Discussion Intel current and future Lakes & Rapids thread

[dullard]

The 13900ks also had 3.2 GHz base. . .

At 150 W for the 13900KS and 125 W for the 14900K. Which supports exactly what you quoted:

At iso clocks will it consume less power

 

[Hitman928]

At 150 W.

Has the 14900k been officially spec'd at 125W? I haven't been following that closely to the refresh.

 

[dullard]

Has the 14900k been officially spec'd at 125W? I haven't been following that closely to the refresh.

Nothing official, but having a refresh that uses more power than the original would make it incompatible with many motherboards (Especially OEM motherboards which usually aren't overbuilt by as much as enthusiast motherboards). So, going over 125 W base would be a really bad move. As for unofficial, this leak just came out showing 125 W TDP:


To me, Raptor Refresh will have nothing to brag about performance wise. Certainly nothing to upgrade over. But, if you need a new computer, might as well get a few extra MHz for about the same power. And if all someone cares about is performance/power, this is delivering exactly that type of improvement--just nothing earth shattering.

 

[Hitman928]

Nothing official, but having a refresh that uses more power than the original would make it incompatible with many motherboards (Especially OEM motherboards which usually aren't overbuilt by as much as enthusiast motherboards). So, going over 125 W base would be a really bad move. As for unofficial, this leak just came out showing 125 W TDP:
View attachment 85371

To me, Raptor Refresh will have nothing to brag about performance wise. Certainly nothing to upgrade over. But, if you need a new computer, might as well get a few extra MHz for about the same power.

Fair enough. We'll see when it gets officially released and reviewed, but a decent efficiency bump would certainly be welcome, even if performance increase isn't really much.

 

[Geddagod]

Don't know if this has been posted already but YukiAnns collected what I'm assume is a collection of ES samples of GNR and SRF sent out to OEMs. I'm a little surprised, tbh, that GNR is on their 2nd? 3rd? (A0, A1, A2?) compared to A0 SRF, especially since GNR is set to launch later than SRF. But perhaps SRF is less "buggy" on its initial silicon, or perhaps Yuki just hasn't gotten those samples yet.
It was already reported that the GNR single tile variants (I'm guessing the 32 core version) has been shipped out to OEMs a while back, but with the 56 core versions, that means the 2 tile variants are being shipped out to OEMs too.
Overall, this isn't anything high impact or anything, but I do find it a bit interesting.

 

[Ajay]

I assume that's 88% Windows PC market share because I'm pretty sure Apple is at or pretty close to 12% share of laptops all by itself given that the Mac market has a far higher share of laptops vs desktops than the Windows PC market.

Apple has in excess of 20% of the laptop market (~25% last time I looked it up). The 12% share is compared to all PCs.l Whew, way behind in this thread.

 

[Joe NYC]

View attachment 85382
Don't know if this has been posted already but YukiAnns collected what I'm assume is a collection of ES samples of GNR and SRF sent out to OEMs. I'm a little surprised, tbh, that GNR is on their 2nd? 3rd? (A0, A1, A2?) compared to A0 SRF, especially since GNR is set to launch later than SRF. But perhaps SRF is less "buggy" on its initial silicon, or perhaps Yuki just hasn't gotten those samples yet.
It was already reported that the GNR single tile variants (I'm guessing the 32 core version) has been shipped out to OEMs a while back, but with the 56 core versions, that means the 2 tile variants are being shipped out to OEMs too.
Overall, this isn't anything high impact or anything, but I do find it a bit interesting.

Wasn't there supposed to be a GNR version with 120 cores?

Or has it, perhaps, not been released yet for sampling?

 

[Geddagod]

Wasn't there supposed to be a GNR version with 120 cores?

Or has it, perhaps, not been released yet for sampling?

I'm guessing the second one. It appears as if only the one tile and two tile GNR variants are out (max ~80 something cores?).
Also wondering if GNR has any monolithic dies like SPR does. I would hope it doesn't, considering all the extra time and money designing the extra dies, but the monolithic advantage has been allowing SPR to compete much better than it should be able too in their lower core count server parts. However, only XCC GNR seems to be sampling rn.

