Discussion Intel current and future Lakes & Rapids thread

[lobz]

Cant really say i'm impressed with either Alderlake or Intel "7nm" considering this took PL2 255w to achieve..

This is my everyday 1 year old 5950x

165w @ 1.1vcore get -> 4550/4400mhz = 30008 points
View attachment 50539

181w @ 1.138 vcore get -> 4600/4500mhz = 30538 points
View attachment 50540

Not sure how wise it is to start comparing these scores without taking into account, how much fine tuning can mean. You also can't know the 12900K's actual power consumption during this test, let alone every other significant detail...

 

[JoeRambo]

What i hate about this CB score bs is the fact that this type of rendering workload has nothing to do with how performance will be in real world.

As long as rendering "slice" working set fits L2 cache, CB will render all day long even if said CPU is not connected directly to memory as proven already by Threadripper 2990WX style abominations from AMD.

You can see that effect in

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

vs

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

If one was to judge how good CPU 32C is by Cinebench and then get burned by 20% slower compile than 16C chip due to lack of sensible memory subsystem?


So that is what we have here today, upper limit of core potential is shown CB23, where sun always shines, workloads fit L2.
Real world will depend on variuos things that can throw a wrench into whole scheme, like OS scheduler, bad inter core latency due to heterogenous nature of the chip, some nasty pitfalls due to DDR5 being immature?
Both Intel's own graphs and GB5 results show regressions. GB5 shows them in workloads that need tight memory.

Let's wait for final results on stock setup with DDR4, DDR5. And then let's wait for finetuned results. Maybe the chip has linear scaling in games going from stock DDR4 3200 to DDR4 4000 for example?

 

[RTX2080]

The memory section of CPU-Z says “quad” channel. Interesting they are framing it that way.

It's dual channel. That's just a wrong recognition by CPUZ

EDIT: just realize this screenshot was posted by a newcomer REHWK on twitter(screenshot was not made by him) and have some other strange posts that hard to understand, which make this leak looks doubtful though......

 

[lobz]

@JoeRambo if it was 'real world enough' for Intel when they kinda sponsored Maxon, then in retaliation it's been 'real world enough' for AMD in the past 5 years to prove the world wrong, let's try to acknowledge it as 'real world enough' at least for a couple of generations more, so we can be fair. It's also not like GeekBench, which uses REALLY questionable weighting, making it the laughing stock of 'objective' benchmarks, right in the same category as UserBenchmark for me.

At least with CB you know 100% perfectly what and how it measures, and it's a perfect indicator for that kind of workload, quite consistently between generations and even between architectures.

 

[ondma]

Cant really say i'm impressed with either Alderlake or Intel "7nm" IF it took PL2 255w to achieve those 30.5k points in Cinebench r23..

This is my everyday 1 year old 5950x

165w @ 1.1vcore get -> 4550/4400mhz = 30008 points
View attachment 50539

181w @ 1.138 vcore get -> 4600/4500mhz = 30538 points
View attachment 50540

If i let my 5950x use above 250w freely i can score up to and above 32k in cinebench r23

*edit*
Bone stock memory + 5950x load-optimized-defaults in bios:

125w = 26182 points
View attachment 50542

Price if the 5950x es 700.00 and up. If the 12900k is priced like the 5900x (which I consider its real competition) it could be a good alternative. If they price it like the 5950x, then not so much (unless it is clearly faster overall, which I don't expect).

 

[SAAA]

Cant really say i'm impressed with either Alderlake or Intel "7nm" IF it took PL2 255w to achieve those 30.5k points in Cinebench r23...

CPUz says 10 nm, speaking the truth!

That aside... this is happening with just 8 big and 8 small cores: consider even using 255 W vs a 11900K at 255 W it's doing great work and efficiency shows. They literally 2x multi-thread in one generation.

 

[Joe NYC]

CPUz says 10 nm, speaking the truth!

That aside... this is happening with just 8 big and 8 small cores: consider even using 255 W vs a 11900K at 255 W it's doing great work and efficiency shows. They literally 2x multi-thread in one generation.

Fixed it:

https://twitter.com/x/status/1420069663879409669

 

[DrMrLordX]

By waving a magic wand, you reckon?

Ask a stupid question, get a stupid answer:

YES

In all seriousness, it's unlikely that they've been working on this update since 2019. They've had plenty of time.

 

[mikk]

Cant really say i'm impressed with either Alderlake or Intel "7nm" IF it took PL2 255w to achieve those 30.5k points in Cinebench r23..

This is my everyday 1 year old 5950x

It's impressive considering there are only 8 big cores and it has 8 threads less than the 5950x. This is unexpected MT performance if it's true.

 

[Edgemeal]

It's impressive considering there are only 8 big cores and it has 8 threads less than the 5950x. This is unexpected MT performance if it's true.

But won't ADL actually be competing with Zen3+ (or whatever its going to be called) ?

 

[thigobr]

Not for a short while. But even then Zen3+ won't offer big performance improvements on every application, just for some things where the L3 size pose big limits to performance (e.g. games). It's nothing like Zen2 -> Zen3 improvements.

If this is true it's indeed a good showing for Alderlake...

