With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.
I don't see how Apple's core philosophy is a liability. There's nothing stopping them from upping the clock speeds significantly, without entering the GHz race mind you. The thing's you talk about in the following paragraph...
...are things anybody would like to have in a processor. The reason why it is not as good as x86 in Handbrake encoding is because there is no hand-tuned assembly code for it, just basic support.
How are they going to up their clock speeds significantly without redesigning the core? As seen in the power scaling chart, making a core that is able to scale in both frequency and power isn't trivial.
Handbrake uses the x.265 codec for HEVC, which does have assembly code tuning since version 3.4. Though I agree it's not nearly to the same extent as x86-64 CPUs, I think the main reason why the M1 doesn't perform very well relative to x86 is due to having lower clocks and less capable SIMD instructions for encoding.
Why are you guys all talking about what Apple could do to add a bunch more cores to the Macbook Pro to compete with the most over the top PC laptop solutions? Apple never has and never will compete in the DTR market where laptops cannot be used on your lap, drain the battery in a couple hours under load, and cannot even reach full performance without being plugged in. If you want that kind of a "laptop" you will never be able to buy it from Apple, so it is pointless to speculate about whether the Macbook's cooling system could be adapted to become one.
Unfortunately, that's the ONLY way the M1max comes (today). On the x86 side, you have actual choices. I don't think it is unfair to compare those choices to Apple's choice to sell you their product one, and only one way.
Though Handbrake is kind of special case since the encoder is running hand tuned assembly code on the x86 side.
Another thing to consider is that even at 50W max, M1 Pro is hitting high 90s in temperature. There is virtually no thermal headroom left for Apple to add more cores since their highly dense SoC is producing too much heat to dissipate quickly. For higher core count, they will need to go with water cooling or die-stacking to create more space between the different hot spots or just wait till 3nm.
The RTX 3080 Ti is an even hungrier beast! I don't know how much of that 141w figure is represented by the GPU.
It looks like the high quality preset for HEVC encoding really punishes the M1. The HWU review used the fast preset and the M1 was competitive unlike this review.
It doesn't. Support has been merged but it isn't enabled in any release version.The ARM-compiled version presumably has hand-tuned NEON assembly code.
They don't need to scale to 100 W and target 5 GHz clock speeds by increasing the pipeline stages. It is just a matter of allocating the transistor budget - I'm sure M2 will have higher clock speeds, with it being on TSMC N4. Now fmax is 3.2 GHz, I wouldn't be surprised if Apple reaches 3.5-3.6 GHz with the M2. As much as Intel's power scaling looks impressive, in practice nobody is going to run a laptop CPU at 100 W sustained.How are they going to up their clock speeds significantly without redesigning the core? As seen in the power scaling chart, making a core that is able to scale in both frequency and power isn't trivial.
It's fair in reality but annoying theoretically. We'd like to test one variable at a time (the CPU) but it's impossible.Unfortunately, that's the ONLY way the M1max comes (today). On the x86 side, you have actual choices. I don't think it is unfair to compare those choices to Apple's choice to sell you their product one, and only one way.
It doesn't. Support has been merged but it isn't enabled in any release version.
Zen 2 got a significant boost in HEVC compared to Zen1/Zen+. This was largely thought to be the addition of full speed AVX256.
They don't need to scale to 100 W and target 5 GHz clock speeds by increasing the pipeline stages. It is just a matter of allocating the transistor budget - I'm sure M2 will have higher clock speeds, with it being on TSMC N4. Now fmax is 3.2 GHz, I wouldn't be surprised if Apple reaches 3.5-3.6 GHz with the M2. As much as Intel's power scaling looks impressive, in practice nobody is going to run a laptop CPU at 100 W sustained.
Alderlake H was not designed to perform without dgpu, there is a reaosn why they put weak igpu in there compare to Alderlake P.I think people would also be curious about the baseline: Under the same power envelope, without adding dGPU, how does Alder Lake perform against M1 Pro (and hopefully, finding Rembrandt as well).
From that baseline, it is fine to say there are options on the x86 side which don't exist on Apple side.
Alder Lake-H is the exact same die as Alder Lake-P, both are 6+8 w/ 96eu iGPU.Alderlake H was not designed to perform without dgpu, there is a reaosn why they put weak igpu in there compare to Alderlake P.
Intel has also said dGPU-less ADL-H designs are comingAlder Lake-H is the exact same die as Alder Lake-P, both are 6+8 w/ 96eu iGPU.
Intel laptops will get MX570/550 paired with those CPUs, which will be faster than any RDNA2 iGPU because only the expensive ultrabooks with LPDDR5 will give comparable performance. So for budget gamers, Intel + NVIDIA will still be the better option.Things are looking very good. AMD doesn't have anything to counter that level of performance other than promote their RDNA2 iGPU, which in these times, IS pretty compelling for the average gamer. So, the laptop choices are an OK CPU (AMD 6000 series) with great iGPU or excellent CPU (Intel 12th gen) with more expensive dGPU. AMD 6000 series laptops should be very popular with budget gamers.
MX570/550 don't have raytracing support. So AMD will have that as an advantage. Of course, the raytracing performance will be nothing to write home about but it could allow game developers to use even that little bit of raytracing power to spice up the visuals a bit.Intel laptops will get MX570/550 paired with those CPUs, which will be faster than any RDNA2 iGPU because only the expensive ultrabooks with LPDDR5 will give comparable performance. So for budget gamers, Intel + NVIDIA will still be the better option.
LOL nobody is going to use RT on an iGPU, not even with FSR.MX570/550 don't have raytracing support. So AMD will have that as an advantage. Of course, the raytracing performance will be nothing to write home about but it could allow game developers to use even that little bit of raytracing power to spice up the visuals a bit.
LOL nobody is going to use RT on an iGPU, not even with FSR.