 

[Saylick]

View attachment 85382
Don't know if this has been posted already but YukiAnns collected what I'm assume is a collection of ES samples of GNR and SRF sent out to OEMs. I'm a little surprised, tbh, that GNR is on their 2nd? 3rd? (A0, A1, A2?) compared to A0 SRF, especially since GNR is set to launch later than SRF. But perhaps SRF is less "buggy" on its initial silicon, or perhaps Yuki just hasn't gotten those samples yet.
It was already reported that the GNR single tile variants (I'm guessing the 32 core version) has been shipped out to OEMs a while back, but with the 56 core versions, that means the 2 tile variants are being shipped out to OEMs too.
Overall, this isn't anything high impact or anything, but I do find it a bit interesting.

Yuck. 1.2 GHz base and 2.4 GHz boost clocks for 56c GNR? That's it?

 

[Hulk]

While any efficiency increase is welcome regardless of the actual power consumption, the fact that it's hard to run a 13900K at max stock speeds without serious cooling means that the 14900K could be a legit upgrade for those thermally constrained with a 13900K, like me. Of course it all depends on pricing...

 

[Markfw]

Yuck. 1.2 GHz base and 2.4 GHz boost clocks for 56c GNR? That's it?

I was going to say something, but after you have commented, I will just say, yes, thats not good. My 9554 Genoas do 3.5 ghz all-core under the heaviest loads (avx-512) %100% load, so even for a ES that seems low.

Edit: and my heat sinks are what I call "sub-optimal" for for a non-server situation, those do not exist yet, until Threadripper Zen 4 HSF's make their debut.

 

[H433x0n]

Yuck. 1.2 GHz base and 2.4 GHz boost clocks for 56c GNR? That's it?

That’s not that unusual at this stage. The final clocks could end up at more than double these values.

 

[Joe NYC]

That’s not that unusual at this stage. The final clocks could end up at more than double these values.

I am going to make a wild guess that the Turin CPUs that are sampling are not running at 1.2 GHz.

 

[H433x0n]

I am going to make a wild guess that the Turin CPUs that are sampling are not running at 1.2 GHz.

I would hope so. Their job is a lot easier comparatively.

also - Isn’t this an Intel thread? You’ve got an entire thread dedicated to Turin.

 

[Henry swagger]

Intel Core Ultra 7 155H processor spotted with 16 cores and 4.8 GHz clock - VideoCardz.com

Core Ultra 7 155H, Intel Meteor Lake with new naming Intel’s next gen mobile chip to feature 16 cores. This leak marks the initial sighting of a new CPU from Intel’s Meteor Lake series, identified as the “Ultra 7 155H.” It stands out from its predecessors, notably the 1002H and 1003 models, due...

videocardz.com

Matches the 13700h .. wonder at what power

 

[Geddagod]

Yuck. 1.2 GHz base and 2.4 GHz boost clocks for 56c GNR? That's it?

It's ES 1....

I am going to make a wild guess that the Turin CPUs that are sampling are not running at 1.2 GHz.

I mean, Turin is gonna release like ~1/2 of a year ahead of GNR, and I'm pretty sure AMD has better ES samples as well since they usually take less steppings to launch their product.

But honestly, if you guys want to believe that GLC ported to a new process and using what looks like a better design (larger tiles means less power loss to EMIB) for power is going to clock like half as low as SPR already does, then I'm not going to try to change your mind lol.

 

[JoeRambo]

Are those GNR variants for 2 compute die ones? I assume so, as there are only 8 MC and 288MB* of L3 total on those chips.
Full chip having 432MB of L3 is substantial lump of cache and combined with 6400 DDR5 might pleasantly surprise in performance department. Also 28 cores is decent sized NUMA domain with plenty of L3 and 4 channels of 6400.

If Intel's 3 is not power hog, i don't see any problems for Intel in non-cloud sales.

*288 might be L2 + L3, not a big reader of hieroglyphs.

 

[mikk]

Core Ultra 5 125H 4.5
Core Ultra 7 155H 4.8/4.8/4.5/4.3+3.8/2.8
Core Ultra 7 165H 5/5/4.7/4.4+3.8/2.8
Core Ultra 9 185H 5.1/5.1/4.8/4.5+3.8/2.8

动态-哔哩哔哩

t.bilibili.com

First Geekbench entries from 125H and 155H appeared.

