Pic of GTX 480M + 3DM06 Test

Is that the GTX 460M? I thought the GTX 460M is supposed to have 240 shaders?

 

Phinagle said the 270M with 240 cores (if one exists). Still, it's very interesting to see the cores/vs. core speed game they are playing.

 

Nice clocks, BTW. Eager to see some benches.

 

Nice find. What machine is that out of? EDIT: NM Found it. ASUS Lambo VX7. So it means the G53 is using something else.

That looks like a GF104 with some cores cut out.

 

what tpd does it have?

 

There must be something weird. I've found this link:
NVIDIA GeForce GTX 480M Powered Notebook Sneak Peek - HotHardware

Considering that the 480M GTX is teamed with a desktop processor, the boost doesn't look that huge. One of these benchmarks must be fake.

 

Quick comparison:


Desktop cards paired with a i7 965:

 

circa 75W.

 

Why does the GTX 480m have GDDR5 if it's only going to be clocked at 600MHz, which could easily be reached on GDDR3?

 

I'm not completely sure of this, but from what I remember, GDDR5 has higher bandwith. It also uses lower power, which Nvidia probably felt would be good, considering the 100W TDP of the 480M.

 

I was wondering if GDDR5 had any additional benefits besides the possibility of being clocked higher, but that answer makes me suspect it doesn't.

It's unfortunate because they could be cheaper, and of course this cost will be passed onto us.

 

GDDR5 is twice the bandwidth of GDDR3 so you can get faster memory speeds at lower clocks that result in less heat.

 

The 460M looks boss...

192-bit, 192 shaders, GDDR5, and 675mhz core. Honestly wouldn't be surprised if it scores better than the 480M in some regards.

 

Uh it's only 192 shaders, around half of that of 480M, I won't be surprized if it's even slower than 5870MR.

 

This makes sense now, since GDDR3 on a 256-bit bus is the same as GDDR5 on a 128-bit bus, even when the GDDR5 isn't clocked twice as high.

 

So you would certainly expect a 256bit card with 2GB GDDR5 (480M) would dust either a 128bit 1GB GDDR5 (5870) or 256bit 1GB GDDR3 (285M) card...and it is also not surprising that it would be quite a bit more expensive

 

At least they could put less memory to make it cheaper. 2 GB is such a waste, not even the desktop 480 has that much.

 

More expensive is the only thing that would be certain.


Different architectures are different and their memory utilization can't be compared clock to clock.

 

Memory bandwidth for reference:

GTX 285M 65.28
GTX 480M 76.8
Mobility Radeon HD 5870 64

 

SemiAccurate :: Nvidia GTX480M SLI notebooks spotted and benched

Already posted?

i7 desktop + 2x 480M GTX = 16620 points in 3DMark06 ?!?

There must be something wrong.

 

That score is for one card, as shown in another review, disabling one 480M gives 16-17k 3dmark06 score, SLI gives 17-18k scores, obviously driver not ready yet.

 

Definitely a driver issue.
M17X with 920XM+5870CF @stock clocks hits 19K in 3DMark06.

 

Mobility driver shouldn't be that different from desktop tho. Consider 480M is very similar to GTX 465

 

More likely it's ~65-70% of the GTX 465

 

that would still be damn fast.

 

All I want to know, at this point, is how well it overclocks.

 

id guess not great since the gpu core is so low.....

 

I think that the low core was more for heat reasons.

 

I found some tests on notebookcheck in game and it's easy to compare with the other GPU like ATI 5870M : Notebookcheck: Computer Games on Laptop Graphic Cards

 

good find, if these are accurate its about 5-25% faster, of course we have no idea what kind of cpu was used. This looks like the first real benchmarks to me.

 

Those benches look rather unimpressive, BUT!...
The drivers aren't polished yet. So we might see some improvement in a few months.
To begin with, I wouldn't rely on notebookcheck here.
In general, it's too early to draw conclusions.

 

Has notebookcheck been proven to be unreliable? their data has at least from ive seen been correct.

 

Not unreliable but they sometimes compare GPU's with different CPU's which makes there benchmarks a little off.
They generally do a good job review laptops but comparing GPU's you have to take with a grain of salt.
Just my 2 cents.

 

well ya thats true, thats why i said who knows what cpus were used, but id guess chances are at the very least the 480 was tested with a mobile i7, and perhaps with something bigger. at any rate with all the various tidbits of benchmarks we have seen, im not blown away by the new card and am officially calling shenanigans on folks who were saying it would be 50-100% faster.

 

Yeah, I agree, their reviews are very decent and far less biased and marketing oriented than most.
When looking at their GPU scores, always click details on the right of the score bar and you'll have the specs of the systems that were used to get those marks. That gives more precision.


For example So this is a D901F with a desktop [email protected].
Suddenly the picture looks worse, IMHO.

 

Hmm... For tessellation enabled games like Dirt 2, the FPS seems to have a 50% gain over 5870MR. Even though it's using the 940, but I seriously doubt Dirt 2 is bottlenecked by a mobile i7.

None tessellation enabled games have very small difference.

 

According to Notebookcheck:
Clevo W880CU
Single 480M
i7-620M 4096 MB
(Standard 1280x1024)
3DMark06
14133 pts

Same but with an i7-920XM - 16000pts

Same test with a D900F:
Single 480M
i7-940 6GB

3DMark06: 16390 pts

Same config with 980X - 18731pts

 

Well, IMO vantage and 06 marks no longer matter, because the 480M is built for tessellation, and nvidia never claimed major performance gain for non-tessellation rendering.

