Support NeoGAF

[GymWolf]

Noob question, is this the revolutionary intel cpu that fans were waiting for? or it's the next one?

 

[nightmare-slain]

Noob question, is this the revolutionary intel cpu that fans were waiting for? or it's the next one?

Well people have been wanting intel to move to 10nm for years. This is it…. they’ve been on 14nm since 2015 atleast for desktop cpus. Some of their mobile/laptop cpus are 10nm but this is the first desktop ones.

 

[nightmare-slain]

My 9900 K destroys anything I throw at it but I’m gonna need to upgrade if I want a PC with DDR5 and PCIE5.0. My 9900K runs HOT and these new ones are 128-228W???? Lololol.

im gonna hold off and wait for reviews and see what AMD do with their next Zen release. Maybe it’s time to try AMD.

 

[Haggard]

Except the 12900K is likely going to beat the 5950X in cinebench and if Intel is smart they will price it at or below the 5900X

With 10nm vs 5nm they simply can`t beat AMD, at least not as long as performance per watt is a factor at all. They´d have to stoop to unheard new pricing lows for a new flagship product.....
If they`d already use TSMC´s 3nm capacities for it then that would be an entirely different thing, but the way things are with Intel being years behind in manufacturing processes i´m not holding my breath for any of their products.
Doesn`t help that AMD`s next gen is already confirmed to be going for those 3nm processes either.

 

[GymWolf]

Well people have been wanting intel to move to 10nm for years. This is it…. they’ve been on 14nm since 2015 atleast for desktop cpus. Some of their mobile/laptop cpus are 10nm but this is the first desktop ones.

So it's finally time to upgrade my 8600k i guess.

 

[nightmare-slain]

with 10nm vs 5nm they simply can`t beat AMD, at least not as long as performance per watt is a factor at all.

You know you can’t compare nodes like that. Different processes

 

[nightmare-slain]

So it's finally time to upgrade my 8600k i guess.

You’ll need to upgrade your entire PC. New motherboard, new cpu, new ram, new ssd if you want to make use of pcie5.0 and ddr5. well I suppose you could keep using a pcie 3/4 ssd but definitely new ram and motherboard.

 

[GymWolf]

You’ll need to upgrade your entire PC. New motherboard, new cpu, new ram, new ssd if you want to make use of pcie5.0 and ddr5. well I suppose you could keep using a pcie 3/4 ssd but definitely new ram and motherboard.

That's the plan.

I need something powerfull enough to not even consider buying anything third party on console because right now my ps5 kinda trade blows with my pc (i still prefer the pc because i can tailor the setting and the games cost almost half the price).

 

[Haggard]

You know you can’t compare nodes like that. Different processes

We`ve seen Intel`s processes and how behind they are ad infinitum these last few years. TSMC`s efficiency level is completely out of reach for them atm.

 

[Kenpachii]

I mean doing even 10 seconds of Google research before talking about it would help.

This is an ARM-style big.LITTLE design. The big and little cores are different cores. Sources are saying the little cores in this are based on Intel Atom cores (Gracemont), the big cores are unknown but unlikely to be same as Rocket Lake because that was such a disaster (Golden Cove, whatever that is).

Also if they are literally quoting 228W at max P-state that's fucking hilarious. The ENTIRE COMPUTER Mac Mini M1 pulls 40W from the wall running full tilt, that's the whole computer with M1 chip and motherboard components. The M1 chip itself, same one in MacBook Pro, is probably pulling around ~15-20W. The M1 is an 8 cores (4 big, 4 small) chip. I mean sure, it's 8 cores instead of 16 putative on this Intel thing, but it's also pulling maybe 20W versus 228W. Intel is an absolute joke at this point.

Good point, didn't readed it clearly enough. but yea 230w for a CPU is nuts.

 

[ethomaz]

They have been doing this for a while already 9900k is the same.

All cores its 4,6ghz i believe, 1 core or 2 cores is 5ghz.

luckely pretty much all of those cpu's are easily overclockable to 5ghz manually.

Seems like this is more of the same. Also pretty sure AMD does the same.

That is Turbo technology.

All Intel processadora released have the exactly same CPU cores… the news here is that now some CPU cores are smaller and others bigger.

 

[ethomaz]

Good point, didn't readed it clearly enough. but yea 230w for a CPU is nuts.

The actual CPUs are 250W for PL2.

 

[ethomaz]

I mean doing even 10 seconds of Google research before talking about it would help.

This is an ARM-style big.LITTLE design. The big and little cores are different cores. Sources are saying the little cores in this are based on Intel Atom cores (Gracemont), the big cores are unknown but unlikely to be same as Rocket Lake because that was such a disaster (Golden Cove, whatever that is).

