Apple-exclusive Intel Core i9-10910 Rears its Head

[btarunr]

Intel is readying an Apple-exclusive Core i9-10910 desktop processor which will feature in an upcoming, unannounced iMac / iMac Pro product, according to a spot by _rogame. The i9-10910 sits between the i9-10900 and the unlocked i9-10900K that's available in the retail market. It has an interesting set of clock speeds. Its nominal clock speeds is significantly higher than the i9-10900, at 3.60 GHz, compared to 2.90 GHz of the i9-10900; however, its max Turbo Boost frequency is lower, at 4.70 GHz, according to Tom's Hardware, compared to 5.00 GHz on the i9-10900. Perhaps 4.70 GHz is the all-core TVB max frequency, a 100 MHz increase over the 4.60 GHz of the i9-10900. Also, its TDP is rated at 95 W (for a locked chip), higher than the 65 W of the i9-10900, but lower than the 125 W of the i9-10900K.

The i9-10910 is a 10-core/20-thread processor, just like the i9-10900, and features 20 MB of shared L3 cache, along with a Gen 9.5 UHD 630 integrated graphics. In related news, the unreleased iMac that was used in this Geekbench run also sports a Radeon RX 5300 discrete graphics solution, featuring 20 RDNA compute units (compared to 24 on the Radeon Pro 5500M), amounting to 1,280 stream processors; up to 1.65 GHz engine clocks, and 4 GB of an unknown memory type. It will be interesting to see if the i9-10910 remains Apple-exclusive after the Ryzen 9 3900XT launches next week.



View at TechPowerUp Main Site

 

[XiGMAKiD]

Most intriguing, I'm curious if this is Apple's x86 swan song

 

[lexluthermiester]

Most intriguing, I'm curious if this is Apple's x86 swan song

Very likely.

 

[ZoneDymo]

imagine making a special processor for your bae only to be dumped soon thereafter

 

[watzupken]

imagine making a special processor for your bae only to be dumped soon thereafter

There is nothing special about this processor, unlike the custom Ice Lake processors that Intel supplies only to Apple. This is basically a same i9 10900 with a base clockspeed bump, which can easily be done since there are no hardware difference.

In any case, whether they are getting dump soon, Intel is getting paid for the products they supply to Apple. While the switch from Intel to ARM is starting this year, it may take another couple of years for them to be able to fully move away from Intel, especially when you are looking at high end chips, where ARM processors are still not able to keep up at this point.

 

[lexluthermiester]

While the switch from Intel to ARM is starting this year, it may take another couple of years for them to be able to fully move away from Intel, especially when you are looking at high end chips, where ARM processors are still not able to keep up at this point.

That's a good point. However, Apple could move very quickly on this.

 

[efikkan]

imagine making a special processor for your bae only to be dumped soon thereafter

It's a different bin, not a special die.

In any case, whether they are getting dump soon, Intel is getting paid for the products they supply to Apple. While the switch from Intel to ARM is starting this year, it may take another couple of years for them to be able to fully move away from Intel, especially when you are looking at high end chips, where ARM processors are still not able to keep up at this point.

And then a few of years of support/spare parts.

ARM will not be able to compete with x86 in generic performance for "normal" desktops or high-end desktops. Just like their mobile products, they will rely heavily on ASIC acceleration to do the heavy lifting. Power users using Macs have been slowly disappearing for years, but after this only the die hard fans will remain. Macs have gone from pricey professional computers to being underpowered expensive fashion choices.

 

[Valantar]

There is nothing special about this processor, unlike the custom Ice Lake processors that Intel supplies only to Apple. This is basically a same i9 10900 with a base clockspeed bump, which can easily be done since there are no hardware difference.

In any case, whether they are getting dump soon, Intel is getting paid for the products they supply to Apple. While the switch from Intel to ARM is starting this year, it may take another couple of years for them to be able to fully move away from Intel, especially when you are looking at high end chips, where ARM processors are still not able to keep up at this point.

Nothing particularly special about those Ice Lake chips either - just different bins of the same silicon. Back when Apple was the only one using Iris Pro/Plus chips there was an actual silicon difference, but no more.

