Discussion Intel current and future Lakes & Rapids thread

[eek2121]

No, that is absolutely not how Intel used to do it. Anyway, since they can't legally and so openly do what they used to do anymore, they have learned and got really good at going after the decision makers in a company. While still not legal in every aspect, it's like illegal prostitution - it takes two and if both individual parties are happy (who cares about the family or the state), you rarely get into official trouble.

They absolutely did do this. It is nothing more than a volume purchasing agreement. The rebate system as I described is legal. Intel is NOT the only company to do it. Many other companies do this as well as an alternative to volume discounts. Volume discounts require a large purchase up front and most companies shy away from buying a ton of inventory.

The other crap about incentivizing purchases of only Intel products is not legal.

 

[Hulk]

If you sign an agreement to sell X amount of units within the next Y amount of time, you'll receive a rebate. At least that is how Intel used to do it. The rebate depends on the number of processors you sell, and the actual rebate amounts aren't privy to the public, but the discount can be significant. In the past a $350 CPU the sold moderately would cost an OEM such as Dell almost half that (around $190-$210), while in their (the OEM's) BOM they would charge you the retail price. I imagine an OEM that sells several hundred thousand 12900ks will be able to lower their ASP to around $300 or so. I do NOT expect the 12900k to cost more than $580 (and less with discount resellers like Amazon, the 11900k is $488 right now for example, but the MSRP is actually $550), but we will see. Note all after rebate are ASP (average selling price), so when you multiply that by X amount of units, it adds up pretty quickly. Note that some companies will offer you discounts for buying a large number of units, but Intel, to my knowledge (at least when I worked with them) they don't do that. Perhaps someone here works with Intel currently and can provide insight as to whether they do this.

Please note that my knowledge is pretty out of date. I got out of that game a few decades ago.

Thanks for the reply. I hope you are right about the 12900K pricing. I'm just hoping they don't start to believe their own hype and try pricing the 12900K at $999!

 

[lobz]

They absolutely did do this. It is nothing more than a volume purchasing agreement. The rebate system as I described is legal. Intel is NOT the only company to do it. Many other companies do this as well as an alternative to volume discounts. Volume discounts require a large purchase up front and most companies shy away from buying a ton of inventory.

The other crap about incentivizing purchases of only Intel products is not legal.

You misunderstood me. What you described, every company does. This was not what allowed Intel to win over customers, and certainly not how they convinced them not to purchase and offer anything from competitors.

 

[Asterox]

Thanks for the reply. I hope you are right about the 12900K pricing. I'm just hoping they don't start to believe their own hype and try pricing the 12900K at $999!

Not this time or today, no 1000$ for power hungry CPU on already expensive platform.

Will you be able(huh lol) to install 12900K on 70-80$ Intel motherboard? In short, you cant kill Gigabyte motherboard it has power limit around 130W.

 

[moinmoin]

Wish they shared the sub scores because the overall does not mean much on its own.

At the end of the article they list three result pages at Geekbench, 1, 2, 3. Isn't what you want among that?

 

[Timorous]

At the end of the article they list three result pages at Geekbench, 1, 2, 3. Isn't what you want among that?

Yes it is, completely missed that. Thank you.

So the INT and FP scores between the 3 results are pretty similar for the SC.

 

[Mopetar]

And the point I was trying to make was you were both correct since you were talking about (slightly) different things.

I understand that, but it's silly to make claims about Intel gaslighting people over the meaning of efficiency while making what I consider a disingenuous argument for the reasons I previously stated.

It's a bit like taking a new Tesla which might have a range of 400 miles if you drive 60 mph and getting it up to top speed where the range might only be 100 mph and then make wild accusations against Tesla and their statements about performance and how an electric vehicle could have the same range as a vehicle with a combustion engine.

It's certainly valid to consider how those small cores perform at the clock speeds Intel ships the product with, but the statements being made were asinine.

