Question Raptor Lake - Official Thread

[Hulk]

Since we already have the first Raptor Lake leak I'm thinking it should have it's own thread.
What do we know so far?
From Anandtech's Intel Process Roadmap articles from July:

Built on Intel 7 with upgraded FinFET
10-15% PPW (performance-per-watt)
Last non-tiled consumer CPU as Meteor Lake will be tiled

I'm guessing this will be a minor update to ADL with just a few microarchitecture changes to the cores. The larger change will be the new process refinement allowing 8+16 at the top of the stack.

Will it work with current z690 motherboards? If yes then that could be a major selling point for people to move to ADL rather than wait.

 

[Hulk]

I've very rarely seen people truly happy with CPU upgrades of only one generation. The speed gains are often measurable, but often are so small that they can not easily be noticed outside of benchmarks. Plus the i9 line almost never makes good sense outside of bragging rights. The 14700K (or KF) seems to be the sweet spot for Raptor Lake refresh. Frequency difference between that and the 14900K is less than 2%--far too small to actually feel. The only time you'd notice the difference with the 14900K is when you really need all 16 E cores which is not common for most users. That is definitely not worth the 41% price premium on Amazon currently over the 14700K.

The reality of the situation is the one application that I use that I wish to be faster, namely Presonus Studio One (mixing down tracks), is not very multithreaded so 6 P's running fast is about the same as 8 P's running fast. I have my 13600K clocked at 5.4GHz on the P's anyway and when I'm mixing down tracks it's only drawing 88Watts. This application is actually faster during this process without HT on. I may turn off 1 core and see if it slows down much. It might only really be stressing 1 or 2 cores in which case I might be able to "find" my best cores and crank them up a bit more.

 

[igor_kavinski]

The reality of the situation is the one application that I use that I wish to be faster, namely Presonus Studio One (mixing down tracks), is not very multithreaded so 6 P's running fast is about the same as 8 P's running fast.

Are you sure that it's only CPU limited and not I/O bound? Easy way to test it is to create a pretty large RAM drive, put the files you work on there and benchmark a typical session. If the program settings have the option to specify a temp drive that it can use for temporary storage, you can point it to a folder on the RAM drive for that too. Even the fastest NVMe SSD is no match for a RAM drive.

 

[Hulk]

Are you sure that it's only CPU limited and not I/O bound? Easy way to test it is to create a pretty large RAM drive, put the files you work on there and benchmark a typical session. If the program settings have the option to specify a temp drive that it can use for temporary storage, you can point it to a folder on the RAM drive for that too. Even the fastest NVMe SSD is no match for a RAM drive.

No, it's compute limited. It takes 20 seconds to render out a 6MB mp3 file. The files it is mixing aren't that large either. We're talking audio files not video.

Also the render speed is directly proportional to the cpu clock, I've checked it. Good thought though!

 

[igor_kavinski]

It takes 20 seconds to render out a 6MB mp3 file.

TBH, I would explore different applications to see which one is the fastest for my workflow, although for 20 seconds, I might not even bother.

 

[Hulk]

TBH, I would explore different applications to see which one is the fastest for my workflow, although for 20 seconds, I might not even bother.

The thing about the 20 seconds is when mastering a project you need to lower a vocal a bit, do it, wait 20 seconds to update the mastering project... rinse, repeat, after 20 times it the wait gets old real fast. Even moving from 20 seconds to 16 is helpful to the workflow.

I hate waiting especially when I'm doing something creative as the wait breaks the creative process. If Studio One was "smart" it would break up the song into 50 slices and assigns processors to each slice, much like how we see CB R23 swarming during a run.

 

[igor_kavinski]

If Studio One was "smart" it would break up the song into 50 slices and assigns processors to each slice, much like how we see CB R23 swarming during a run.

That thought must have occurred in some creative's mind who just happened to also be a developer so there's a good chance that something better than Studio One is out there.

 

[Hulk]

That thought must have occurred in some creative's mind who just happened to also be a developer so there's a good chance that something better than Studio One is out there.

Like most software they all have their advantages and disadvantages. I've been using Studio One since inception and have used many others. I like it. No software is perfect unfortunately.