Ultra 5 125H

HP OMEN Transcend Gaming Laptop 14-fb0xxx - Geekbench

Benchmark results for a HP OMEN Transcend Gaming Laptop 14-fb0xxx with an Intel Core Ultra 5 125H processor.

browser.geekbench.com

Ultra 7 155H

HP HP Spectre x360 2-in-1 Laptop 16-aa0xxx - Geekbench

Benchmark results for a HP HP Spectre x360 2-in-1 Laptop 16-aa0xxx with an Intel Core Ultra 7 155H processor.

browser.geekbench.com

 

[Geddagod]

Core Ultra 5 125H 4.5
Core Ultra 7 155H 4.8/4.8/4.5/4.3+3.8/2.8
Core Ultra 7 165H 5/5/4.7/4.4+3.8/2.8
Core Ultra 9 185H 5.1/5.1/4.8/4.5+3.8/2.8

动态-哔哩哔哩

t.bilibili.com

First Geekbench entries from 125H and 155H appeared.

Ultra 5 125H

HP OMEN Transcend Gaming Laptop 14-fb0xxx - Geekbench

Benchmark results for a HP OMEN Transcend Gaming Laptop 14-fb0xxx with an Intel Core Ultra 5 125H processor.

browser.geekbench.com

Ultra 7 155H

HP HP Spectre x360 2-in-1 Laptop 16-aa0xxx - Geekbench

Benchmark results for a HP HP Spectre x360 2-in-1 Laptop 16-aa0xxx with an Intel Core Ultra 7 155H processor.

browser.geekbench.com

Not super encouraging tbh for MTL tbh, PPC looks to be marginally worse than Zen 4. However, GB has typically had higher scores for AMD anyway, so not too surprised.
This would put single core ~5% higher for Raphael, and multi core (for what little that's worth based on the no power consumption given), appears to be higher for MTL than Raphael.

 

[mikk]

There is always a big fluctuation from mobile CPUs in the Geekbench scores, it's impossible to draw a conclusion from this at the moment.

 

[FangBLade]

Not super encouraging tbh for MTL tbh, PPC looks to be marginally worse than Zen 4. However, GB has typically had higher scores for AMD anyway, so not too surprised.
This would put single core ~5% higher for Raphael, and multi core (for what little that's worth based on the no power consumption given), appears to be higher for MTL than Raphael.

That's right, but Intel is so inefficient that even if they only reduce power consumption and keep the same performance, it's considered a 'success.' I don't like the direction Intel is going in. New generations of mobile and desktop processors practically offer minimal improvements in terms of performance. When I buy a new generation, I don't want it to only offer lower power consumption; I want better performance as well. Otherwise, I won't buy them just for the sake of lower power consumption. Intel is still searching, experimenting, trying to find the winning formula for chiplet design to break out of their rut, while AMD has a PhD in chiplet design and is moving full steam ahead. Also, when we talk about chiplet design, Intel's solution is much more expensive than AMD's, and we'll see what the prices will be. However, considering they have their own factories and produce much larger quantities, they can afford minimal profit per unit.

 

[João Bortolace]

GNR-AP

 

[Geddagod]

There is always a big fluctuation from mobile CPUs in the Geekbench scores, it's impossible to draw a conclusion from this at the moment.

I don't think it's that bad lol, is it?

That's right, but Intel is so inefficient that even if they only reduce power consumption and keep the same performance, it's considered a 'success.'

That's cuz Intel has an yearly release schedule and AMD's is like ~1.5 years.

Intel is still searching, experimenting, trying to find the winning formula for chiplet design to break out of their rut,

Their server CPUs are changing how they use chiplets a lot, but their client platforms use of chiplets (bad LNL) looks pretty straight forward until prob post ARL or post PTL.

while AMD has a PhD in chiplet design and is moving full steam ahead.

Zen 6 is apparently a huge change in how AMD uses chiplets lol

Also, when we talk about chiplet design, Intel's solution is much more expensive than AMD's, and we'll see what the prices will be.

Ye that's true.

However, considering they have their own factories and produce much larger quantities, they can afford minimal profit per unit.

Think MTL stock is gonna be pretty bad in 2023 tbh. But AMD's mobile launch has been pretty slow as well, so....

 

[dullard]

GNR-AP

There are more photos released today. https://www.intel.com/content/www/us/en/newsroom/resources/global-manufacturing.html#gs.5d9no9

That particular screenshot came from this video: https://www.intel.com/content/www/u...he-way-with-advanced-packaging.html#gs.5da3kb

 

[Geddagod]

GNR-AP

That's gotta lend itself to a pretty accurate estimation of the die size of the chiplets, nah? Based off the package size...
Also weird that one of the IO chiplets is ever so slightly offset lol, but whatever