Also if they are literally quoting 228W at max P-state that's fucking hilarious. The ENTIRE COMPUTER Mac Mini M1 pulls 40W from the wall running full tilt, that's the whole computer with M1 chip and motherboard components. The M1 chip itself, same one in MacBook Pro, is probably pulling around ~15-20W. The M1 is an 8 cores (4 big, 4 small) chip. I mean sure, it's 8 cores instead of 16 putative on this Intel thing, but it's also pulling maybe 20W versus 228W. Intel is an absolute joke at this point.

228W if for PL2 when the whole CPU is at max turbo… it is less than previous Intel processors that have 250W for PL2.

Your comparison of a Desktop CPU with an mobile CPU is nuts… you should wait the mobile version of these processors to compare with M1.

It is a improvement for Intel but still seems to lag behind AMD with 150W PL2 for 5900X… even so it is 12 vs 16 cores.

 

[Drew1440]

x86 isn't designed for the ARM-style big.LITTLE, will all 16 cores be able to run concurrently like ARM? Windows assumes every CPU is identical, how is load even going to be allocated between big and little cores? How is this even supposed to work, Intel?

Windows Phone supported this kind of setup with ARM (Qualcomm), shouldn't be too much trouble to get it to work with ARM

 

[ZywyPL]

Why on earth would ever anyone need a big.LITTLE design in a stationary PC that's plugged into the wall?? I don't even like the philosophy behind the design in smartphones to begin with. In the meantime recent rumors say that Zen4 will be up to 16C, up to 170W, at 5nm process node.

 

[Armorian]

Why on earth would ever anyone need a big.LITTLE design in a stationary PC that's plugged into the wall?? I don't even like the philosophy behind the design in smartphones to begin with. In the meantime recent rumors say that Zen4 will be up to 16C, up to 170W, at 5nm process node.

Yeah it's stupid but this 8\16 big cores will be enough for high end games trough this gen

 

[VolticArchangel]

Leaked top secret footage at Intel labs captured during the R&D of their 10 nm fab:

Spoiler

 

[MikeM]

That's great. Can someone dumb this down as to what the benefits are?

 

[jaysius]

You’ll need to upgrade your entire PC. New motherboard, new cpu, new ram, new ssd if you want to make use of pcie5.0 and ddr5. well I suppose you could keep using a pcie 3/4 ssd but definitely new ram and motherboard.

Sometimes it seems like PC gaming is a worse addiction then gambling with the money you need to burn to keep up to date.

 

[martino]

Looking to upgrade my i7 7700k but I'm def waiting for pci 5.0. 228 watts is just wild tho

same plan here

 

[Armorian]

That's great. Can someone dumb this down as to what the benefits are?

You have 8/16 high performance cores great for gaming and most work tasks, plus 8 smaller cores that will probably be only useful for heavy threaded programs.

This will work good only if OS sets correct affinity on cores/threads, here W11 comes along.

Edit: Plus DDR5 and PCIE5 support and integrated Xe GPU I belive. This CPU won't be bad but probably Intel will fuck up prices/availability or something.

 

[DonkeyPunchJr]

Sometimes it seems like PC gaming is a worse addiction then gambling with the money you need to burn to keep up to date.

Sometimes it does. Although over the last few years, used PC parts have retained their value surprisingly well. So net cost to upgrade is not so bad if you sell your old stuff on eBay.

 

[JohnnyFootball]

a DDR5 motherboard should be able to also handle DDR4 ram?

No. Most definitely will not happen.

 

[JohnnyFootball]

Except the 12900K is likely going to beat the 5950X in cinebench and if Intel is smart they will price it at or below the 5900X

I'm always amused by how many people automatically assume Intel will regain the performance crown. It's a completely new architecture that isn't remotely close to being as mature as 14nm. It will take a few generations of optimization. Intel fanboys automatically think that 10nm will automatically have all the refinements that 14nm had.

In fact, it was speculated that one of the reasons it has taken so long to get 10nm on desktops was that they could not hit the performance that matured 14nm CPUS were hitting.

Time will tell of course.

 

[Bitmap Frogs]

I just read the confirmation on Tom’s.

The specs are final but the clocks are Enginer Sampler yet.
The clock will probably change in the retail version.

The biggest news is Intel buying GlobalFoundries… $30 billion.

That’s ironic considering those used to be AMD old fabs. Bet Jerry Sanders won’t be happy about his fabs being sold to Intel.