 

[eek2121]

It's impressive considering there are only 8 big cores and it has 8 threads less than the 5950x. This is unexpected MT performance if it's true.

The people still swearing ADL-S will be horrible at this point are anti-Intel. Don't bother. You can't have an actual objective conversation with them. For me personally, I suspect power usage will be well above and beyond my AMD chip (a 5950X), so unless it absolutely destroys my machine in every way, shape, and form, I'm not going to bother.

But won't ADL actually be competing with Zen3+ (or whatever its going to be called) ?

Yes, but I wouldn't expect Zen3D to perform miracles. The rumor mill indicates slightly lower clocks, but better performance thanks to the additional cache. Neither vendor will have a clear winner from what I personally gather. Even if AMD manages to meet or exceed clocks, You are looking at less than a 20% overall boost to Zen 3. AMD will likely retake the crown for a few key benchmarks, and lose others where cache is not as important. A lot of people claim that CBr23 doesn't care about memory bandwidth...we'll find out for sure once Zen3D comes out.

 

[Zucker2k]

Yes, but I wouldn't expect Zen3D to perform miracles. The rumor mill indicates slightly lower clocks, but better performance thanks to the additional cache. Neither vendor will have a clear winner from what I personally gather. Even if AMD manages to meet or exceed clocks, You are looking at less than a 20% overall boost to Zen 3. AMD will likely retake the crown for a few key benchmarks, and lose others where cache is not as important. A lot of people claim that CBr23 doesn't care about memory bandwidth...we'll find out for sure once Zen3D comes out.

The key thing here is that with Zen3D rumored to be a bit more expensive piece of tech to manufacture, how much more is AMD going to tag onto the Zen 3 tax, and are enthusiasts going to respond favorably to it? As it is, with the 12600kf rumored to be 5800x killer and listed at a paltry $261 price, things are not looking good for Zen3D pricing.

ADL-S and next line of releases are going to be a real problem for AMD. ADL-S looking this strong this early, and with RPL with 8 raptor cove cores and 16 e-cores, and all of a sudden, it's AMD that's under pressure in the "moar cores" war.

 

[lobz]

The key thing here is that with Zen3D rumored to be a bit more expensive piece of tech to manufacture, how much more is AMD going to tag onto the Zen 3 tax, and are enthusiasts going to respond favorably to it? As it is, with the 12600kf rumored to be 5800x killer and listed at a paltry $261 price, things are not looking good for Zen3D pricing.

ADL-S and next line of releases are going to be a real problem for AMD. ADL-S looking this strong this early, and with RPL with 8 raptor cove cores and 16 e-cores, and all of a sudden, it's AMD that's under pressure in the "moar cores" war.

If they have any sense, they will pull a Zen+ and price the parts really competitively, since 7nm production is really-really stable now.
How much sense they have left is hard to say though. 😔

 

[coercitiv]

The key thing here is that with Zen3D rumored to be a bit more expensive piece of tech to manufacture, how much more is AMD going to tag onto the Zen 3 tax, and are enthusiasts going to respond favorably to it? As it is, with the 12600kf rumored to be 5800x killer and listed at a paltry $261 price, things are not looking good for Zen3D pricing.

ADL-S and next line of releases are going to be a real problem for AMD. ADL-S looking this strong this early, and with RPL with 8 raptor cove cores and 16 e-cores, and all of a sudden, it's AMD that's under pressure in the "moar cores" war.

My understanding is that Zen3 and even Zen3D should still be cheaper to manufacture than a large monolithic die such as ADL-S when looking at the flagship SKUs, alas production volume vs. demand will complicate the equation quite a bit. We'll discuss this in greater detail as ADL-S launches, for now I'd rather wait and refrain from stealing any spotlight from ADL as this is it's rightful thread. I'll say this for now though: I don't expect AMD to follow up in the "moar cores" escalation. Just like with vanilla Zen, they will play their strengths and change their marketing narrative. I expect P cores to be under assault from the competition, not E cores.

Alder Lake will definitely be a refreshing and competitive product for the end of this year, although you probably know by now that in this stage I value P cores much-much more than E cores when it comes to consumer desktop.

 

[lobz]

My understanding is that Zen3 and even Zen3D should still be cheaper to manufacture than a large monolithic die such as ADL-S when looking at the flagship SKUs, alas production volume vs. demand will complicate the equation quite a bit. We'll discuss this in greater detail as ADL-S launches, for now I'd rather wait and refrain from stealing any spotlight from ADL as this is it's rightful thread. I'll say this for now though: I don't expect AMD to follow up in the "moar cores" escalation. Just like with vanilla Zen, they will play their strengths and change their marketing narrative. I expect P cores to be under assault from the competition, not E cores.

Alder Lake will definitely be a refreshing and competitive product for the end of this year, although you probably know by now that in this stage I value P cores much-much more than E cores when it comes to consumer desktop.

the
Still, don't forget the infamous warchest. Took some years for them to realize, but Intel knows well, that they can't get away with everything anymore. They will swallow what they must to price ADL low enough to be desirable, but still high enough not to induce utter panic among investors. After all the magical shuffling, finances will be just fine at the end. I'm sure AMD knows and sees this a hundred times better and clearer than I (or any of us) do, so I must assume they have something in the works or they just have full confidence in their mid-term plan.

