Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Raw CPU power will not save you now. Moore's Law just delivers more cores now not more GHz. We're in that many-core future many people predicted. Software will have to conform. As will the OSs.

OSs and plenty of software already support multithreading. I bet the support for multithreading is much bigger than support for OpenCL. Besides, GHz nor cores are everything. Architectural changes in SB make it ~15% faster clock for clock than Nehalem.

It is in no way marketing BS. AMD have thrown their weight behind it wholesale, and increasingly so are Intel. Even NVIDIA are supportive despite CUDA.

It is BS with low-end GPUs. Sure, it is useful if you have a tower with some real GPU horsepower in there but I doubt it has any impact on performance when dealing with low-end GPUs, like Intel IGP and Nvidia 320M. I don't see it being a huge deal for Apple when talking about 13" MBP, in the end it's more or less meant for consumers who have no idea what OpenCL even is. I'm sure Intel would have added support for OpenCL if it actually helped something.
 
Damn. No USB 3.0 until 2012.
Why didn't Intel put USB 3.0 in Sandy Bridge?
The Intel 6 Series PCH isn't much of a dramatic change from the previous one. Intel had plenty of time to spin out something new but using the tested design ensures reliability.

It is in no way marketing BS. AMD have thrown their weight behind it wholesale, and increasingly so are Intel. Even NVIDIA are supportive despite CUDA.
It makes for a nice sticker on your hardware and not much else.
 
nilka said:
I have now read through the entire thread, and I must say that to my dissapointment not one of the posts mention the impact that this can be for the iMac.

With both new processors out on the market and the new gfx cards from both nVidia and ATI out I think we will see a new iMac in the next 2 months. What is really enticing me is that we might actually get desktop graphics into the iMac. The new ATI cards use way less power than previous generation and these new cpus looks nice. With a nice handeling of switching between the inegrated and the discreet gfx card this would also limit the total poweroutput and fan noise after what you are doing.

I thought I read that, when a discrete GPU is connected to a PCI-E slot the IGPU is disabled. In the case of the iMac the ATI or NVidia GPU's are actually connected via PCI-E, the chip just happens to be soldered to the board and traces made directly to the controller. So i doubt there will be the switching you describe. To be clear tho I read this on the desktop SB articles, not sure if it is mentioned in the mobile articles. If you go to dailytech.com they have provide a list of SB articles some including the mobile versions.
 
OSs and plenty of software already support multithreading. I bet the support for multithreading is much bigger than support for OpenCL. Besides, GHz nor cores are everything. Architectural changes in SB make it ~15% faster clock for clock than Nehalem.

That is just prolonging the old regime. We're on the edge of a step-change in how we get performance from hardware. This is driven by more cores and the need for much smaller power envelopes (mobile and server).

It is BS with low-end GPUs. Sure, it is useful if you have a tower with some real GPU horsepower in there but I doubt it has any impact on performance when dealing with low-end GPUs, like Intel IGP and Nvidia 320M. I don't see it being a huge deal for Apple when talking about 13" MBP, in the end it's more or less meant for consumers who have no idea what OpenCL even is. I'm sure Intel would have added support for OpenCL if it actually helped something.

With the PowerVR stuff we'll even see this used much more extensively in iOS too. This is not the preserve of the high end. No device can afford to have any of its hardware sat there idling when it can usefully be used for compute.

Let's pop back to this and see how widespread OpenCL is in a year's time shall we? ;)
 
How come no one has mentioned that Intel has put DRM on the chip to satisfy Hollywood? I bet you all aren't too excited about that.

http://www.theinquirer.net/inquirer/news/1934536/intels-sandy-bridge-sucks-hollywood-drm
Don't use it. Have we talked about the remote shutdown feature, yet?

Not exactly. AMD 6870 has TDP of 151W which is exactly the same what 5850 has. The performance difference between those two is relatively small as well, 6870 is a bit faster. With desktop GPUs, Apple would have to use something like 5570 which is much worse than the current 5850M is.
I'm still waiting for the next generation of mobile GPUs from ATI and nVidia. It's old seeing relabels of the 400M and HD 5000 Series.
 
