while, let alone the "flagship" 480.

Uh? It's one thing to see how far you can push the limits when you are overclocking or modding, but it's another story when the damn thing is IDLE at higher temps than a 5870's load. I don't know about you but, I like my equipment cool and do not enjoy burnt palmrests/laps/whatever else. "You don't want a racecar in the redlines..."

 

We haven't seen what the notebook parts can do yet, so I wouldn't be too quick to judge power consumption based on TDP. At least from the desktop version point of view, Fermi is just as efficient as ATI's offering on idle power, which is the most important thing for battery life. While on load, users will be plugged into the A/C, so here it depends more on the cooling capability of the notebook (though of course Fermi will require more).

 

If it has 5850 performance, I think it's worth it. And if they can keep the idle power consumption down, that will be key. Regardless it will still retain portability. The only thing that changed in the w880cu is it gained like 0.3" in thickness or something on the bottom over the w870cu. Hardly anything to whine about I think if you do get 5850 performance in a laptop. But I doubt it will get 5850/gtx275 desktop performance like some claim.

 

That's 100% Clevo's fault though, with the decision to use such small batteries.

 

As long as the laptop manufacturers don't have to come up with a ridiculously thick "laptop" to get this GPU in there, I'm OK with this.

In fact, this should consume a similar amount of power to the D900F if you replace the desktop CPU (130W) with a mobile Core i7 (55W) and replace the mobile GPU (70W) with this desktop GPU in disguise (100W).

Also, the clock speeds are MUCH lower than the rumoured desktop GTX 465, just like the GT335M clocks in the Alienware M11x are low, even though it has 72 shaders.

However, if NVIDIA claims that this is the "fastest" mobile GPU in the world, the claim is a little hollow if it uses 40W more than the next fastest.

 

They are smart enough not not hint anything about heat and power on their website.

Also its most-likely that the clevo battery is short because they don't want to make the laptop fatter. Why else other manufacturer like msi or alienware having problems fitting the HD5870 in (also heat manage management)

 

Yeah...

Anyone else find this odd?

HD5870 has 800 Stream Processors which can be overclocked to 800 mhz easily with very low temps and at 50 watts.
- The HD5870M at 700 mhz, power is 1.12 Teraflops
- The HD5850M at 625 mhz is around 900 gigaflops (AMD website range from .8-1.0 teraflops)
- The HD5870 @ 700 mhz texture fill rate 28 gigatexels/sec
- The HD5850 @ 625 mhz about 23 gigatexels/sec
- The HD5870 as said can easily be overclocked to 800. I imagine that means about 1.25-3 Teraflops
- The HD5870 @ 800 mhz guess 30 gigatexels/sec

The GTX 480M is at 897 Gigaflops.
The GTX 480M listed texture fill rate 18.7 billion/sec

Looks to me the 480M is going to follow my prediction, while tessellation may be good, but it will be hampered on Direct Compute and other shader duties. I'd like to see how well this 480M performs in game in reality not just some tessellation benchmark.

Nvidia thinks DX11 is about Tessellation, I disagree, I think Direct Compute is what excites game developers more. As others have said, I don't see tessellation impacting games for another few years yet. BC2 is a DX11 game because of it's use of Direct Compute.

Looks to me the HD5870M could still win on texture fill rate and computational power or pure brute power. Like said, this will be interesting to see how these actually go head to head in games not just tessellation benchmark.

 

While they're harping on tesselation, apparently AMD has been looking into driver updates that will use unused Stream processors to do tesselation, so tesselation will theoretically get a pretty big bump soon here. Those are of course rumors.

 

I have no real preference between ATi and Nvidia, except the prices. I'm happy that they finally released something, since at least ATi will have to compete again.

As for all the comments complaining about the possible size increases of laptops; I see this as a non issue. It's the biggest laptops getting marginally bigger, and this hardly matters because any laptop where size mattered before won't be getting a GTX 480M.

 

Say what? Where are you getting this from? The temperatures are not quite as bad when idle, but the Fermi cards are still hotter and more power hungry than the Evergreens, even when idle.

 

Yeah and i read up somewhere that the first public "Fermi" offerings don't even have the full features of the Fermi architecture implemented yet.

