rMBP: successor for the 650m

[ppone]

I think the biggest upgrade for next year's 15 inch rMBP will be the discrete GPU. I hope it is much more powerful than the 650, to cure any hiccups the computer has in driving the retina display.

What do you guys think the successor will be?, 680MX?

The cpu side performance, won't be a huge gain. Not sure if Haskell will bring in DDR4 memory, which will also help in terms of performance, but I am already getting around 11000 on the 32 bit geek bench for the base rMBP.

The bottleneck of this year's 15 retina is the GPU. What are the chances that next year's dGPU will drive the retina flawlessly and not just alleviate minor lag and stability issues, but cure them?

 

[Freyqq]

Probably the 750m or whatever comparable chip from AMD that is out by May-ish of next year.

 

[psykick5]

Why would the successor be the 680?? The TDP of that chip is much higher than the rMBP's. The new chips will make it the 750m or if they do a TDP reduction maybe a 760m. The 650m is as good as can go in here without new chips coming out.

 

[nontroppo]

What are the chances that next year's dGPU will drive the retina flawlessly and not just alleviate minor lag and stability issues, but cure them?

100% because, unless you have proof this is a hardware problem (all the evidence points to it being a software issue), the 650M is already more than capable of running the retina display...

 

[icecoldart]

I think the biggest upgrade for next year's 15 inch rMBP will be the discrete GPU. I hope it is much more powerful than the 650, to cure any hiccups the computer has in driving the retina display.

I love how little people know and go out loud with it.
ANY hiccups that come from a fine ( not defected rMBP ) are related to the HD4000. Scroll lag etc, if you switch to 650m i doubt you'll see any slow down, the downside is shorter battery run time.

and 680mx is not possible in that form factor, the next gen successor like previous poster have already said, something like 750m or equivalent.

 

[wiznet]

I love how little people know and go out loud with it.
ANY hiccups that come from a fine ( not defected rMBP ) are related to the HD4000. Scroll lag etc, if you switch to 650m i doubt you'll see any slow down, the downside is shorter battery run time.

Actually, no it's not.

Contrary to popular belief, the hardware is actually capable enough to handle the retina resolution well. Software is the cause for said hiccups.

 

[DeusInvictus7]

Actually, no it's not.

Contrary to popular belief, the hardware is actually capable enough to handle the retina resolution well. Software is the cause for said hiccups.

Exactly. Even the HD 4000 is capable of driving the retinal display just fine, it's just the software and apps that have to be updated properly to support it.

 

[Fatt]

It appears nvidia will be releasing a kepler revision early next year, and not maxwell just yet. So expect only a 15% to 20% performance increase from say a 650m to a 750m. Should see a larger performance jump when maxwell is ready. However, apple has a tendency to switch GPU vendors each gen, the rMBP revision may very well be an AMD chip. Which AMD is also rumored to be working on a revision of southern island, before releasing sea island.

Basically, for both nvidia and AMD don't expect more than another 15% or 20% bump in performance for the next release of consumer cards. As they will be revisions of current architectures and not the new ones we have been expecting. But when we do finally see maxwell or sea islands expect a significant jump in performance.

 

[leman]

I love how little people know and go out loud with it.
ANY hiccups that come from a fine ( not defected rMBP ) are related to the HD4000. Scroll lag etc, if you switch to 650m i doubt you'll see any slow down, the downside is shorter battery run time.

I did some testing with Safari and Quartz Debug and I didn't see any noticeable difference between the HD 4000 and 650m. Actually, HD 4000 has all the cards to be faster for UI rendering, as it can optimise many things like avoiding texture copies. HD 4000 has direct access to the system RAM and it also uses the CPU L3 cache. UI rendering is still a mixture of CPU and GPU work, and because HD 4000 is basically part of the CPU, it can sync data more efficiently. The actual workload the GPU has do perform (which mostly boils down to texture filtering) stays relatively low compared to GPU insensitive applications like games. This is of course speculation, I don't have anything to actually prove or back up my theory.

 

[whitedragon101]

It appears nvidia will be releasing a kepler revision early next year, and not maxwell just yet. So expect only a 15% to 20% performance increase from say a 650m to a 750m. Should see a larger performance jump when maxwell is ready. However, apple has a tendency to switch GPU vendors each gen, the rMBP revision may very well be an AMD chip. Which AMD is also rumored to be working on a revision of southern island, before releasing sea island.

Basically, for both nvidia and AMD don't expect more than another 15% or 20% bump in performance for the next release of consumer cards. As they will be revisions of current architectures and not the new ones we have been expecting. But when we do finally see maxwell or sea islands expect a significant jump in performance.

Gah beaten to the punch....

Incidentally Maxwell will be an insanely large leap forward if they deliver on their roadmap. Kepler (ie the 650m) delivers 5 GFlops per watt, Maxwell will (they say) deliver 15 Gflops per watt !!

 

[stevelam]

So the consensus is apple is just bad at writing software and apparently has been terrible at it for a long time. So what's that say about any real improvements to the RMBP? The 2.0 update basically did nothing.

 

[derbothaus]

So the consensus is apple is just bad at writing software and apparently has been terrible at it for a long time. So what's that say about any real improvements to the RMBP? The 2.0 update basically did nothing.

I think it says that you should always listen to people on when to take a pass on version 1 tech. It is a golden rule. They'll fix it but probably not the model you have. I have an iPad v1 that can attest to that.

 

[stevelam]

I think it says that you should always listen to people on when to take a pass on version 1 tech. It is a golden rule. They'll fix it but probably not the model you have. I have an iPad v1 that can attest to that.

Well this is apparently a software issue. Obviously the iPad 1 limitations were purely hardware based.

 

[derbothaus]

Well this is apparently a software issue. Obviously the iPad 1 limitations were purely hardware based.

Not obviously. They wrote everything since introduction to waste more cycles and need more power than was necessary. Rev 1 OS and apps were just fine. It got functionally hobbled by updates.