 

[DonkeyPunchJr]

I'm always amused by how many people automatically assume Intel will regain the performance crown. It's a completely new architecture that isn't remotely close to being as mature as 14nm. It will take a few generations of optimization. Intel fanboys automatically think that 10nm will automatically have all the refinements that 14nm had.

In fact, it was speculated that one of the reasons it has taken so long to get 10nm on desktops was that they could not hit the performance that matured 14nm CPUS were hitting.

Time will tell of course.

I guess we’ll see. What I’ve read is that Intel’s 10nm was very similar to TSMC’s 7nm, and part of the reason for the delay was because Intel was very aggressive with what they were trying to do.

Hard to believe they originally were supposed to launch 10nm desktop CPUs in freaking 2016.

 

[ResilientBanana]

I'm always amused by how many people automatically assume Intel will regain the performance crown. It's a completely new architecture that isn't remotely close to being as mature as 14nm. It will take a few generations of optimization. Intel fanboys automatically think that 10nm will automatically have all the refinements that 14nm had.

In fact, it was speculated that one of the reasons it has taken so long to get 10nm on desktops was that they could not hit the performance that matured 14nm CPUS were hitting.

Time will tell of course.

This particular design has been in the works for quite some time. Sure, AMD is in the optimizing phase, but Intel will regain the performance crown with this design most likely.

 

[bbeach123]

Ryzen 5000 series already like 1 year old when Alder lake release so I think its safe to assume alder lake can beat it .

Zen 4 vs raptor lake though . Idk

 

[ethomaz]

That’s ironic considering those used to be AMD old fabs. Bet Jerry Sanders won’t be happy about his fabs being sold to Intel.

Well GF is not tied to AMD anymore.
In fact AMD have a contractual clause that force them to use x% of GF production… AMD is trying for years to get out of GF to the point that all high-end chips today are produced at TSMC.

Maybe that will be a easy way to AMD get out of GF.

 

[TheThreadsThatBindUs]

Seems like a interesting approach like ARM processors with performance cores and energy saving cores... so you get both the power and low power consumption.

I understand the approach but it doesn't appear they're all that successful in it. The low power profile (presumably when the larger performance cores are switched off) reveals this vividly. Gotta assume they're still burning power wasting transistors on x86-64's increased instruction set cores for the low power option.

Would have made sense to get an ARM license and follow what Apple has done with the new MacBook processor, using small low power ARM cores to emulate x86 for the lower-power cores.

I get that it's easier for Apple because they control the entire software stack and don't need to worry about legacy third party enterprise software that still requires the inclusion of ancient CPU instructions. But maybe there's a wider need for Intel to shake up their product lines here and branch enterprise cores from retail consumer desktop cores so as to trim the fat on the latter and be able to better compete against the encroaching ARM.

 

[ethomaz]

I understand the approach but it doesn't appear they're all that successful in it. The low power profile (presumably when the larger performance cores are switched off) reveals this vividly. Gotta assume they're still burning power wasting transistors on x86-64's increased instruction set cores for the low power option.

Would have made sense to get an ARM license and follow what Apple has done with the new MacBook processor, using small low power ARM cores to emulate x86 for the lower-power cores.

I get that it's easier for Apple because they control the entire software stack and don't need to worry about legacy third party enterprise software that still requires the inclusion of ancient CPU instructions. But maybe there's a wider need for Intel to shake up their product lines here and branch enterprise cores from retail consumer desktop cores so as to trim the fat on the latter and be able to better compete against the encroaching ARM.

It is a top desktop processor after all.
Apple M1 is a mobile focused processor.

There is no leak about the new Intel mobile processor that will be compared with Apple M1.

The processors in the OP are on the same segment of Ryzen 5900X or Intel 11900X.

 

[TheThreadsThatBindUs]

It is a top desktop processor after all.
Apple M1 is a mobile focused processor.

There is no leak about the new Intel mobile processor that will be compared with Apple M1.

The processors in the OP are on the same segment of Ryzen 5900X or Intel 11900X.

I'm not really commenting on the specific Intel product line here, but rather the more general approach.

Consuming less power while providing similar or even better performance (as in the case of the Apple M1) is also beneficial at every level and in every product segment.

At the top end, as is discussed in the thread here, using significantly lower power ARM cores for the 8 E-cores while providing similar performance, would not just enable a lower power consumption in the lowest power profile, but also more importantly provide more overall power availability for the larger cores in it's highest performance mode allowing those cores to be possibly clocked higher or even allowing them to include more SRAM cache on-die to decrease further latency for memory accesses (increasing overall performance further).

 

[ethomaz]

I'm not really commenting on the specific Intel product line here, but rather the more general approach.

Consuming less power while providing similar or even better performance (as in the case of the Apple M1) is also beneficial at every level and in every product segment.