 

[coercitiv]

Still, don't forget the infamous warchest.

As I said, we can get into more details once we meet ADL-S properly, for now I merely wanted to point out that final pricing is a moving target for both vendors, not just one. Personally I doubt we'll see prices that seriously dent margins from either company, mainly due to the current state of affairs in the global market (crazy demand, global supply issues etc). Innovation will be far more exciting than pricing, or at least that's how I see 2021-2022.

I would definitely welcome agressive pricing from Intel, but I'm not holding my breath after seeing how they approached RKL-S.

 

[JoeRambo]

I would definitely welcome agressive pricing from Intel, but I'm not holding my breath after seeing how they approached RKL-S.

I'd say if leaks are correct and 12600KF is really ~$250, that will force AMD to stop milking customers with "proper" 8C and come up with either better product in the form of Z3D or to adjust pricing accordingly. IF 8+8 match 16 AMD cores in MT then 6+4 will definately beat 8C in MT while also providing better ST performance as well. $400 5800X will need adjustments and AMD 6C will be hurt even more so.

Good that AMD still has ton of lots for SKUs to release at whatever prices they want. What about 5700X with 100mhz lower boost clock on paper for 330$ ?

 

[lobz]

(...) final pricing is a moving target for both vendors, not just one.

That is the most accurate way of putting it

 

[uzzi38]

So, from the looks of it GNR doesn't seem to employ the use of Foveros at all

Dylan Patel on Twitter: "According to the presentation I am listening to, Intel will not be using their 3D stacking technology (Foveros) on Granite Rapids (GNR), targeting launch in 2023 AMD with TSMC "start production on our highest end products with 3d chiplets by the end of this year" $INTC $TSM $AMD https://t.co/H12Mgij85M" / Twitter

EMIB only, same as SPR. Meteor Lake was confirmed earlier to be using regular Foveros (36um). Looks like Foveros Direct which makes use of hybrid bonding like SoIC is as of the end of this year isn't exactly close on the roadmap for any high performance products, which is a shame.

 

[coercitiv]

I'd say if leaks are correct and 12600KF is really ~$250, that will force AMD to stop milking customers with "proper" 8C

Et tu, Brute? If you think the 12600KF could offer strong value for the money, then wait till you see the rest of the i5 SKUs. The 12600K will need cut-down 8+8 silicon while the rest of the i5 range will be able to use the far more price effective 6+0 silicon.

I cannot help but smile watching this irony unfold: for more than a decade Intel cursed the small cores to be cheap alternatives to the real "Intel Inside", now with Alder Lake they will probably sell the Atom for more $$ / mm2 than their crown jewel Coves. True poetic justice, if such a thing can be envisioned for CPUs.

 

[Mopetar]

The key thing here is that with Zen3D rumored to be a bit more expensive piece of tech to manufacture, how much more is AMD going to tag onto the Zen 3 tax, and are enthusiasts going to respond favorably to it? As it is, with the 12600kf rumored to be 5800x killer and listed at a paltry $261 price, things are not looking good for Zen3D pricing.

ADL-S and next line of releases are going to be a real problem for AMD. ADL-S looking this strong this early, and with RPL with 8 raptor cove cores and 16 e-cores, and all of a sudden, it's AMD that's under pressure in the "moar cores" war.

Really depends on how Intel ends up doing. If AMD can beat them again, expect higher prices. If it's more competitive then expect the same for their prices.

It's not a coincidence that Zen, Zen+, and Zen 2 which were still behind Intel in many areas where more cores weren't helpful were priced far more reasonably and Zen 3 which basically ruled the roost saw price bumps.

 

[jpiniero]

Intel historically has not changed prices that much at a given brand level. Now what you get at brand level is another story.

 

[dullard]

now with Alder Lake they will probably sell the Atom for more $$ / mm2 than their crown jewel Coves.

Hard to tell right now based on preliminary pricing, but a crude price estimate is $32/core (regardless of type of core) + ~$50 for being the top of the category (for example the top i9, the top i7, the top i5, top i5 non K, etc). If that is what it ends up being, then yes, they are selling Atom at 4x the price/area.

That said, Intel doesn't really charge based on area. The integrated graphics on the last few chips are a massive amount of area and the F chips don't cost much less. Meaning, they aren't charging much at all for that functional area.

 

[Hulk]

My understanding is that Zen3 and even Zen3D should still be cheaper to manufacture than a large monolithic die such as ADL-S when looking at the flagship SKUs, alas production volume vs. demand will complicate the equation quite a bit.

I don't have much insight into this and would love to know more? At the top of the stack I assume the two AMD chiplets are smaller in total area than the 8+8 ADL die? Also when there is a defect in ADL you could lose the whole die, which is less of a loss than one chiplet. But then it seems as though there might be more assembly time/cost in the chiplet approach? I know, I'm jumping the gun here but I'm curious about the economics of these approaches specifically in the case of 7nm 5950X vs 10nm 12900K production?