 

The specs suggest that this is a downclocked desktop GTX 465, which would make sense as it's their best bet considering what nVidia has now. Funny enough that this will be the first consumer available notebook card with GDDR5 and a 256 bit memory bus. However, nVidia's use of higher bandwidth is inneficient in this new fermi architecture, which is evident in desktop comparisions against the ATi cards.

Also, if you need a reference point for how this card will perform, Xbitlabs ran some tests with the desktop GTX 465 (beta drivers), which can be seen in the following:

Nvidia GeForce GTX 465 Preview: Worthy Successor or a Weak Link? - X-bit labs

So we should expect 75-80% of GTX 275 performance (I'd say GTX 280 performance because of drivers).

 

The ultimate goal of lowering power consumption and heat isn't to fit the top of the line card into the smallest notebook....that would be a waste of the cooling potential larger notebooks offer.

If one day a GPU that offers Mob. HD5870 performance can be designed to run cool enough to fit in a 10.1" netbook then a much more powerful card better be designed to take advantage of the cooling potential a larger 17.3" chassis offers.

If a notebook can be designed to cool a GPU with a 100w TDP then a 100w TDP GPU should be designed to make use of it. If a notebook can't be designed to cool a 100w TDP GPU then better cooling technology needs to be invented.

 

Nice post, Phinagle.

Would you expect an overclock to make up the difference, or is there more at play than clockspeeds? This is assuming that the Core clock can be brought near the reference 608MHz.

 

To bring the clock near the reference speeds, you're probably going to have to increase the voltage. Note how the stock GPU that this is based on sinks 50% more power.

 

Mmm yeah, I def won't to be the first in line to guinea pig that one.

 

The clock speeds appear to be the only difference between the GTX 465 and 480M. Both cards have 352 shaders and GDDR5 with a 256-bit bus. That said, I don't see where the power to overclock this will come from: even if you can do it without touching the voltage, the power consumption of a semiconductor scales linearly with its frequency. If you do have to up the voltage, that goes into the same equation quadratically. I'm not sure how the motherboards and PSUs of laptops will deal with this; my guess is that they simply won't.

 

it just seem, crude. too much is sacrificed for that little extra performance imo.

not sure if this applies to semiconductors.

 

I think it still does; here's a wiki article about it.

 

I think people should read that GTX 465 xbit review. And realize that the GTX 480M won't be as powerful, not nearly as they are speculating now.

xbit concluded while the GTX 465 may have good tessellation power, it's texture fill rate and just power is lacking. I think this will be even more noticeable in the mobile version. Power is significantly reduced with underclocking.

Which means looks to me Nvidia fans will be disappointed. I have no doubts it will be the king of notebook GPU. But not by much and I think a lot users will ask, for larger notebook, more weight, more heat and tons more power consumption, it's pretty disappointing compared to the HD5870M.

Or rather this is the perfect time for AMD to release Cross-Fire on the HD5870M full force. Which will really put the pressure on Nvidia as then people will be able to get that at the same price and power as the 480M... but better performance. But then some say 480M SLI, yeah I don't think even a lot of hardcore high end notebook users would consider that.

 

You are just speculating.... and you talk about the 5870 like it would run very cool, with tests on furmark putting it close to 100C.... im not trying to bash the 5870 at all, i think its a wonderful card performance/price, but dont put down the GTX480 yet, until you really see it. If it fails so be it, if it screams and leave the ati as no2 contender then so be it also, just let it be, its already coming, so lets just wait and see it runing before drawing conclusions.

BTW that 465 review is based also on beta drivers, probably not even optimized for that card, let it mature some, as the desktops 5870/5850 have like lets say.... 10.3.

 

I have run Furmark extreme burn never went over 87C with a 800/1100 overclock without a notebook cooler or back cover on. Sager owners reported the same, around 86-87C.

In game with 800/1100 overclock I played BC2 for 2 hours with notebook cooler, never went over 77C.

Sorry to say you are very much mis-informed about the HD5870M. For the power it has, it runs phenomenally cool and I have no doubts it runs even cooler in a Sager in game.

 

I'm guessing about 30 minutes battery life for these in a laptop SLI'd

 

The Clevo gets less than 50 minutes if you are lucky with a single HD5870M now, with 128 bit GDDR5.