At the top end, as is discussed in the thread here, using significantly lower power ARM cores for the 8 E-cores while providing similar performance, would not just enable a lower power consumption in the lowest power profile, but also more importantly provide more overall power availability for the larger cores in it's highest performance mode allowing those cores to be possibly clocked higher or even allowing them to include more SRAM cache on-die to decrease further latency for memory accesses (increasing overall performance further).

I’m not sure what are you trying to say because Apple M1 is very weak in performance compared with desktop processor of years ago.

It is not even on the same level that a mobile Ryzen.

 

[TTOOLL]

It doesn’t matter if devs don’t use them well.

 

[RNG]

The smart thing to do would be to wait for benchmarks.

 

[Bitmap Frogs]

Well GF is not tied to AMD anymore.
In fact AMD have a contractual clause that force them to use x% of GF production… AMD is trying for years to get out of GF to the point that all high-end chips today are produced at TSMC.

Maybe that will be a easy way to AMD get out of GF.

Real men have fabs.

 

[TheThreadsThatBindUs]

I’m not sure what are you trying to say because Apple M1 is very weak in performance compared with desktop processor of years ago.

It is not even on the same level that a mobile Ryzen.

I think you're being a bit disingenuous now. In your immediately previous post, you recognised the importance of acknowledging a core's performance wrt its competitor's within the same product category.

But now, here you suddenly start making a meaningless comparison between the Apple M1 and other Intel cores in all other product categories.

Clearly, if you consider the product category the Apple M1 is competing in, it punches well above its weight, outclassing x86 cores in the same performance band when those cores are running x86 code natively while consuming less power.

You're telling me that with Intel's design chops they can't engineer an ARM core that performs at a similar performance level as the low power cores in the top-end product listed in the OP? I'm not arguing they should integrate Apple M1s. I'm arguing they should adopt a similar approach and include their own bespoke ARM core design at the same performance level as the existing product cores and these will be more power-efficient by reason of the inherent fact that ARM's RISC base uses fewer transistors for the same level of performance and thus consumes less power.

 

[Polelock]

Lol, freaking Intel. Every year get to buy a new board. So glad AMD is kicking ass.

 

[ethomaz]

I think you're being a bit disingenuous now. In your immediately previous post, you recognised the importance of acknowledging a core's performance wrt its competitor's within the same product category.

But now, here you suddenly start making a meaningless comparison between the Apple M1 and other Intel cores in all other product categories.

Clearly, if you consider the product category the Apple M1 is competing in, it punches well above its weight, outclassing x86 cores in the same performance band when those cores are running x86 code natively while consuming less power.

You're telling me that with Intel's design chops they can't engineer an ARM core that performs at a similar performance level as the low power cores in the top-end product listed in the OP? I'm not arguing they should integrate Apple M1s. I'm arguing they should adopt a similar approach and include their own bespoke ARM core design at the same performance level as the existing product cores and these will be more power-efficient by reason of the inherent fact that ARM's RISC base uses fewer transistors for the same level of performance and thus consumes less power.

Me? I never ever talked about M1 before somebody tried to put it in comparison with 12900X lol
Anybody that tries to use the M1 power consumption to say this top desktop processor has too high power consumption has not ideia of what is talking about.

 

[Cryio]

First new node in like a decade as well, should be a quantum leap in performance.

New node of what?

If you're talking about 10nm, this will be Intel's 4th product on 10 nm (Cannonlake, Ice Lake, Tiger Lake and Alder Lake).
If you're talking about uArch, this is 4th new uArch since Skylake in 2015. So ... ?
If we're talking PURELY nodes, this is still 10 nm, still the same node as the last previous 3 generation of products.
As for desktop ... sure, 10 nm finally after Skylake at 14nm since 2015. I somehow still have no faith in them to deliver, lol.

 

[J-Rod]

I have no faith in them delivering for a while. They delayed 3 generations of chips through a succession of spectacular manufacturing failures which is why their stock shed $40+ billion in value, why Apple is shifting its macs away from Intel to chips made by TSMC, and why NVIDIA and AMD(which both outsource to TSMC) are gaining in areas Intel used to dominate.

They were flirting with the idea of outsourcing to TSMC themselves at one point, but this new ceo committed to building new foundries and keeping flagship manufacturing in-house. They are laying the groundwork for turning it around, but it’s going to take time and a prayer.

 

[ethomaz]

I have no faith in them delivering for a while. They delayed 3 generations of chips through a succession of spectacular manufacturing failures which is why their stock shed $40+ billion in value, why Apple is shifting its macs away from Intel to chips made by TSMC, and why NVIDIA and AMD(which both outsource to TSMC) are gaining in areas Intel used to dominate.