 

[trom89]

Macs have gone from pricey professional computers to being underpowered expensive fashion choices.

This!

 

[claes]

That's a good point. However, Apple could move very quickly on this.

They claim two years.

 

[$ReaPeR$]

the naming scheme... ridiculous. this doesn't look like anything special, as mentioned its just a different bin for the same CPU. it's very intriguing though the comparison that so many people make between arm and x86 as if things are in a vacuum, apple will, as it has done until now, optimize their software for arm making x86 basically irrelevant. they will have total control and the benefit of that is extreme optimization.

 

[Valantar]

the naming scheme... ridiculous. this doesn't look like anything special, as mentioned its just a different bin for the same CPU. it's very intriguing though the comparison that so many people make between arm and x86 as if things are in a vacuum, apple will, as it has done until now, optimize their software for arm making x86 basically irrelevant. they will have total control and the benefit of that is extreme optimization.

Why is the "naming scheme" (can you call it a scheme when it's a single CPU?) ridiculous? Giving a new bin with different clocks a new name is exactly how CPUs are named everywhere... Are you saying they should have called this the i9-10900 even though it doesn't match the specifications of that part?

As for performance comparisons with Arm and optimizations: you have far too much faith in what can be achieved with software optimizations. Apple still needs to come up with an Arm CPU that can compete in raw performance overall, which thus far doesn't exist from any vendor in any market segment beyond mobile. Having control over the software stack doesn't make the CPU dramatically faster, and if their Arm chips don't perform well one could still build a Hackintosh with the current MacOS and get more performance for less, as X86 MacOS will still be supported for years to come. Heck, going by the current mobile market one could make the opposite argument: Apple has complete control of software and hardware, have significantly faster hardware than the competition, yet in real world use cases their phones are no more responsive or faster than the competition.

 

[$ReaPeR$]

Why is the "naming scheme" (can you call it a scheme when it's a single CPU?) ridiculous? Giving a new bin with different clocks a new name is exactly how CPUs are named everywhere... Are you saying they should have called this the i9-10900 even though it doesn't match the specifications of that part?

As for performance comparisons with Arm and optimizations: you have far too much faith in what can be achieved with software optimizations. Apple still needs to come up with an Arm CPU that can compete in raw performance overall, which thus far doesn't exist from any vendor in any market segment beyond mobile. Having control over the software stack doesn't make the CPU dramatically faster, and if their Arm chips don't perform well one could still build a Hackintosh with the current MacOS and get more performance for less, as X86 MacOS will still be supported for years to come. Heck, going by the current mobile market one could make the opposite argument: Apple has complete control of software and hardware, have significantly faster hardware than the competition, yet in real world use cases their phones are no more responsive or faster than the competition.

I was talking about the Intel naming scheme in general, and I think we can all agree that, in general, is a tad ridiculous. As for software optimization, I don't have "faith",I just judge from what we can see, Apple, although I despise it as a company, has one of the most optimized OS s in the industry and possibly the best "experience" smartphones and PC's in the industry due to the whole ecosystem and the extreme level of optimization. So when they pull a move like this one I trust that they will make something out of it. Time will tell if I'm right or wrong, I don't know the future, I'm just judging from accuired knowledge.
Btw, even the 400$ iPhone can match much more expensive Android phones in the responsiveness department. And I will reiterate that I despise Apple.

 

[Valantar]

I was talking about the Intel naming scheme in general, and I think we can all agree that, in general, is a tad ridiculous. As for software optimization, I don't have "faith",I just judge from what we can see, Apple, although I despise it as a company, has one of the most optimized OS s in the industry and possibly the best "experience" smartphones and PC's in the industry due to the whole ecosystem and the extreme level of optimization. So when they pull a move like this one I trust that they will make something out of it. Time will tell if I'm right or wrong, I don't know the future, I'm just judging from accuired knowledge.
Btw, even the 400$ iPhone can match much more expensive Android phones in the responsiveness department. And I will reiterate that I despise Apple.