 

[JoeRambo]

New 12700 results seem to be pulling away from my Skylake, even if memory still seems to have horrible latency.

Default string Default string vs Micro-Star International Co., Ltd. MS-7C71 - Geekbench Browser

One has to wonder about previuos vs these Crypto results -> either there was some problem with BIOS advertising VAES instructions or part of previuos test was run on small core. If latter it's bad news for Alder Lake on Windows 10 scheduler.

 

[Abwx]

New 12700 results seem to be pulling away from my Skylake, even if memory still seems to have horrible latency.

Default string Default string vs Micro-Star International Co., Ltd. MS-7C71 - Geekbench Browser

One has to wonder about previuos vs these Crypto results -> either there was some problem with BIOS advertising VAES instructions or part of previuos test was run on small core. If latter it's bad news for Alder Lake on Windows 10 scheduler.

In single thread Ram is not of big importance, so overall it s 28% faster in ST clock/clock than your CML.

 

[lobz]

In single thread Ram is not of big importance, so overall it s 28% faster in ST clock/clock than your CML.

But Rocket Lake covered much of that leap in GeekBench already, so these are still interesting results at most.

 

[Hulk]

Two points that might be worth tossing around while we wait for the next leak

I'm really optimistic about the ADL UP4, 2P + 8E CPU.

I'm currently running a Surface Laptop 2/Kaby Lake. 4E alone *should* nearly have the compute of my current rig, perhaps more if it can clock up higher. Then add in another 4E and 2P and things are looking pretty good.

Also I've been thinking about Intel's claim that Golden Cove will be the P core for the next decade. I'm thinking they mean that as software transitions more and more to MT and therefore ST performance becomes less important, Golden Cove (and future derivatives) might "see them through" the Big core era. Perhaps the future is lots of efficiency cores and Intel is predicting that will occur in the next 10 years?

 

[itsmydamnation]

I've been wondering how much of the increased concurrency of ddr5 can make up for trash timings, along with bigger ooo windows more prefetch. ADL benchmarked with DDR 4 vs 5 vs tight vs loose I would be very interested to see.

 

[eek2121]

But Rocket Lake covered much of that leap in GeekBench already, so these are still interesting results at most.

For Anandtech, the 11900k scored 1821/10281. The 12700 right now scores around 1763/11895.

The thing is, unless Intel changes pricing from previous gen, the 12700 is a $350 part while the 11900k is a $550 part. The 12700 likely has a 65W PL1, and the 12700 is going to consume significantly less power than than the 11900k.

Anandtech does not appear to have 11700 GB5 results, but a stock system score with stock RAM (JEDEC timings) appears to be around 1665/8900.

 

[Hulk]

For Anandtech, the 11900k scored 1821/10281. The 12700 right now scores around 1763/11895.

The thing is, unless Intel changes pricing from previous gen, the 12700 is a $350 part while the 11900k is a $550 part. The 12700 likely has a 65W PL1, and the 12700 is going to consume significantly less power than than the 11900k.

Anandtech does not appear to have 11700 GB5 results, but a stock system score with stock RAM (JEDEC timings) appears to be around 1665/8900.

I don't trust those leaked scores. Especially this early. Golden Cove, with all those architectural improvements scoring ST lower than 11900K? Pricing will ultimately be strongly affected by performance.

Performance is unknown and so it pricing. We'll see.

 

[coercitiv]

Ian Cutress on Twitter:

https://twitter.com/x/status/1431643300226818057

 

[JoeRambo]

In single thread Ram is not of big importance, so overall it s 28% faster in ST clock/clock than your CML.

I did a run of DDR4 2400CL20, to "simulate" DDR5 4800CL40. Of course it is not real deal, cause secondaries are still tight, but the impact on ST is still visible:

Micro-Star International Co., Ltd. MS-7C71 vs Micro-Star International Co., Ltd. MS-7C71 - Geekbench Browser

I'd expect at least 5% ST difference with bad memory for AlderLake.