They were flirting with the idea of outsourcing to TSMC themselves at one point, but this new ceo committed to building new foundries and keeping flagship manufacturing in-house. They are laying the groundwork for turning it around, but it’s going to take time and a prayer.

Well they are trying to buy Global Foundries.

 

[Reallink]

New node of what?

If you're talking about 10nm, this will be Intel's 4th product on 10 nm (Cannonlake, Ice Lake, Tiger Lake and Alder Lake).
If you're talking about uArch, this is 4th new uArch since Skylake in 2015. So ... ?
If we're talking PURELY nodes, this is still 10 nm, still the same node as the last previous 3 generation of products.
As for desktop ... sure, 10 nm finally after Skylake at 14nm since 2015. I somehow still have no faith in them to deliver, lol.

I'm aware, I was referring to desktop parts.

 

[Black_Stride]

with 10nm vs 5nm they simply can`t beat AMD, at least not as long as performance per watt is a factor at all. If they`d already use TSMC´s 3nm capacities for it then that would be an entirely different thing, but the way things are with Intel being years behind in manufacturing processes i´m not holding my breath for any of their products.

You dont know what you are talking about when you say 10nm vs 5nm.
perf per watt? As long as you can cool your 12900K if its faster its the better chip to get.

I'm always amused by how many people automatically assume Intel will regain the performance crown. It's a completely new architecture that isn't remotely close to being as mature as 14nm. It will take a few generations of optimization. Intel fanboys automatically think that 10nm will automatically have all the refinements that 14nm had.

In fact, it was speculated that one of the reasons it has taken so long to get 10nm on desktops was that they could not hit the performance that matured 14nm CPUS were hitting.

Time will tell of course.

Im amused by people assuming Intel is like 1000 years behind on IPC.
Are their chips hotter and have a higher TDP? sure.
But the leaked benches already show that the 12900K will be on par or faster than the 5950X now im assuming Intel wont price themselves out of contention which would lead me to believe the 12900K will be the better buy versus the 5900X.
Heck if the 12700K has the same P-cores it should be a better buy than the 5900X.

But as you said time will tell.

 

[Haggard]

You dont know what you are talking about when you say 10nm vs 5nm.
perf per watt? As long as you can cool your 12900K if its faster its the better chip to get.

Looks more like you've been living under a rock for some time. Intel is years behind TSMC by now....
Also, kiddie gamers who don't mind a second heater in their room are just a tiny fraction of the market and efficiency is a major factor in the rest of it.

 

[Skifi28]

I could be wrong, but as others have said, I don't see the point of a phone design on a desktop CPU. Seems more like wasting silicon on something needless. Even back then my 2500k was running at something like 1200mhz when not doing anyting demanding to save energy, is it really necessary to have separate cores for something like that?

 

[SmokSmog]

rip AMD

8 big cores will score around 8250points in cinebench r20 at 5ghz all core, while Zen3 8core 5800x scores 6250 points at 4,8ghz.

Looks like Alder Lake has 27% better IPC than Zen3 in CB r20. 12900k with 8 big and 8 smaller cores will be faster than 16 core Zen3

 

[Armorian]

If this IPC is true we have a new performance king until RL and Zen 4.

But the fun thing is that only W11 will work with CPU correctly so sales will be limited by number of new OS users? It will work in W10 obviously but system will fuck up core/thread affinity.

 

[TheThreadsThatBindUs]

Me? I never ever talked about M1 before somebody tried to put it in comparison with 12900X lol
Anybody that tries to use the M1 power consumption to say this top desktop processor has too high power consumption has not ideia of what is talking about.

Now you're intentionally misconstruing my arguments. That's incredibly disingenuous.

Nowhere did I compare the Apple M1 with the 12900X and in my immediate post, I clarified exactly what my argument is. You've simply chosen to ignore it.

The mention of the Apple M1 was an attempt to explain how Apple proved it possible to engineer an ARM chip with considerably higher performance than equivalent intel cores. I don't consider that to be a difficult concept to grasp, ethomaz.

And I think you do understand the point. You're just choosing to argue in bad faith for the sake of "winning".

 

[SantaC]

rip AMD

8 big cores will score around 8250points in cinebench r20 at 5ghz all core, while Zen3 8core 5800x scores 6250 points at 4,8ghz.

Looks like Alder Lake has 27% better IPC than Zen3 in CB r20. 12900k with 8 big and 8 smaller cores will be faster than 16 core Zen3

Nobody is using R20 anymore. R23 is the current benchmarker.