Alder Lake-S leaks

[LePhuronn]

rip Zen 3

Care to actually explain what you're posting, or are you just regurgitating some random post doing its round on the internet? Because I really don't think a single thread at 5.1GHz being a larger number in a single synthetic benchmark than a Zen 3 CPU running at stock even remotely qualifies as "RIP".

Oh and by the way, you know this rumoured 25% performance boost coming from the 3D stacked cache in Zen 3? That takes the 5900X and 5950X to 810 points. Whoops! RIP Alder Lake...

So essentially no, there is no RIP Zen 3. Intel will have done an incredible job to close the monumental deficit they have to AMD in a single generation (2017 all over again, anybody?), bringing some proper competition back to the CPU space.

And besides, Alder Lake is SUPPOSED to kill Zen 3, otherwise what's the point? Just like Zen 4 will kill Alder Lake and mostly likely Raptor Lake. This is what true progression is about.

 

[SmokSmog]

Care to actually explain what you're posting, or are you just regurgitating some random post doing its round on the internet? Because I really don't think a single thread at 5.1GHz being a larger number in a single synthetic benchmark than a Zen 3 CPU running at stock even remotely qualifies as "RIP".

Oh and by the way, you know this rumoured 25% performance boost coming from the 3D stacked cache in Zen 3? That takes the 5900X and 5950X to 810 points. Whoops! RIP Alder Lake...

So essentially no, there is no RIP Zen 3. Intel will have done an incredible job to close the monumental deficit they have to AMD in a single generation (2017 all over again, anybody?), bringing some proper competition back to the CPU space.

And besides, Alder Lake is SUPPOSED to kill Zen 3, otherwise what's the point? Just like Zen 4 will kill Alder Lake and mostly likely Raptor Lake. This is what true progression is about.

25% gain for Zen3D ? Where did you get it from ? AMD shown 15% on average in 5 games, in games which are memory sensitive. You will not get more than 5% in CPU-Z from v-cache.

Alder Lake GldenCove 6-Way decoder x86, ROB 512, 5xALU, 5xAGU

Rocket Lake SunnyCove 4-Way decoder x86, ROB 352, 4xALU, 4xAGU

Zen3 4-Way decoder x86, ROB 256, 4xALU, 3xAGU

 

[Cyber-Mav]

Hope new cpus will be available in large quantities. Otherwise it be just reviewers who get to play with them.

 

[wookiee87]

rip Zen 3

love to know what power the new intel chip pulls, plus what type of cooling was used?

 

[humbug]

rip Zen 3

You look like you waited a very long time to say that.

You're going to have to keep waiting, Zen 3D around the same time and Zen 4 not long after that.

 

[cliffy]

25% gain for Zen3D ? Where did you get it from ? AMD shown 15% on average in 5 games, in games which are memory sensitive. You will not get more than 5% in CPU-Z from v-cache.

Alder Lake GldenCove 6-Way decoder x86, ROB 512, 5xALU, 5xAGU

Rocket Lake SunnyCove 4-Way decoder x86, ROB 352, 4xALU, 4xAGU

Zen3 4-Way decoder x86, ROB 256, 4xALU, 3xAGU

You obviously like your Intel. I wonder what the actual IPC gain is though as that Intel chip was running at 5.1Ghz my 5950X gets 674 running at 4.8Ghz turbo boosting itself. Would also like to see the power draw of the Intel chip and heat. As you said 3D Cache gains on average 15% but you have forgotten to factor in also the probable speed bump from the refined process and also forgotten it wont be that long till Zen 4. You are basically getting excited about an as yet unreleased CPU with no proven benchmark results, thermal or power draw beating a 1 year old AMD chip. I should also add my system is running memory at 3200 cl16 and i suspect the Intel is running higher memory speeds

 

[LePhuronn]

25% gain for Zen3D ? Where did you get it from ? AMD shown 15% on average in 5 games, in games which are memory sensitive. You will not get more than 5% in CPU-Z from v-cache.

