gt 445m performance benchmark

from here
¡¾K580ÔËÐÐ3Dmark06µÄÕæÊÇ·ÖÊý¡¿-ÉñÖÛ¾«¶Ü K580ÂÛ̳-ZOLÖйشåÔÚÏß

 

Well if they should run 3D Mark 05 1280x768 doesn´t count since standard res when benchmarking 3D Mark 06 is 1280x1024.

 

We shouldn't really use 3DMark06 anymore. It bunches different performing GPUs together too much compared to Vantage.

 

GPU GF106M

GPU Clock 570MHz

CUDA Cores 144

Processor Clock (MHz) 1140 MHz

Memory Interface Width 128-bit

Memory Clock (MHz) 2500 Mhz

Memory Type GDDR5

 

Nice, will this be a cheap mid range card or more expensive like the 260GTX?

 

The GT are usually Nvidia's mid-range cards.
GTX are the high-end ones.

 

Nice its a very good performance for a midrange card. Pretty much like the 260GTX which is high end today. I might get on of these

 

Nvidia may devour the mid-range space, if this thing is 35W or less.

Makes the 480M look even more silly, especially if the GT 445M is faster than the GTX 260M.

 

Comparing to results from this thread:
http://forum.notebookreview.com/gam...rds/507342-gt-445m-no-faster-than-gt335m.html
something seems strange...
CPU is weaker, but still, almost 3000 points difference-maybe hard OC?

 

that is strange...TBH I was very surprised to see the 445 break the 10000 pt barrier...I'm calling BS. Its an OC or something. The record for the 335 is 7700, and that was OC'd so I dont thing this is possible (with stock clocks).

 

Those previous benchmarks were done with a massivly less powerful CPU on the 445.

 

Why would you expect the 445 to be comparable to the 335? Nvidia is known for its shady renaming, but they've never made a GPU with the first 2 digits higher than a previous one without it being better. I'd still wait for a review with official drivers, but this thread strikes me as far more plausible than the one claiming it is no better than a 335.

 

My GTX 260M, Vantage GPU score, overclocked to 600/1500/900:

This has double the shaders, 72 -> 144, and GDDR5. Why are you surprised?

 

The lower res probably gives the score a good boost too. I reckon its 9000 which isn't surprising as Kevin says as it is a new architecture and GDDR5 and double shaders.

 

I went ahead and ran 06 @ 1280x768:




I've hit 12k before, but not today. It's probably a driver thing. Again, my GPU is overclocked to 9800 GT spec.

This 445M is really impressive. I'd like to see what the 460M can do.

 

Well, since we're using a comparison to the GTX 260m, have a look at mine. (CPU is a T9900 C2D at 3.06 GHz, GPU clocks as seen on GPU-Z in the screenshot)

The 445m looks pretty decent.

 

Forget 3dmark06. Way old. Nowhere near resembles modern games let alone dx10 or dx11.

This should be a bit faster than a GTX260M. Anything else will tell us there are some architectural improvements or flaws.

And we can't forget drivers may be in an embryonic state which may mean there's plenty of optimisation (and hacks :/) to be done to bump scores.

The shaders (processors or whatever) are only 2x core speed unlike the GTX260M which is 2.5x so that should pretty much cancel the shader count advantage. Memory bandwidth is increased obviously but only by 25% or so due to 128bit memory bus. ROP performance I'd be interested to see especially if there's only 16 of em.

OCing prospects should be good though. Esp if this does fit 35W - but I think THAT's a bit optimistic... would mean 60W to 35W for similar performance in one process shrink + new uarch.

I think 460M will be ~ 5870M. And Nvidia's marketing as a result will be all about tesselation and boring stuff about GPGPU...

edit: hah! seen it already!
http://www.asusforum.com/showthread.php?p=14789
"The G53jw-A1 with the GeForce GTX 460M has 4 times faster FPS versus Radeon HD 5870 in Microsoft DirectX 11 SDK subd11 tessellation test with maximum tessellation level set."

 

I Wont be surprised if GTX460m was actually on par or faster than 5870m
Remember that Ati is releasing Radeon 6000 by the end of the year so nvidia gotta compete with that not their 5000m cards unless they like to lead sales for 2months only

 

Nvidia's tesselation performance is good but drops off when the core is very busy so simple benchmarks with high tesselation for no good reason (think heaven benchmark) are not very accurate.

 

so far these are lineups we know of

gtx 480m 352 core 256bit gddr5
gtx 460m 192 cores 192 bit gddr5
gt 445 144 cores 128 bit gddr5

any other line up specs?

clearly there is no space for gts 450m i really hope there is a gtx 470m with 240 cores/256 bit and gtx 485m with 1gb/256 bit/336 cores

 

There is supposed to be a GTX 470M w/ unknown specs, GTS 450M with 160 SP and 128-bit GDDR5 interface, GT 425M with 96 SP and 128-bit GDDR3 interface but until they're seen, we'll just have to wait to confirm specs. Though the mobile workstation variant of the GTX 480M, the Quadro 5000M is slightly weaker at 320 SP but keeping the same 256-bit GDDR5 interface.

 

GTX465 was phased out, dont see why the 480M its based off cant be.

 

If a GTS 450M exists it would be GF106 based and since you can't get 160 SP out of clusters of 48 the core count would have to be either 144 or 192.

I wouldn't be surprised if GTX 470M was in fact using the same 320 core GPU that the Quadro 5000M uses...that would mean though that it's a GF100 derivative, and that's the hot and crappy Fermi.

 

Still, I fins these scores a bit weird now.

For example, the new GT445m has 144 shaders 128 bit GDDR5 memory bus width amounting to 40GB/s and 580 clocks or something like that right?