Would you like to guess again? But yeah I don't think Clevo even cares, they will probably start advertising their notebooks as laptops for your desk, a desktop replacement, both figuratively and literally.

Enough for me on this, I'll wait and see how it is, not that it matters I have no plans to buy a desktop labeled a notebook.

 

Thus the overclocking discussion.

It all depends on whether the 425/850/1200 can come close to the ~600Mhz/1200MHz/1600MHz reference.

It's just like how we compared the 5870M to the 5770.

 

After all the talk about power requirements, price and desktop comparisons, has anyone actually seen what the notebook variants do in practice? Since the most powerful 5 series cards are out for ATI, it would only make sense to wait until the 480m turns out in a notebook and then do a proper comparison test.

Some users actually rank performance over every other factor, and it's only fair to those users that proper comparisons between the 2 notebook variants are conducted before baseless(?) speculation is bandied about.

 

It will considering that the specs are the game (which means usually that its using the same chip at the GTX 465 but with lower voltage running through it), but the probably of it getting to that point with reasonable tweaking is a question. Most Fermi cards need a voltage tweak to get the core past 10% of stock speeds. To get this card to that speed, it'll definitely need some voltage tweaks and also cooling tweaks to manage the increased heat. Also throw in PSU tweaks too if you're really determined. On the plus side, Fermi cards do scale better in performance when voltage tweaked compared to Radeon cards.

But again, the GTX 465 is really GTX 280 performance once drivers iron out. Yet an OC'd Mobility 5870 @ 900mhz core and 1100mhz memory is as fast as a HD 4890, which went head to head successfully against the GTX 280. This should be some healthy competition in terms of overclocked performance.

With that said however, 8500 Vantage GPU @ stock settings or BUST.

 

Well, 465GTX is reported be range vastly, from desktop 5830 performance to even faster than desktop 5870 performance depending on the game, and it's still using a beta driver so the potential is definitely there. The 480M is expected to be around 280M SLI performance, so it should be only 10-20% slower than a 5870MR in CF, and considering the myriad of problems for dual GPU solution, 480M could be worth it's price for a powerful single GPU setup, like W870CU/W880CU.

Also, for the TDP, you do know that ATI's stated TDP is only for the chip right, while Nvidia's stated TDP is for the whole card, so you should know that the actually TDP difference is not that large.

 

very nice. Time to upgrade from my 9800M GTX SLI. 100W TDP? who cares? - only an ATI fanboy. fortunately the new 480M will have the " ATI fanboys" option in the drivers. ^^ cant wait for it and teh results from putting that card into SLI.

 

haha I was just poking fun. I have an affinity for all cars,graphics cards, etc. I just go toward the bang for buck deals usually since Im not made of money.

 

I'd say I lean toward the AMD/ATI side of things (despite the hardware in my sig), but even still, we can't be making these snap judgements on hardware that has yet to be seen in the wild.

Until a NBR forumgoer, NBR itself, or Anandtech/PCPer/MaxPC/Bjorn3D/etc get one of these and tests it, I'm going to reserve my judgement.

Sure, I'd *expect* the 5870 to be a more efficient chip, but there is no real way to know until we do some objective and verifiable comparisons! And as others have said, we should all WANT Nvidia to do well, simply so that ATI has a good reason to keep competing.

How many of us expected ATI to take so much of the mobile market in such little time compared to 2 years ago? Things can change fast!!

 

I don't think I'll get this after all.. I can't imagine laptop can last long with such powerful component inside. I'd say 1 1/2 years max then all are starting to give problem.. oh well, uneducated guess from my part XD

Oh and, for that price to upgrade from 5870M? nty..

 

I deleted a lot of bickering posts; please try to have a constructive conversation from this point forward. Thank you.

 

My post was constructive. Anyhow, ATI has outdone Nvidia in the 5800 vs 480 war.

 

How is ATI going to respond to this? No doubt the 6000 series will be reaching upwards on 100TDP as well to match this power in a single configuration. This sort of power is only for the chosen few its like one person chooses to buy a Mini an one choose to buy a WRX one is going to use more fuel but in the end the buyer knows and expects that "Battary life" Its the price you pay for more power.