Alder Lake GldenCove 6-Way decoder x86, ROB 512, 5xALU, 5xAGU

Rocket Lake SunnyCove 4-Way decoder x86, ROB 352, 4xALU, 4xAGU

Zen3 4-Way decoder x86, ROB 256, 4xALU, 3xAGU

Oh I see, you're taking a very specific synthetic benchmark that has little bearing on the real world and claiming it as Zen 3's death knell. Good job kid, you win today's internet bullcrap cookie.

That 25% performance boost is a top-end figure being rumoured for operations that benefit from lots of cache. Yes, AMD showed 15% in games, yes there will be some things that only see a 5% boost, but there will also be significant boosts as well.

RIP Zen 3 lol, I suggest you go find Dave and 4K8K and go have a little chuckle in the corner if you're going to post utter drivel like you just did.

 

[Grim5]

Lol @ looking at a cpu-z, geekbench, cinebench result and thinking that qualifies for a rip

 

[Grim5]

love to know what power the new intel chip pulls, plus what type of cooling was used?

probably 200w

the power usage will be super close to what 10th and 11th gen has. All the savings in power from a smaller node is used to stack little cores on - overall power draw will be so so similiar to 14nm chips.

And lol @ cpuz not showing any info for the little cores

 

[Zarax]

I prefer to wait until the real world testing is done. It’s this simple, how much better is it in gaming than the 10900k, paired with a 3090 at say 3440x1440. It’s not all about the CPU, DDR5 is also a factor.

I think someone on here is getting a little carried away stating “RIP Zen3”. This better be faster than Zen3 or it would be embarrassing for Intel. But again the DDR5 is also a factor, which AMD won’t have until 2022. Let’s wait before posting rubbish, with very little info to back it up.

Pretty much this. PC gaming is not supposed to work like football. Pick your budget, decide what features you want and what you're ready to take compromises on.
One year brand A will offer the best CPU and brand B the best GPU, next year things might reverse and sooner or later another player will come.

I personally applaud AMD for proving that chiplets are a good idea to improve yelds and performance and I'm glad Intel is willing to try something new to regain the performance crown.
At the same time, I'm closely watching VIA/Zhaoxin as a potential player in the next 5 years, hoping Intel can make a decent GPU for a good price and would pay any company to make a decent voxel accelerator if that could help that kind of game engines come back in higher definition.

I've been an Intel user since 1988 and Ati/AMD GPU user in the last decade, however if I had to buy today I would probably go for a 5900x/3060 combo as that would give me the best bang for the buck on my budget.

 

[humbug]

probably 200w

the power usage will be super close to what 10th and 11th gen has. All the savings in power from a smaller node is used to stack little cores on - overall power draw will be so so similiar to 14nm chips.

And lol @ cpuz not showing any info for the little cores

Tested in Aida64 the 12900K was pulling 250 Watts continuous power, 400 Watts peak.

This CPU is on Intel's new "Intel 7" process node, it seems they still haven't figured out how to make a power efficient X86 architecture.

 

[jigger]

Tested in Aida64 the 12900K was pulling 250 Watts continuous power, 400 Watts peak.

This CPU is on Intel's new "Intel 7" process node, it seems they still haven't figured out how to make a power efficient X86 architecture.

Intel are between a rock and a hard place. The primary driver will be power efficiency, but at this end of the market Intel also need wins in charts. The strategy is likely to sell two chips in one and have one sat idle 90% of the time.

What Intel need is 16 big cores that can be the Alpha and Omega.

 

[pcuser]

Getting a 5900X which I know will last me a long time made me lose my excitement. Will wait for 50% uplift in per core performance at the very least before I want to upgrade again.

That could be a long wait.

 

[cliffy]

Tested in Aida64 the 12900K was pulling 250 Watts continuous power, 400 Watts peak.

This CPU is on Intel's new "Intel 7" process node, it seems they still haven't figured out how to make a power efficient X86 architecture.

Really? That's insane. Still trying to figure out why Intel are calling it "Intel 7" process node, i am guessing its to hide the fact they are not on 7nm yet

 

[jigger]

That could be a long wait.

50% gain per core, over 12 cores and across the board… Yeah might take some time.

 

[jigger]

Really? That's insane. Still trying to figure out why Intel are calling it "Intel 7" process node, i am guessing its to hide the fact they are not on 7nm yet

It’s not even 10nm in the true sense. But neither is TSMC 7nm. Intel are adding market lies to marketing BS though.