My GTX 260m is 112 shaders, 500mhz clock 256 bit GDDR3. Save for a bit of difference in the bandwidth, the new card has higher everything, yet it doesn't score more. It's like GF100+ is less eficient than G92b?

I don't have high hopes for GTX460m anymore

 

GT 445M has 144 CUDA cores but if it works like GF104 then it's only got the schedulers for 96 CUDA cores. The remaining 48 CUDA cores have to share schedulers if they become available...and that doesn't always happen.

On top of that GT 445M has a lower memory clock than the GTX 260M and if the GPU-Z screenshot is to be believed it's only got 1 ROP and a lousy fillrate. (I don't believe that ROP count is correct )

 

OMG can you imagine only a 1 single ROP?!!? And your explanation makes more sense about the cores, which only leads me to the usual disappointment

So the GT445 is more akin to a GTS360m with DX11 capabilities.

Man, I so want an HD5870m now.

 

Remember, the TDP is a lot lower and you can expect a significant difference in power consumption. For that drop, I could take the hit in performance.

 

Yeah, but the performance hit means these new cards, with more shaders, are hardly more powerful.

I mean... the 480m barely beats the 285m and HD5870. The 460 will probably be on the same level as 285m with DX11 capabilities.

It's not really an upgrade. Man, I expected more but well now I can properly plan my next purchase and not go expecting sick things from the current GPUs.

 

well i think nvidia wont release any new gf100 core for mobile or even for desktop parts and its good.

i really hope if there is a gtx 470m it should be of gf 104 and also gtx 485m iif they do there will be a gpu technology conference in sept 20 to 23rd hopefully it will be exposed by then still its too late .

 

If the GT 445M is going after the GT 335M's market, in that sense, it's a huge upgrade

If the GTX 460M is as faster than the 285M, it's pretty much as fast as the Mob. 5870. I see the 460M as Nvidia's new mainstream performance, 15" chip. The 480M is already EOL, and Nvidia will release a GF104 chip which takes the crown, while being more efficient and less expensive. I expect it to come from the desktop GTX 460; the question is, how long will the wait be?

 

Yeah I understand you, and that's what I initially hoped until I sat down and actually looked at the specs. At this point I think the 460m will be around the performance of the 285m but with DX11 capabilities.

That is a lousy increase for the "high end" area, while the medium range got a superb increase in performance, matching a GTS 360m.

I am now considering waiting until 2011 for Sandy bridge+ Radeon 6000 series.

 

i think gt 445m will be faster than gts 360m

because the desktop gts 450 is 192 core and the probable gt 440/445 which is against desktop 5750 is 144 core which in turn is gt 445m

now the gts 360m is a desktop gt 240 which is under desktop 5670 so gt 445m's desktop counterpart is going againg desktop 5750 so gt 445m might very well be the entry level high end card for mobile or at least upper mid range

it might even be the 3rd fastest card in the lineup of g 400m series

though i hope the lineup will be

gtx 480m
gtx 470m (really want this one if 240 sp 256bit gddr5)
gtx 460m
gt 445m (might beat or equalize with gtx 260m)

anything below doesnt matter as the next probable card will be 96sp or less and will be of gts 360m caliber or less

 

Nvidia bringing something to market that's equivalent to a Radeon except it draws more power and is 8 months late... pretty much the story of GF100 from the start.

Not feeling too bad about the gtx260m in this thing, even if it is a shrunk & rebadged 8800. Seems GPUs will only be improving by increments in the next few iterations as they have in the past, rather than the massively huge improvements we've seen in jumps like from Geforce MX (5) to 6 and then 7 to 8. They're all hitting power usage limits or going way past them (e.g. desktop 5970 downclocked to fit 300W, GTX480M breaking the record with 100W)


(Gimme GTA: San Andreas on a 8600M GT over GTA4 on the fastest GPU, anyday. Powerful GPUs can only polish turd games so much.)

 

According to several reviews (such as HardOCP and TechReport, for example), the new GTX 460 is a better value and performer than it's ATI counterparts, especially when in SLI (HD 5830/5850). Better late than never, but the new GF104 cards are much better and more efficient than even the last iteration GF100 cards.

 

How about "the 5870m is 25% faster then the 460m GTX in every other game"

NV fanboys trying to brag while AMD is clearly in the lead...

 

put one in the next MBP please!

 

Not necessarily. The compiler bundled with drivers that came out after GF104's release should mitigate that somewhat. Of course, the performance won't equal a 144 CUDA core version of the GTX 2xxM series, but it really shouldn't be worse than the 112 core GTX 260M, should it?

Ha yeah, the desktop GTX 460 is just begging to be shrunk into a mobile card. Those power and temperature results are pretty amazing.

 

Don't know if anyone is interested in benches anymore for this card but I'll post my ones:

Last 3DMark Vantage I did today: P7162

Last 3DMark 06 done (best was 12kish): 11470

Also did a Furmark test and go 2488 points there.

The card is overclocked at setting 725/931, using MSI afterburner.

So far I find this card to be very nice.

 

hi redlance,

I got my L701x a few days ago, seems very nice so far. it runs crysis 1 very smoothly - would you happen to know the performance in crysis 2? I want to find out if its playable at high settings with similar framerates to crysis 1 before i buy it.

thanks!

 

Without overclocking, you can play smoothly at high settings. If you overclock it, should be able to play smoothly at very high settings. After overclocking, you will get about 20-30 fps at extreme settings which is playable but I always prefer higher fps >35.

 

very good to know, thanks mate!

 

No prob Have a look at this thread http://forum.notebookreview.com/del...driver-vs-modified-nvidia-driver-445m-23.html
It has got more information about the performance and overclocking of GT 445m.