I think we all agree that the Intel naming scheme is a mess, but I don't see how this is the thread for discussing that - it was discussed aplenty back when Ice Lake and Comet Lake launched.

As for optimizations... well, the ecosystem has nothing to do with optimized software, just having a broad hardware portfolio with self-made software (from OS to a lot of the apps) and making the effort to have it all work together. They're very good at that, but it's proprietary nature makes it impossible to discuss whether it's well optimized - it's not like we can run it on Windows or Android to compare the performance after all.

You seem to either mean something else than software optimization (which generally means improving the code base of software to make it run as optimally as possible on the intended device(s), typically meaning increased performance), or to be conflating other advantages and strengths of Apple with this. Nothing I said goes against the quality of Apple's software or user experience, I am only saying that your line of reasoning presented here - "discussing comparisons between X86 and Arm is meaningless as Apple's software will be highly optimized" - makes no sense. The performance comparisons in question are also typically well optimized on every relevant platform - but that is down to the software vendor (say, UL for 3DMark, etc.) and not Apple - Apple doesn't make cross-platform benchmarking software, after all, nor do they optimize 3rd party software (though they might tweak the OS to improve performance running certain apps, but the effect of this is relatively small).

 

[$ReaPeR$]

I think we all agree that the Intel naming scheme is a mess, but I don't see how this is the thread for discussing that - it was discussed aplenty back when Ice Lake and Comet Lake launched.

As for optimizations... well, the ecosystem has nothing to do with optimized software, just having a broad hardware portfolio with self-made software (from OS to a lot of the apps) and making the effort to have it all work together. They're very good at that, but it's proprietary nature makes it impossible to discuss whether it's well optimized - it's not like we can run it on Windows or Android to compare the performance after all.

You seem to either mean something else than software optimization (which generally means improving the code base of software to make it run as optimally as possible on the intended device(s), typically meaning increased performance), or to be conflating other advantages and strengths of Apple with this. Nothing I said goes against the quality of Apple's software or user experience, I am only saying that your line of reasoning presented here - "discussing comparisons between X86 and Arm is meaningless as Apple's software will be highly optimized" - makes no sense. The performance comparisons in question are also typically well optimized on every relevant platform - but that is down to the software vendor (say, UL for 3DMark, etc.) and not Apple - Apple doesn't make cross-platform benchmarking software, after all, nor do they optimize 3rd party software (though they might tweak the OS to improve performance running certain apps, but the effect of this is relatively small).

I just made a comment, it wasn't my intention to start a conversation exactly because that horse have been beaten to death.
Now,as for the topic, my point is that they have absolute control over everything so it's much easier to optimize to such a point that the hardware differences will not matter. You know how people say "the computer that took us to the moon is slower than a game boy"? It is, but it doesn't matter because it's made specifically for one job with custom software. I hope this clarifies my train of thought. Of course until they release the product all this is purely theoretical so, just friendly conversation.

 

[Valantar]

I just made a comment, it wasn't my intention to start a conversation exactly because that horse have been beaten to death.

No problem

Now,as for the topic, my point is that they have absolute control over everything so it's much easier to optimize to such a point that the hardware differences will not matter. You know how people say "the computer that took us to the moon is slower than a game boy"? It is, but it doesn't matter because it's made specifically for one job with custom software. I hope this clarifies my train of thought. Of course until they release the product all this is purely theoretical so, just friendly conversation.

That is exactly where I believe you have a bit too much faith in the power of optimized software. The Apollo computer is as such a good analogy: it did its job, but due to its lack of power it did so with what would today be seen as entirely unacceptable margins of error.

Spoiler: Going off on a tangent about societal attitudes underpinning why this was acceptable then and isn't now

(This is largely due to changing attitudes towards risk and death in the line of such endeavors - today there is a general imperative in most if not all levels of society to keep people alive no matter the cost, while at the time of the Apollo missions the relatively high risk of death for astronauts was well known and broadly accepted, as it was for the test pilots for experimental aircraft that were the predecessors of early astronauts (and what many of them were before they became astronauts). Largely thanks to developments in medicine, the 20th and 21st centuries has seen a radical evolution from seeing death as an inevitable and relatively common part of life to something almost preventable and worth nearly any effort to avoid. But that's enough of a sociological tangent for today!)