Let's pop back to this and see how widespread OpenCL is in a year's time shall we? ;)

Sure. I'm just suspicious because I've already waited nearly 1.5 years and the support isn't any better. When SL was released, people went crazy about OpenCL and it was supposed to deliver huge performance upgrades. As you can see now, it was adopted quite poorly and I would be surprised if there was a sudden boom of OpenCL in a year like you're saying.
 
Sure. I'm just suspicious because I've already waited nearly 1.5 years and the support isn't any better. When SL was released, people went crazy about OpenCL and it was supposed to deliver huge performance upgrades. As you can see now, it was adopted quite poorly and I would be surprised if there was a sudden boom of OpenCL in a year like you're saying.

The 3d apps are already leaning towards nVidia because the drivers suck less. Throw in CUDA renderers and physics engines and nobody really cares.

It's Beta and VHS.
 
Intel Insider

I think one exciting feature of Sandy Bridge (and Intel ecosystem) is new feature called Intel Insider. As I understand it provides better hardware based encryption which Hollywood trusts enough to allow streaming of movies in 1080p. At this time the details are rather sketchy. Apple is not part of Intel announcement and an Apple spokesman declined to comment. It looks like Apple prefers to stick with iTunes and its 720p thus falling further and further behind from PC world in video world (first BlueRay, then AirPlay instead of DLNA, and now that)
 
Sure. I'm just suspicious because I've already waited nearly 1.5 years and the support isn't any better. When SL was released, people went crazy about OpenCL and it was supposed to deliver huge performance upgrades. As you can see now, it was adopted quite poorly and I would be surprised if there was a sudden boom of OpenCL in a year like you're saying.

The problem here is that this isn't being driven by this market (yet). The software industry (outside of gaming) moves slowly, this is clear when watching this change from single to many-core.

As has been pointed out NVIDIA have a stranglehold on compute right now with CUDA but expect OpenCL eventually overtake that lead. No-one likes lock into one vendor but CUDA brings the performance and ease of use right this minute. Oh, and marketing hype.
 
So... because the IGP is not in an actual Macintosh ... the rating of the IGP sucks? Sounds flawed. If anything, it would be Apple/Intel's falt for cruddy support on the driver (or whatever is need for a processor die IGP) and reflect nothing of the IGP's true performance.

I believe it is faster than the 320M. But really, the 320M is pretty old now I think as of Jan 2 2011.

Learn to read. I'm reserving judgement on the new IGPS until they find their way into a shipping Macintosh and that shipping Macintosh is reviewed. Nothing is final prior to that.

Also the 320M is only a year old, nothing compared to how long Apple drew out use of the 9400M before switching to the former.
 
If you just bought a Macbook Pro, I want this thread to be the reason that you should have waited.

Just assume that this will be the biggest update ever, and that you'll miss it.
:rolleyes:
 
If you just bought a Macbook Pro, I want this thread to be the reason that you should have waited.

Just assume that this will be the biggest update ever, and that you'll miss it.
:rolleyes:

That's what everyone said about Arrandale :p
Remember the thread; waiting for arrandale
 
Do you care to explain why it matters if the benchmarks are performed on "(Apple) computers"?

Benchmarks show the performance of particular CPUs and GPUs __relative__ to other CPUs and GPUS. Hence, why most people aren't clamoring for Linux or Windows XP or Windows Vista or OS X 10.5 or OS X 10.6 benchmarks; it's superfluous...

Unless you think Apple is going to under-clock the CPU/GPU then your concerns aren't warranted.

First off, the Intel card only beat the 320M on low settings on the 5 games, already not a stellar sign. Secondly, I care as to how it appears in a shipping Mac versus in its first official tech demo as things can change between now and then and Apple's implementation of the chips may change between now and then. To say that drivers are the only difference and that its an inconsequential one is short-sighted. It all affects final system performance when we finally get the Mac that ships with it in the mail.
 
If you just bought a Macbook Pro, I want this thread to be the reason that you should have waited.

