HURRAY: Nvidia 600 series not just Fermi!! (Kepler)

What worries me is the fact that nvidia probably will need more months to come up with something since theres zero traces of a 680m. And Clevo on the other hand will probably be pissed and it will not update the old 150hm bios to suit the newer card.
I mean, there's the EM now which probably is on their priority list. If things would come out faster probably they will still care about the existing models.
Realistically in, lets say, 4-5 months current generation of clevos will not have any reason to exist anymore with ivy.
And that sucks.

 

Its not just Nvidia who is slow though. AMD have been out with 7870 and 7850 for a while now and absolutely nothing about 7970M.

While TSMC could have major problems with 28nmm, AMD atleast have IBM producing chips for them

 

Its just human nature.

Its like when the government buys a hammer for $300, when the exact same one is $20 retail at Lowes. You and I both know its the same thing, but the fact that they can profiteer in one particular area is a bit frustrating.

 

lol the TSMC yields are @#!#

 

Testla is a completely different thing that what is mentioned in his post. Workstation cards like the Quadros and the FirePros from AMD are exactly the same hardwarewise to their gaming counterparts.

The Quadro6000 is a GTX470. The difference is driver related to how you access the GPU functions. Drivers for workstation cards are optimized for completely different operations, resulting in different performance compared to regular gaming cards. Additionally, workstation cards offer massive support in both drivers and technical support, even going as far as same day exchange faulty hardware. You pay for stability, support, and drivers more than the hardware.

Until not so long ago, one could simply get a gaming card and softmod it to become a workstation card. When it started to become a popular practice, Vendors started hardware locking the capability to change drivers to workstation type.

Tesla is made for HPC computing, with much higher number crunching capabilities and different core counts due to arrangements. Not a workstation card like the quadros or firepros.

 

Interesting, will wait for even more info on the subject

 

Even the tesla cards are built on the same tech as all the other cards. Besides the extra memory the tesla's have, a gaming card is technically just as capable. The main reason you see a difference is because nvidia have nerfed the double precision performance on the gaming cards to seperate the market, so they can charge insane prices on the tesla cards.

 

even the length of the cards (quadro 6000 vs gtx 470) is different, so you saying they made it different for fun purposes... I will get a very very detailed answer on this, already started researching, problem is I am very busy usually I don't have time to do serious research on this kind of a matter, but I WILL get to the end of this, I will write what I learned to here sometime soon.

EDIT: Memory bandwidth and interface are different for gtx 470 and quadro 6000, again we boiled down to CUDA cores means the same card.. I will get to the bottom of this, and I am quite confident that I will find they are entirely different except for the alignment of CUDA cores.

please read..

http://www.xbitlabs.com/articles/graphics/display/nvidia-quadro-5000_2.html

 

What must be clarified: They are made from the same die, sometimes parts fused off, at least for Fermi and older architectures

 

wait so do we know if the clevo w110er will have GDDR3 or GDDR5 ram?

 

yes it has GDDR3 and the W150ER (15.6 inch model) has a GDDR5 650m

 

The architecture is the same. The full enabled cuda cores, vs the same physical cuda cores 2 of the 32 stream cores disabled, makes it to the gaming card.

I am not saying each and every workstation card has a gaming equivalent, because they do have very different variations in the workstation realm (mainly lower end "gaming carrds" with more aggressive memory interface and quantity).

But in the end, the PCB board can be larger or lower, with the exact same architecture and reduced/increased memory interface. They migh have a GTX480, then double it's memory size and end up making a slightly bigger card.

The point is, the technology is exactly the same. They just acomodate it for a different purpose. You will hardly find dramatic differences between workstation cards and gaming cards, specially high end ones. Some will be exact matches, but in general the workstation cards on stock have higher memory installed.

Similar to how some vendors sell you their GTX560 version with twice the memory etc for example.

Or the whole different versions of the GT555 etc. But it's good you research about it, it's always good to learn new stuff and find other tibits. I have been researching GPUs for over 10 years now, ever since the rise of ATi into the gaming world.

Exactly this.

 

i wonder what next gen quadro M GPU's comes out and when! quadro 3000/4000M are monsters in OC, like 50-60% boost with easy tuning)

 

They do have the 5010M out, but its still a Fermi Part... Wicked.. But not Kepler.. I haven't seen any word on a "6000M" part.

 

Could we see Intel CPUs with Nvidia IGPs in the near future?

Nvidia Wants to Make its Chips Inside Intel's Factories

 

Sounds like it's not going to happen according to what Jon Carvill commented...

 

I was thinking that Intel bought Nvidia and used their graphic knowledge to combat AMDs Fusion APUs. Like AMD bought ATI.
It would then be two companies, AMD vs Intel, both in GPUs and CPUs

 

Huang is dreaming... and I'm very serious.

 

They need to buy nVidia for their technology. They probably can't because of Anti-Trust issues.

Would be interesting if Intel Spun off their APU/SOC processors into a separate business and bought nVidia with it. That would be quite cool.

 

wait wait wait, nobody is buying nvidia, they are going to tackle intel for cpu business with tegra, and they are seriously going at it

 

Not sure if it has been posted, but I just noticed update on Anandtech
AnandTech - NVIDIA's GeForce 600M Series: Mobile Kepler and Fermi Die Shrinks
GT 650M now lists clocks of 850MHz with DDR3 and 735MHz with GDDR5.
NVIDIA suggested that their goal is to keep products with the same name within ~10% performance range.

