06-09-2012 11:04 AM
Hi, all --
My purchase decision is between the T530 and the W530.
Does anyone know the difference in performance between the T530's NVS 5400M and either the W530's K1000M or K2000M graphics??
Unfortunately, the sales guy I talked with drew a complete blank.
06-09-2012 08:15 PM
I'm also curious about this. I've found some numbers, but they mean nothing to me, since I know nothing about graphics cards.
This thread is a discussion of of the K2000M. They say that it should essentially be a GeForce580M with slower RAM, the specs of which can be found in this enormous Wikipedia page. This NVIDIA doc says that the K1000M has half as many cores as the K2000M, which makes sense, I suppose.
The specs for the 5400 are here. It has even fewer cores, which I'm assuming means it is less powerful.
I'm aware that I sound like I have no idea what I'm talking about, mostly because I don't. I'm hoping these docs will enable someone to explain it in terms of "It will play [LoL/Portal 2/ SC2/ TF2](I know that isn't very adventurous in terms of graphics intensitivity, but hey, it's what I've got.) on [settings] [at X fps/ without stuttering or lagging or bursting into flames]."
In short, I've done legwork and found some numbers but I have no context for them. I hope this will help someone more knowledgeable than I to answer the question.
06-10-2012 05:00 PM - edited 06-10-2012 05:15 PM
The assertion that the K2000M is based on the GeForce 580M is incorrect. The K2000M is built on the same hardware as either the GeForce 650M or 660M. This can be discovered because the K2000M has the same shader count as both the 650M & 660M and is based on the same architecture (named Kepler by Nvidia). The Kepler-based cards are based on a smaller process technology than the most recent previous generation (called Fermi), which means Kepler cards are more powerful and also can do more within the same or reduced power consumption, which is good for notebook owners.
Now we know the shader count and process technology of the K2000M, but Nvidia has not yet released the clock speeds. Based on last year's cards, Nvidia placed the 2000M between the performance level of the 550M & 560M, and so it is reasonable for the K2000M to be somewhere in the middle between the 650M & 660M, in terms of performance. So that means it's likely to have at least the performance of a 650M and slightly lower than or equal to a 660M using DDR3 memory (the 660M can also be made more powerful by using GDDR5 memory, which is not offered on the K2000M).
In terms of what it can handle, the K2000M should be able to handle every game currently out there at native resolution and medium-high settings, and should be capable of running many lesser-demanding games at maximum settings. Notebookcheck.net has an extensive list of mobile graphics cards and comparisons between them, if you want to know about the level of performance for a particular game.
The NVS 5400M from the T-series is a different matter. That's based on the 630M, which is a lower-performing card and in which most current games are playable on low-medium settings, but some more demanding games are not playable with acceptable frame rates (30 frames per second or more) even on low settings.
06-10-2012 07:01 PM
That gives a decent qualitative sense of it all ... and not terribly surprising, I think.
Any idea how something like Second Life would run on each graphics system??
Second Life recommends an NVIDIA 9600/9800, which are unknown to me.
(Second Life is not a requirement ... I'm just trying to triangulate.)
06-10-2012 07:11 PM
Another perspective ... from the viewpoint of a Quadro FX 570M, which is what's on my T61p.
It looks like the T530 graphics would be about 2-3 times faster
And the K2000M would be about 6-7 times faster.
... and that's based on fill rate and memory bandwidth.
06-10-2012 07:47 PM
Thanks, Djembe! So I guess that puts the K1000M somewhere between the two, since it has fewer shaders than the K2000M, but more than the 5400M? Is there a comparable GeForce card for the K1000M?
06-11-2012 06:24 PM - edited 06-11-2012 06:25 PM
06-12-2012 02:06 PM - edited 06-12-2012 02:09 PM
I will also join this debate cause I am also curios about the Quadro K2000M. Considering the shader count and power envelope, it is pretty clear we are dealing with Kepler
And here comes my rant against Nvidia. The 660M has same TDP as the 560M[75 Watt. confirmed by Dell/Alienware, Eurocom, etc.] Performance increase is not that worthy of praise. It isn't a bad thing, but having the same level of performance at the same power consumption will not convince me as a future customer.
Now, the big problem steps in. We all read the GPGPU tests where even the GTX 680 cannot beat the old GTX 580, although it has 3xShader Count. So far, from a gaming point of view: 384 Kepler Shaders = 192 Fermi Shaders. So K2000M will have roughly the same performance as the old 2000M. +/- 5%. But the most important feature of a Quadro cards is computing. That is where the Kepler gaming core has failed to convince anyone. It is my personal fear that the K2000M will be just a GK107 with optimized drivers. The Drivers alone cannot compensate the lack of GPGPU power, hence the K2000M would end up slower than the current 2000M.
However, I am willing to give it a shot and assume that the K2000M will derive from the big Tesla die. It is only my own assumption on how this Quadro chip should look like. If it holds to be true, that would be good. If not, well, pray to the God of Thunder that Nvidia drivers will save the day and give it equal performance to the old-gen Fermi. Or I'll owe everyone a beer for being wrong on this one
For the K1000M, the above will apply. Gaming performance somwhere around 540M. Computing-wise, ideal: faster than the current 1000M. Less than ideal: slower or on par with 1000M.
06-12-2012 06:27 PM
06-12-2012 08:22 PM