Just assume that this will be the biggest update ever, and that you'll miss it.
:rolleyes:

I just bought a MBP in November. I don't regret it at all (especially at $979 tax free). I simply don't need Sandy Bridge on my MBP. If I want to game, I'll play on my PC instead.
 
You'd be surprised....and wait until Lion - I expect extensive OpenCL use all over the place in the OS. By that point the software producers will be using it more widely too.
The question is whether AVX and Sandy Bridge's dedicated media encoder is an acceptable alternative to OpenCL running on an nVidia IGP or low-end discrete GPU. The most common use case for GPGPU is video encoding, and Anand's benchmarks have shown that Sandy Bridge's dedicated media encoder is faster than even a GTX465M using CUDA which is more mature than OpenCL. In other words, lack of OpenCL GPGPU acceleration is in no way an impediment for Sandy Bridge's video encoding abilities. As for everything else, especially third-party software, it's much easier for them to transition existing code from SSE acceleration to AVX acceleration for a noticeable speedup than to rewrite code in OpenCL. The AVX speedup may not exceed what OpenCL can accomplish, but if it's easier to realize, I think that will be most beneficial for most users.

EDIT: As an observation, it appears from Anand's review that Sandy Bridge brings noticable improvements in battery life where a quad core Sandy Bridge can get similar minutes of internet usage per MWhr of battery capacity as dual core Arrandale which is a significant improvement over Clarksfield quad cores. More likely than not, battery life concerns will be the primary factor in Apple deciding whether to go pure Sandy Bridge as opposed to Sandy Bridge + discrete GPU or Core 2 Duo + IGP or any other combination rather than GPU performance concerns for space constrained models like the 13.3" units. In other words, much better CPU performance than Penryn Core 2 Duo, equivalent GPU performance as the 320M, and noticeable improvement in battery life seems like a compelling argument even if OpenCL support is quietly relegated to CPU support only. Apple will no doubt focus on the video encoding acceleration performance improvement anyways.
 
Last edited:
The question is whether AVX and Sandy Bridge's dedicated media encoder is an acceptable alternative to OpenCL running on an nVidia IGP or low-end discrete GPU. The most common use case for GPGPU is video encoding, and Anand's benchmarks have shown that Sandy Bridge's dedicated media encoder is faster than even a GTX465M using CUDA which is more mature than OpenCL. In other words, lack of OpenCL GPGPU acceleration is in no way an impediment for Sandy Bridge's video encoding abilities. As for everything else, especially third-party software, it's much easier for them to transition existing code from SSE acceleration to AVX acceleration for a noticeable speedup than to rewrite code in OpenCL. The AVX speedup may not exceed what OpenCL can accomplish, but if it's easier to realize, I think that will be most beneficial for most users.

You're mixing layers here. AVX is, very broadly speaking, a vector processing unit. OpenCL is a software layer and will be able to target that too as a compute resource. I also think Intel will bring an IGP with OpenCL/DirectX 11 support at some point. It has to to bridge the gap between now and the commoditization of the Larrabee-derived Knights Corner/Ferry.

I've not said anywhere that Sandy Bridge is of no use, or that AVX doesn't have some significance. My point is that Apple do focus on OpenCL and that has some bearing on when, and if, they adopt a hardware choice that doesn't accommodate that well.
 
Whatever processor is inside the MBPs. I don't fancy buying any of them until the screens are IPS. If the iPad and the iPhone sport them, then the margins should be good enough for the MBPs to have them by now.
 
I always find the chip naming schemes funny. What's next 'rocky road'? :)
Anyway, improvements are a plus.
 
They just started the whole GPU stuff a few years ago. You cannot do this from one day to the next.

Wasn't the GMA 950 released like 5 years ago? The fact is every IGP release has seen very dismal performance gains. The sandy bridge IGP is a prime example, a IGP that barely tries to stack up against a IGP that was released 2 years ago.

Who really wants to play games at low settings? There's a lot of games you can play at medium with the 320m yet the sandy bridge IGP ALTHOUGH paired with a faster CPU is 26% slower than a 320m at medium settings? Screw intel and their lack of focus in their GPU sector.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.