It also had very specific tasks, and ones that were ultimately rather straightforward (if complex) mathematical calculations. This is of course a radical difference from the essentially unknown workloads the general-purpose computers of today must be capable of completing. As the total capabilities of the system become broader and more diverse, the opportunities for radical optimization will inevitably shrink. In such a scenario Apple is still situated optimally, with control over hardware (though not entirely on the architectural side, with it being licensed from Arm even if Apple does have the right and ability to modify the base architecture heavily), firmware (to a large part), OS and application software. But this nonetheless limits the effects of optimizations in the vast majority of scenarios to "make things work slightly better" rather than "increase performance by an order of magnitude" (which is entirely possible in simpler systems). So, as I have said all along, they still need to deliver hardware that is significantly more powerful than what they have today if they are to deliver a performance increase over their current, Intel-based solutions. Their mobile chips are great, but real-life application testing doesn't show that necessarily translating into perceptible performance advantages in iOS when compared to Android (though it does in some cases). Similarly, Apple has for years managed to make Final Cut Pro on Intel-based Macs outperform all competitors in Prores format handling, playback and editing, despite this being a general purpose compute platform - but no other application shows anything near that level of performance optimization, indicating that Apple (as the creator and owner of the codec) likely knows of something that third parties don't when it comes to efficient decoding. Expecting that kind of advantage to be more broadly applicable is unrealistic IMO.

 

[$ReaPeR$]

No problem

That is exactly where I believe you have a bit too much faith in the power of optimized software. The Apollo computer is as such a good analogy: it did its job, but due to its lack of power it did so with what would today be seen as entirely unacceptable margins of error.

Spoiler: Going off on a tangent about societal attitudes underpinning why this was acceptable then and isn't now

(This is largely due to changing attitudes towards risk and death in the line of such endeavors - today there is a general imperative in most if not all levels of society to keep people alive no matter the cost, while at the time of the Apollo missions the relatively high risk of death for astronauts was well known and broadly accepted, as it was for the test pilots for experimental aircraft that were the predecessors of early astronauts (and what many of them were before they became astronauts). Largely thanks to developments in medicine, the 20th and 21st centuries has seen a radical evolution from seeing death as an inevitable and relatively common part of life to something almost preventable and worth nearly any effort to avoid. But that's enough of a sociological tangent for today!)

It also had very specific tasks, and ones that were ultimately rather straightforward (if complex) mathematical calculations. This is of course a radical difference from the essentially unknown workloads the general-purpose computers of today must be capable of completing. As the total capabilities of the system become broader and more diverse, the opportunities for radical optimization will inevitably shrink. In such a scenario Apple is still situated optimally, with control over hardware (though not entirely on the architectural side, with it being licensed from Arm even if Apple does have the right and ability to modify the base architecture heavily), firmware (to a large part), OS and application software. But this nonetheless limits the effects of optimizations in the vast majority of scenarios to "make things work slightly better" rather than "increase performance by an order of magnitude" (which is entirely possible in simpler systems). So, as I have said all along, they still need to deliver hardware that is significantly more powerful than what they have today if they are to deliver a performance increase over their current, Intel-based solutions. Their mobile chips are great, but real-life application testing doesn't show that necessarily translating into perceptible performance advantages in iOS when compared to Android (though it does in some cases). Similarly, Apple has for years managed to make Final Cut Pro on Intel-based Macs outperform all competitors in Prores format handling, playback and editing, despite this being a general purpose compute platform - but no other application shows anything near that level of performance optimization, indicating that Apple (as the creator and owner of the codec) likely knows of something that third parties don't when it comes to efficient decoding. Expecting that kind of advantage to be more broadly applicable is unrealistic IMO.

Fair enough. I don't think we disagree anywhere, but written communication has it's limits. Anyway, we will see what they can deliver and we can continue the conversation then.