Nvidia will probably ask it’s legal dept to take look at that when Intel enter the graphics card market. Ponte Vecchio in particular will potentially hurt Nvidia badly.

 

[humbug]

Intel are between a rock and a hard place. The primary driver will be power efficiency, but at this end of the market Intel also need wins in charts. The strategy is likely to sell two chips in one and have one sat idle 90% of the time.

What Intel need is 16 big cores that can be the Alpha and Omega.

Its not always about performance, the trouble is Intel are having to run them at ridiculous power level's to keep up with AMD, like AMD used to have to do with Bulldozer.

While Intel has a monopoly on software compatibility in a lot of aspects in Data centre, making those peoples only option Intel i think if people had a choice they wouldn't be choosing Intel, well, probably the old stuck in their ways duffers might but in this climate (no pun intended) people just aren't running hundreds of watts per server more power than AMD only to match their performance.

Intel per core performance was never really behind AMD's, with Zen more often than not still ahead if only by a bit, the exception being Zen 3 which is an exceptional architecture, high clocking, very high IPC and very power efficient all in one.

As time goes on and less and less people are dependant on Intel's decades of *IfAMDdetected =JunkCodePathOrLockOut/ on software AMD can only grow at Intel's expense.

We evaluated Intel’s latest generation of “Ice Lake” Xeon processors. Although Intel’s chips were able to compete with AMD in terms of raw performance, the power consumption was several hundred watts higher per server - that’s enormous. This meant that Intel’s Performance per Watt was unattractive.

https://blog.cloudflare.com/the-epy...n-in-cloudflares-11th-generation-edge-server/

AMD learned a hard lesson from Bulldozer.

 

[Perfect_Chaos]

That could be a long wait.

+15% for Zen 3D (games and some other tasks only?). Zen 4 rumoured about 20% IPC increase so more perf with higher clock speed.

Not difficult to believe when Intel's chips will supposedly be 100% quicker than current stuff alone just <4 years from now.

 

[jigger]

Its not always about performance, the trouble is Intel are having to run them at ridiculous power level's to keep up with AMD, like AMD used to have to do with Bulldozer.

While Intel has a monopoly on software compatibility in a lot of aspects in Data centre, making those people only option Intel i think if people had a choice they wouldn't be choosing Intel, well, probably the old stuck in their ways duffers might but in this climate (no pun intended) people just aren't running hundreds of watts per server more power than AMD only to match that their performance.

Intel per core performance was never really behind AMD's, with Zen more often than not still ahead if only by a bit, the exception being Zen 3 which is an exceptional architecture, high clocking, very high IPC and very power efficient all in one.

As time goes on and less and less people are dependant on Intel's decades of *IfAMDdetected =JunkCodePathOrLockOut/ on software AMD can only grow at Intel's expense.



https://blog.cloudflare.com/the-epy...n-in-cloudflares-11th-generation-edge-server/

I completely agree. Personally I would like to see what Intel could offer with its low powered cores and push some low price/low power parts in the 8-16 core range. That seems about the optimum desktop core count currently.

As you point out performance isn’t Intel's biggest problem, the issue is power use and core count scaling. From what I’ve seen Intel haven’t addressed those issues with Alderlake and everything hinges on Nova Lake.

Intel have lost the HEDT, Workstation and DC market for the foreseeable future now. The guys making the financial decisions have been burnt badly by Intel and once you compare AMD and Intel solutions in a spreadsheet it’s clear. Even if Intel hit all the right notes with Nova Lake in the DC market they still have 5-10 years of legacy investment to overcome and that is plenty of time for AMD to push its R&D.

Where Intel have a potential advantage is in it’s Ponte Vecchio GPU acceleration and that seems to be Intel’s strategy to making its DC recovery. Push into that market and corner it first, then leverage its CPU platforms for some extra value and offer a single vendor solution.

 

[jigger]

+15% for Zen 3D (games and some other tasks only?). Zen 4 rumoured about 20% IPC increase so more perf with higher clock speed.

Not difficult to believe when Intel's chips will supposedly be 100% quicker than current stuff alone just <4 years from now.

It’s very difficult to see how Intel achieve that.