This will also be beneficial for the cooling situation. New technological advances will be made in that area as well to handle all this extra power. The 100w TDP step had to be beached at one point or another if this is the only way to move forward to keep these gamers happy then so be it. Cant wait for the bench tests.

 

I don't think ATI has to "respond," to be honest. They're in a pretty damn good setup, right now. Considering the price of these things, it makes so much more sense to go 5870CF.

 

If you can go for 5870MR CF that is, for single GPU notebooks like W870CU/W880CU, 480M is the best performance choice available, and for those who goes for pure performance without regards to price, 480M in SLI will still win out in performance.

Nevertheless, I predict that 480M will have a price drop soon, the current price is a bit on the ridiculous side.

 

I'm going to go out on a limb and guess money and portability are no object to you. Jolly good for you then, but something tells me it's more than just "ATi fanboys" who find those things important, especially in a notebook. More power consumption + more heat = less battery life, thicker/heavier machines, shorter lifespans, larger power bricks, and higher power bills.

I'm afraid this is the same story as the GTX 480 on the desktop; sure, it's faster than the 5870, but I'd wager very few people would put a 10% performance gap over a $100 price difference, twice the power consumption, and a heat output you could literally boil water off of. But even then, more people are going to be willing to make that jump on a high-end desktop, where figures like performance/watt and heat output are easier to forget. Performance/watt and heat output, though, are simply things you can't just throw aside on a notebook.

 

1. Clevo W880CU with 480M is lighter and smaller than even Asus G73. So the more TDP = Larger and heavier is moot.
2. It's a lot more than 10% performance difference.
3. For power consumption, ATI has high power consumption on load as well, won't be much lower than 480M, it's only ATI's idle power consumption that's low.
4. Doesn't matter what's the heat output if the notebook can handle it, and I won't have any doubt about handling the heat if it's Clevo.

But I do agree that the price sucks.

 

I'd definitely like to see some comparisons as far as temps in the w870cu once the 480m comes out.

Even with a 800/1100 overclock on the 5870m, and playing bad company 2 for a couple of hours, my highest temp on the 5870m was 67c according to amd gpu clock tool. Highest temp reached on the 920xm was 66c.

 

yes, 5870MR on my 8690 reaches 72C max on 12min of furmark, so 480M is no problem even without a cooling mod, and Clevo's response of not giving 480M to 8690 due to heat is BS since 8690 has just as good cooling as 8760.

 

i don't know if the assumption that resistance stays constant holds.

 

As long as the temperature is kept constant, the resistance should stay at the same value.

As for the performance of the 480M GTX, I also agree we should wait for real world benchmarks before making any judgement on it.

 

I take it you guys have seen the correct specs?

Mobile 480

352 cores
425mhz core
850mhz shaders
600mhz ram (GDDR5, 256bit)

100 Watts (GPU + Mem)

From Home - AnandTech :: Your Source for Hardware Analysis and News who confirmed the memory clocks with Nvidia.

 

It's no way about the core, its about the VRMs and notebook cooling. Which gives out first depends on the cooling.

That or they have castrated the voltage so it may not clock at all.

 

If ATI really wanted to respond to this, all they'd have to do is release a binned MR 5870 (call it 5875 or something) that runs at higher voltage and has higher clocks. I bet they could raise the clocks enough to beat the 480M and still have lower power consumption and temperatures.

That said, I don't see any reason for them to respond. This card is going to be in very, very few laptops (Clevo, Alienware, maybe a couple of others) whereas the MR 5800 series is in multiple mainstream machines. Even in the laptops that support the 480M, I suspect all but the "must have the highest performance" people will go for the MR 5870 because it's so much cheaper. This has practically no impact on ATI.

Only if ATI's management has gone completely insane. First, they don't need 100W to beat this -- they can do it with less power even with their current architecture. Second, this goes against their philosophy.

 

I'm a noob at this stuff but even i can tell that's not true.

I also like how most people here seem to have an ATI GPU in their sig.
I'm not saying that what people claim is false, but let's wait for the thing to COME OUT first?


edit: personally i prefer laptops under 17" so i'll pass this one just because of that :F

 

5870 mobility crossfire will perform like a desktop 5850.