Gotta reboot this slugfest of 2400CL20

 

[Joe NYC]

Ian Cutress on Twitter:

https://twitter.com/x/status/1431643300226818057

I would take the one on the right.

 

[moinmoin]

If we now post random tweet takes on topic, I'll take these two:

https://twitter.com/x/status/1428898427090440192

https://twitter.com/x/status/1428903793148768256

 

[Ajay]

Ian Cutress on Twitter:

https://twitter.com/x/status/1431643300226818057

Yes, on the right - 10P. No fancy scheduler needed. No AVX512 issues. No brainer to me.

 

[Det0x]

Intel Alder Lake support added to Extreme Tuning Utility (XTU), first stability test leaks - VideoCardz.com

Intel Alder Lake at 86°C and 248W average Lenovo China Gaming Desktop Product Planning Manager known as Wolfstame on Chinese media platform Weibo shares information on unreleased hardware on a frequent basis. This time around, he shared the power and heat stability test of the undisclosed Alder...

videocardz.com

Lenovo China Gaming Desktop Product Planning Manager known as Wolfstame on Chinese media platform Weibo shares information on unreleased hardware on a frequent basis. This time around, he shared the power and heat stability test of the undisclosed Alder Lake CPU from AIDA64.

According to the screenshot that he posted, a processor which is not named heats up to 93°C with an average of 86°C after 51 minutes of testing. The user is careful not to name the product as it obviously breaks the embargo. He notes that GeForce RTX 3090 is no longer the only thing in the system that consumes 350W of power. (to run full) This part is an Intel Core processor (as seen on a screenshot) with an unknown number of cores and configurations.

To achieve 248W average package power consumption, the CPU would have to run in PL2 (power limit 2) mode for an extended time (in other words it has to be overclocked beyond the maximum load time). Out of all leaked Alder Lake CPUs, only the 125W series is known to have a 241W PL2 limit, those are the K-series SKUs. In other words, Wolfstame is testing either Core i9-12900K, i7-12700K, or i5-12600K.

WolStame states that it's not only the NVIDIA GeForce RTX 3090 graphics card that would need 350W to run at its fullest potential & follows up with what seems to be early numbers of an Intel Alder Lake-S chip running at Lenovo's test labs.

That's why earlier leaks made sure to mention that watercooling was used for that leaked scores, but it was done without "overclocking"

https://twitter.com/x/status/1417527787695448069

Seems like these cpus will use up-to and above 250++ w if you have the cooling needed. (just to beat regular zen3(?)

 

[Hulk]

Yes, on the right - 10P. No fancy scheduler needed. No AVX512 issues. No brainer to me.

If true that Gracemont=Skylake then I'm not sure. 4 Golden Coves + 24 Gracemonts (Skylake) might run well on some well threaded apps. Can't wait for actual reviews.

 

[Ajay]

If true that Gracemont=Skylake then I'm not sure. 4 Golden Coves + 24 Gracemonts (Skylake) might run well on some well threaded apps. Can't wait for actual reviews.

Different strokes I guess.

 

[AMDK11]

Gekkbench 5.4.1 supports AVX512 as well as Rocketlake but has been disabled in Alderlake so Alderlake's results are lower.

 

[Accord99]

Seems like these cpus will use up-to and above 250++ w if you have the cooling needed. (just to beat regular zen3(?)

The 5950X will use 250+ watts too if you want to extract the last 10% of all-core performance.

 

[eek2121]

I don't trust those leaked scores. Especially this early. Golden Cove, with all those architectural improvements scoring ST lower than 11900K? Pricing will ultimately be strongly affected by performance.

Performance is unknown and so it pricing. We'll see.

You are basically saying “It barely beats the 5950X so it must be bad.” You are wrong. What we are about to see is an intriguing shift in computing where even the “lowly” 12600k (which will likely retail for under $300) has more than 8 cores.

AMD will have to step it up in order to compete. That is a good thing.