 

[Timur Born]

What the use of an AVX offset (-1) looks like on a 13900K when per core ratios don't match Turbo ratios.

1-core load SSE vs. AVX:


2-core load SSE vs. AVX:


4-core load SSE vs. AVX:


6-core load SSE vs. AVX:


8-core load SSE vs. AVX:

 

[igor_kavinski]

7.73 GHz all core LN2 record: https://www.tomshardware.com/pc-com...ke-refresh-cpus-and-plenty-of-liquid-nitrogen

 

[igor_kavinski]

R-Batches, 13900KSS and IMC regressions? – Intel Core 14th Gen Binning Results from almost 600 CPUs | igor´sLAB

Another CPU generation, another binning session along with an article for you all! We also had the pleasure of opening up almost 600 silicon loot boxes for the launch of Intel’s 14th generation and…

www.igorslab.de

Igor declares 14th gen IMC a sham.

 

[poke01]

R-Batches, 13900KSS and IMC regressions? – Intel Core 14th Gen Binning Results from almost 600 CPUs | igor´sLAB

Another CPU generation, another binning session along with an article for you all! We also had the pleasure of opening up almost 600 silicon loot boxes for the launch of Intel’s 14th generation and…

www.igorslab.de

Igor declares 14th gen IMC a sham.

14th gen as whole is useless. No performance uplifts on the flagship. 10nm node is saturated, time for a new node.

 

[igor_kavinski]

14th gen as whole is useless. No performance uplifts on the flagship. 10nm node is saturated, time for a new node.

They COULD have made some improvements, like more cache or better tuning of the IMC or heck, a 40 thread monster powered by more E-cores. But they chose not to, probably coz the decision was made at the last minute. Now they gonna try to survive the onslaught of Zen 5 with unattractive, boring desktop silicon, depending primarily on the stupidity of the common buyer, especially the corporate hardware acquisition "experts". Intel is probably the luckiest corporation in the history of the US. They should have been hanging by a thin thread by now but it seems they have too many thin threads keeping them hanging.

 

[igor_kavinski]

| פיסי אונליין

| פיסי אונליין

1-pc--online-co-il.translate.goog

It seems the i9-14900KS may be real.



That CANNOT be a typo, especially the 6.2 GHz part.

Guess this is Intel's Zen 5 counter strategy.

You know what's funny though?

Intel's top CPU being sold in an office PC

 

[Hulk]

| פיסי אונליין

| פיסי אונליין

1-pc--online-co-il.translate.goog

It seems the i9-14900KS may be real.

View attachment 88597

That CANNOT be a typo, especially the 6.2 GHz part.

Guess this is Intel's Zen 5 counter strategy.

You know what's funny though?

Intel's top CPU being sold in an office PC

Good find! Maybe it's aimed at traders that need the fastest ST performance to analyze and make trades!

 

[coercitiv]

Good find! Maybe it's aimed at traders that need the fastest ST performance to analyze and make trades!

With 16GB DDR4 3200?!

 

[igor_kavinski]

With 16GB DDR4 3200?!

Those trader workloads fit in the cache!

 

[Hulk]

With 16GB DDR4 3200?!

My post was intended to be humorous.

 

[igor_kavinski]

Overclocking "Raptor Lake refresh" -...

Just in case :p

www.overclock.net

Interesting theory on why 14900K is so underwhelming. Intel may have aggressively binned the 14900K production batches for accumulating enough CPUs for the 14900KS release.

 

[igor_kavinski]

Only one out of 22 chips that managed DDR5-8400: https://www.overclock.net/threads/o...lts-bins-and-discussion.1807439/post-29261329

 

[Hulk]

I was expecting the 14900K's to be more efficient than the 13th gen. They are turning out to be furnaces.

 

[igor_kavinski]

This guy reports the Honeywell paste to be good enough to give up on the idea of delidding: https://www.overclock.net/threads/o...lts-bins-and-discussion.1807439/post-29261369

Intriguing!