But for tessellation enabled games, the performance gain is indeed quite huge.

 

So how many games are there ready for tessellation benefits? How many are scheduled for the next year?
As someone said:
If you want to drop your frame rates really really quickly for little to no gain in visuals, then tesselation is what you want
When tessellation is mature - we'll have helluva lot more benefits, in a few years, when both 480M and 5870M will be a long forgotten history...

Another thing regarding Dirt2:
Look at the FPS: even though there's a huge gain 113 vs 170+, - will you be able to see the difference? Starting from 30FPS the game play is considered smooth. IF a single 5870 can offer a 30FPS on ultra, then with CF you're golden
And for a gamer that's more than enough (to a certain extend I know, I know.. Need bigger guns ).

And the worst yet, in the overwhelming majority of your favorite games you won't see a difference.

And all that disregarding the fact that a desktop CPU was used in those tests.

Dammnn, I still want that card,...

 

The same argument could be used in defense of 480M and tessellation.

Even with Dirt 2, a tessellation heavy game, the FPS is smooth enough, what games out there without tessellation will you ever need the GPU power for? So does it matter if 480M doesn't perform much better in non-tessellation enabled games if it doesn't even matter?

The point for tessellation is not how wide it's implemented, but the fact that tessellation enabled games will be the GPU intensive games that will actually need the extra power.

An analogy is: My missile launcher sucks against human targets compared to your M16, but it owns tanks, even though tanks are rare but there's tons of human targets, the point is, it is the tanks that matters, because my missile launcher will blast a human target up no problem as well, but your M16 can't handle that rare tank that pops up.

 

So basically, you're saying that you would pay twice for a card that runs much better in a few titles (when other cards can run there smooth ), most likely more power hungry and still can't beat Crysis and Metro, even when SLI'ed? That's your missile launcher?

Tessellation is a big word, too big for 2 games that you will abandon in 2 days to return to WOW

If the nearest future proves more favorable for these new cards (better optimized drivers, no power issues) and we'll see better results in games - I might even grab one.

But so far - 0-5% gain in 98% of the current games
up to - 50-80% gain in 2 games
on a desktop platform vs a mobile platform

up to -100-140% higher price.
up to - 20-30% more power consumption.

Dunno....

 

Tessellation can also be turned off in those few titles that support it.

 

Turn off and you don't have the eye candies.

Well, Metro is smooth on high, only on ultra does it get low FPS, so if it's in SLI I think it will handle it quite well.

Well, never said it's worth it, but it might just be worth for some who likes to blast those tanks that's popping up once a while. My cousin is still using an E8400 with 4850 on his desktop because he only plays WoW, so it depends on whether if it's worth it for you.

 

Interesting benchmarks ...

Im still sorta shocked how close the benches are for non Tessellated games...

I remember when the 5870m was officially 128bit and everyone was screaming about how bad this was including myself to a degree but honestly that has not held the card back at all and these DX11 benches are very similar to what i see with my 2x5870s in my desktop vs my friends 2x480s except that we are talking a 128bit card vs 256bit on the mobile end...

 

Totally agree.
I still very much like to see how deeply these scores are affected by a stronger CPU.

 

Yeah i cant wait for the head to head W880CU vs the W870CU, same CPU etc benchmarks ....

 

If you look on notebook check attentively - Notebookcheck: NVIDIA GeForce GTX 480M, you will notice that they have most of their benchmarks based on the D901F with the i7-940. For the ATI I think it's the i7-820QM because that's the only other Clevo they have in the reviews with the 5870MR or maybe it's an average based on all other reviews, it's hard to tell .

Either way, it looks like my predictions on performance were correct. The 480M is about 20% faster than the 5870MR in average and it's shinning in tessellation. The Vantage score is about 9000-9200 points which is exactly what I have predicted. I feel somewhat sad that I was right, I really wanted more, but it gives me confidence in trying to predicted the performance of other future mobile cards based on their desktop counterparts. It is also possible that future drivers will improve performance although the same could be said about ATI.

 

The GPU score is what I want to see. I suspect the CPU score is skewing the overall score up, but I don't know how much.
EDIT: Well, what do you know. Looking at the benchmark list, the GPU score is only ~8k. Looks like my latest guess of 8.5k was actually too high, though I think 8k will be towards the lower end of the spectrum for stock 480M GPU scores. In any case, that's not much of an advantage over the 5870.

@Blacky - unless you were predicting 9k for overall Vantage score with the i7-940 in mind (and if you did so your skills are pretty amazing, because I don't even know what Vantage's formula for combining GPU and CPU scores is), the GPU score is the one to focus on in Vantage if you're comparing GPU performance. With the same CPU, the 5870 would not be far off that score, but Vantage seems somewhat biased against Fermi as compared to real-world games.

I too have been predicting roughly 20% average improvement, albeit with great tessellation performance; it's a nice feeling to see one's predictions confirmed.

 

http://forum.notebookreview.com/sag...gddr5-dx11-physx-100w-gpu-13.html#post6215758

But I really got the number of shaders wrong and yes that I was taking into consideration a very powerful CPU because it was based on desktop benchmarks which use desktop CPUs not mobile ones (PhysX off).