rMBP Thunderbolt GPU Good?

[lixuelai]

My reply was more in question of this statement of yours:

"It's like saying you can take 2 different train tracks that go to different stations and make it so that all trains going through those tracks end up at the same station"

But your subsequent explanation explained what you meant by that.

It would be much easier than what you propose in this scenario. The easiest way to fix latency issue is to buffer the data which GPUs already does via the VRAM. PCI-E is in this application is just transferring data from CPU/RAM to the VRAM. So they do end up in the same station, the VRAM. It is identical to the concept of multi-gpu rendering as well as link aggregation of pci-e network devices. as latency would also be a factor in that case as well.

 

[ScholarsInk]

Hey!

So im thinking about replacing my windows desktop with a rMBP, but the gt650 bothers me a bit, meanwhile iv been looking around for external gpu solutions that uses the thunderbolt standard but havent found any... yet!!?? my question is: how much will a thunderbolt gpu improve the already not to shabby gaming performance on a rMBP, and will it be able to use both ports for extra "bandwidth". If it was possible to push the rMBP 3dmark06 score over 22000 id be all over it!! fo sure!

The GT650M is actually clocked to faster than a 660, and I've been playing quite a few games on high/highest settings at full resolution. I very much doubt that Elder Scrolls Online could pose a problem.

 

[gentlefury]

Well, for starters, the rMBP graphics are smoking fast! Secondly, the classification as gt650 is kind of inaccurate considering it is overclocked to be faster than the GTX750...the OC is obviously to accommodate the extra screen res, but outputting to a 1080p display gives some really nice performance!

 

[MCAsan]

The 15" GPU infinitely beats not have one at all.

 

[Maven1975]

It's my understanding that you can not use an eGPU on the rMBP because it does not show the HD 4000 as selectable/boot video card under bootcamp.

I can also tell you that eGPUs always run better when connected to an external monitor.

 

[ssn637]

It's my understanding that you can not use an eGPU on the rMBP because it does not show the HD 4000 as selectable/boot video card under bootcamp.

I can also tell you that eGPUs always run better when connected to an external monitor.

How about the 13" rMBP in Boot Camp?

 

[imperialnavy]

There isnt a way to disable thunderbolt and use the 4 freed up pcie slots for the gpu, is there?

 

[Mr MM]

There isnt a way to disable thunderbolt and use the 4 freed up pcie slots for the gpu, is there?

nope, and it wouldnt matter at all.

 

[bill-p]

My reply was more in question of this statement of yours:

"It's like saying you can take 2 different train tracks that go to different stations and make it so that all trains going through those tracks end up at the same station"

But your subsequent explanation explained what you meant by that.

It would be much easier than what you propose in this scenario. The easiest way to fix latency issue is to buffer the data which GPUs already does via the VRAM. PCI-E is in this application is just transferring data from CPU/RAM to the VRAM. So they do end up in the same station, the VRAM. It is identical to the concept of multi-gpu rendering as well as link aggregation of pci-e network devices. as latency would also be a factor in that case as well.

Nope. Again, doesn't work that way. Buffer means the ability to store stuffs for processing later. It's not a magical way to fix latency problems.

There are many things to address to make 2 separate buses transmit the same data. I won't detail it because it pretty much requires a whole article to explain.

It's the same problem with parallel processing. You don't get 200% performance using 2 CPU cores to process stuffs. Same problem here. Theoretical bandwidth should be doubled, but due to latency and sync problem, the actual bandwidth that can be utilized is much lower. It's faster than a single bus for sure, but it's still not like a single bus that's twice as fast.

 

[Maven1975]

How about the 13" rMBP in Boot Camp?

Should work. I know it works with the 13" cMBP.

 

[bogatyr]

It's my understanding that you can not use an eGPU on the rMBP because it does not show the HD 4000 as selectable/boot video card under bootcamp.

I can also tell you that eGPUs always run better when connected to an external monitor.

Doesn't need to show the HD4000 unless you're looking to use Optimus to display the video on the internal LCD.

 

[dagamer34]

Well, TB is PCI-e 2.0 4x while GPUs are usually installed in a PCI-e 3.0/4.0 16x slot... so high performance gaming is out of question anyway. Besides, the rMBP is able to run games exceptionally well for the laptop its weight. I am getting 40+ fps with Guild Wars 2 on high settings at 1920x1200 and the gameplay is very smooth.

Anyway, using google gave me this: http://www.tomshardware.com/reviews/pci-express-graphics-thunderbolt,3263.html

Ridiculously expensive, but maybe its what you are looking for?

Disproven: http://www.anandtech.com/show/5458/the-radeon-hd-7970-reprise-pcie-bandwidth-overclocking-and-msaa

 

[leman]

Disproven: http://www.anandtech.com/show/5458/the-radeon-hd-7970-reprise-pcie-bandwidth-overclocking-and-msaa

Thanks for the link, very interesting! I assumed modern games would already surpass the bandwidth provided by PCIe-1 x16, but it seems that I was wrong. When I think about it again, it kind of makes sense... there is usually not that much need to stream data back and forth between the GPU and CPU (although I am surprised that its actually this little data ^^). What I would be very interested in seeing though is the FPS density instead of just an average - I'd like to know whether lag spikes are more pronounced with slower PCI-e setups.

 

[Mr MM]

Thanks for the link, very interesting! I assumed modern games would already surpass the bandwidth provided by PCIe-1 x16, but it seems that I was wrong. When I think about it again, it kind of makes sense... there is usually not that much need to stream data back and forth between the GPU and CPU (although I am surprised that its actually this little data ^^). What I would be very interested in seeing though is the FPS density instead of just an average - I'd like to know whether lag spikes are more pronounced with slower PCI-e setups.

That would be interesting a normal distribution would do wonders. or just be simple, put a min, avg and max fps in those charts

 

[IceHot]

Disproven: http://www.anandtech.com/show/5458/the-radeon-hd-7970-reprise-pcie-bandwidth-overclocking-and-msaa

So this means that i can run a 7970 and get pretty much the same performance as i would if i had a desktop with the same card?

 

[ssn637]

Should work. I know it works with the 13" cMBP.

Are you saying that you have a 13" MBP AND a Sonnet Express Chassis? If so, I've got a few questions for you...

 

[bill-p]

Disproven: http://www.anandtech.com/show/5458/the-radeon-hd-7970-reprise-pcie-bandwidth-overclocking-and-msaa

I think it's worth noting the 4x result in that article is PCIe 3.0, not 2.0.

PCIe 2.0 4x is actually about the same bandwidth as PCIe 3.0 2x. So to get the right image, look at the 2x results.

And you'll see how much more performance drops at that level.

Not to mention Thunderbolt doesn't exactly behave as PCIe does:

http://www.anandtech.com/show/6023/the-nextgen-macbook-pro-with-retina-display-review/11

You're looking at just a fraction of the theoretical PCIe 2.0 4x bandwidth.

Like I have repeatedly said so many times now: "it just doesn't work that way!"

 

[IceHot]

Well iv seen people getting 6k in 3dmark 11, my gaming pc can do 4,5 so if the retina can pump out 6k++ with.. say a 7970 im all over it.