 

[TheELF]

Now they gonna try to survive the onslaught of Zen 5 with unattractive, boring desktop silicon,

How exiting do you expect zen 5 to be?!
They can't add another ccd because that would add 33% to the cost, they would be making 33% fewer CPUs out of the wafers they get.
They also wouldn't be able to power them since the 7950x already hits the maximum ppt for the platform.
If they add a bunch of zen 5c cores it will add a lot of complexity and issues like the thread director had, or like their own "game mode" , it will also force the main cores to use less power since the whole CPU is still bound to the max ppt, that is already being maxed out by the 7950x.
If they increase PPT people will be pissed because AMD promised a long life cycle for the platform.
Unless they have a complete redesign of the architecture up their...sleeves it's going to be a 10-15% increase in multithreaded at best.
15% if AMD manages to get every single bit of improvement out of the new node.
And while that is more exciting than 3-4% it's not exciting enough to make many people throw away their current systems to upgrade.
All this while hoping that TSMC doesn't increase their prices yet another time or even worse something happens that stops AMD from getting product from them again.

 

[TheELF]

I was expecting the 14900K's to be more efficient than the 13th gen. They are turning out to be furnaces.

To measure efficiency you would have to test both under the same conditions.
If you are testing max power draw you are just testing how much the resilience of the node changed.
With both running at the same power (253W ) the 14900k is 4% better than the 13900k and only a little bit better than the 13900k-S wich already was a refinement.
It's not much but considering that it's the same node it's about what was expected.

Intel Core i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks

i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks / Testmethodik / Multi-Core-Leistung / Single-Core-Leistung

www.computerbase.de

 

[Hulk]

To measure efficiency you would have to test both under the same conditions.
If you are testing max power draw you are just testing how much the resilience of the node changed.
With both running at the same power (253W ) the 14900k is 4% better than the 13900k and only a little bit better than the 13900k-S wich already was a refinement.
It's not much but considering that it's the same node it's about what was expected.

Intel Core i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks

i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks / Testmethodik / Multi-Core-Leistung / Single-Core-Leistung

www.computerbase.de

The problem with measuring power draw between 13th and 14th gen is the difference looks to be so small that it is within the variation of sample binning.
For example, 14900K parts depending on SP rating have 6Ghz VIDs that range from just over 1.5 volts to just under 1.4 volts. Of course 13900KS and some 13900K parts will also hit 6GHz. If they are all using the same voltage under load the efficiency will be the same. But if the higher binned parts are requesting lower VIDs and/or under volted a bit then they will measure as more efficient.

Thing is it looks like there are no process changes from 13 to 14 gen, just more binning, perhaps better yields. It's a lottery for the 14900K. You could get a part that needs 1.5V for 6GHz or you cold get one that needs 1.388 for 6GHz. That's a big difference.

I guess my point is that since there is fundamentally no difference between generations apart from binning (I believe) then it's a lottery as far as efficiency and differences will be minor.

If there is going to be a 14900KS I'll be very interested to see they they fare as far as 6GHz VID specs and the v/f curve points in general compared to the 14900K and 13900KS. From what I've been reading it seems the 13900KS are more consistent than the 14900K. Meaning they are all pretty good. With the 14900K you can get one that has a better v/f curve than a 13900KS or you could get a furnace (~1.5v VID @6Ghz).

Intel 7 is tapped out.

 

[Timur Born]

To measure efficiency you would have to test both under the same conditions.
If you are testing max power draw you are just testing how much the resilience of the node changed.
With both running at the same power (253W ) the 14900k is 4% better than the 13900k and only a little bit better than the 13900k-S wich already was a refinement.
It's not much but considering that it's the same node it's about what was expected.

Intel Core i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks

i9-14900K, i7-14700K & i5-14600K im Test: Anwendungs-Benchmarks / Testmethodik / Multi-Core-Leistung / Single-Core-Leistung

www.computerbase.de

But we don't learn anything about power/efficiency in this table, but just how different single-core frequencies unsurprisingly affect the results. And there are no "running at same power (253 W)" results in this table, as it is a single-core test.

My undervolted 13900K reaches 2360ish points at 38 W package power when running CB23 at 6.0 GHz, that's less than 1.35 V for a (real) single (not dual) core. And my CPU is just a "good enough" bin, not a stellar one.