 

And that is why Intel should buy Nvidia. I doubt it will happen though

Yeah I know, they are pretty serious about the Anti-trust laws. But AMD was allowed to buy ATI. Why can`t Intel buy Nvidia?

 

AMD is a much smaller company compared to Intel. Even buying ATI, they are still a little fish. AMD is worth $5.6bn, Intel is worth $139.5bn! Intel is already in the hot seat for decades of anti-competitive behavior.

Nvidia and AMD are both competition, and AMD has publicly stated that Nvidia is their primary competitor now. If Intel snatches them up, it would basically put AMD in even more trouble from a competitive standpoint.

 

LOL seriously, is Intel that much bigger?! Man that was shocking to see

You make pretty good points. Intel and Nvidia merger would mean big trouble for AMD.

 

lol we should root for the underdog: amd

 

So, how much better will the 660m be compared to 650m?

 

Nobody knows yet

LOL pretty hefty turbo.

In case you people wondered how big the turbo of GT650M is, here it is at full load:



Here is the idle clocks

 

GPU Turbo has both positive and negative points. The positive I'm sure you guys know. However, it is the perfect excuse for manufacturers to use poor cooling solutions in some cases. It is like adding a new performance state and say Turbo clocks are not guaranteed to occur. So, beware of laptops with poor cooling system for your next purchase!

 

You bring up a very good point.

Turbo Speeds are becoming similar to refresh rate numbers on TVs. They are a selling point and they are using jacked-up numbers for all of the time.

I have seen Alienware Advertise 4ghz+ CPUs, when its the Turbo Freq they are advertising like its the steady clock rate. Bad juju.

 

Nah, it's not changed that much, manufacturers used to release machines with bad cooling that would fail, they still do.

 

Now, the cooling fails even more xD

 

W150ER GT650 1GB DDR5 benches:

http://forum.notebookreview.com/sag...ries-w110er-w150er-w170er-23.html#post8409920

 

I see there's still no GT 650m 2GB GDDR5 benches, but so far it looks pretty tough for mid-range

so I guess it's equal or just a bit faster than the GTX 560m

C'mon April, where are you, we need those laptops

 

35 fps Open GL is weak for a mid-upper card. So it is the path already expected from new Nvidia.

For example the HD7690M : HP Envy 15-3040nr Laptop Review - Notebookcheck.net Reviews get 43fps.

The GTX560M could get more also.
NVIDIA GeForce GTX 560M - Notebookcheck.net Tech

 

OpenGL performance have gone down with Kepler compared to Fermi. GTX 680 review showed that. But 5FPS under 560M isn`t bad.

OpenGL is very little used anyway, and the few games that use it is not demanding. Kepler shine where it matters, DirectX, especially DirectX11 which is great

 

yeah in directx 9 the difference is alot smaller then in directx11 where the difference is huge to fermi

 

Proper OpenGL performance has been one of Nvidia's biggest advantages over AMD so its kinda wierd to see such low potential on Kepler (Might be driver related?), but im guessing the Quadro versions will be better on that.

 

Very true. GT 650M score slightly under GTX 570M with 3DMark11 but the difference between them increase with 3DMark Vantage which is DirectX10.

Yeah since OpenGL is very much used with Quadro models they must have some other Kepler versions with better OpenGL performance.

 

do u think we will be able to oc the 650m to 570m performance?

 

I guess for 3dmark11, I don't think so for games though.. comparison is clear I guess 2x336 shaders as opposed to 384 shaders.. who knows, we will see as soon as someone gets hands on it

 

well i dont think the gt 650m is bad considering that same kinda performance last year was in a big bulky laptop

 

I don't have the time to look through this thread right now, so I have a few questions which are probably already answered hundreds of times:

1. Any release dates on the 650M/660M?

2. Will there be both fermi and kepler versions of the GPU's? If so, which of the two are they including in the upcoming Lenovo Y480/Y580?

3. Should I buy a gaming laptop as soon as 650M/660M rolls out or wait until we see some benchmarks from HD 78xxM and HD 79xxM?

 

1. Nothing Firm, just "Soon"
2. 650 and 660 are Kepler.
3. Tough Call. The 650 is on par with 7690M already out. The 660M is a better part, for sure, but we have no idea what AMD will release, performance wise.

 

Remember 660M probably isn't Kepler, just will be 40nm rebadge.

I was excited about it, but likely will wait to see if AMD's 7800 series will be GCN 28nm tech.

 

It actually is a confirmed Kepler Part, the 670M/675M are the Fermis.

NVIDIA GeForce GTX 660M - Notebookcheck.net Tech

 

650 is on par with the 7690m? I thought the 7690m is the rebadged 6770m ? so is the 650m on par with the 555m

 

well according to the Benchmarks for GT 650m(GDDR5) http://forum.notebookreview.com/sag...ries-w110er-w150er-w170er-23.html#post8409920

if you compare with NVIDIA GeForce GTX 560M - Notebookcheck.net Tech , it seems to be more or less on par with the GTX 560m

 

My Bad. .

 

Intel is the damn reason why we don`t see any laptops with the 650M and 660M yet. Damn you Intel and your meh attitute toward the release of Ivy bridge. And screw you AMD for not keeping up with Intel thus not pushing them to work harder. And screw you AMD for not telling us a single thing about the upcoming GCN mobile GPUs.
I curse you all

/rant

 

nice cursing there cloudfire, just don't curse too much as what we have is all thanks to intel