Discussion Intel current and future Lakes & Rapids thread

Page 311 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Thala

Golden Member
Nov 12, 2014
1,355
653
136
January would have been less ugly.

Makes you wonder, considering Alder Lake is supposed to be a 2021 product as well, why they investing energy and releasing Rocket Lake at all.
I am afraid this is a totally irrelevant product in particular in the light of the Ryzen 5000 series.
 

DrMrLordX

Lifer
Apr 27, 2000
21,620
10,829
136
Makes you wonder, considering Alder Lake is supposed to be a 2021 product as well, why they investing energy and releasing Rocket Lake at all.
I am afraid this is a totally irrelevant product in particular in the light of the Ryzen 5000 series.

Probably a limited release? Also Alder Lake-S may slip into 2022, but if it launches in Nov/Dec 2021 then maybe Intel can save face.
 
  • Like
Reactions: spursindonesia

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Looks like Ocean Cove is dead and replaced with "Redwood Cove", whether this is large increase like Sunny/Golden Cove, or a smaller incremental upgrade akin to Willow (If I had to guess I'd say it's more like Willow. Intel seems to be on a cadence of Large Increase -> Small Increase -> Large Increase -> etc...)

Redwood Cove being like Willow Cove gains is going to be laughably bad. Unless Golden Cove is a titanic jump and its 35-40% faster than Willow Cove to make up for the fact they have a year of minimal gains versus AMD that is doing 15-20% every year.

I do think the cache hierarchy change in WC could be preparing for future uarch expansions. Similar to Nehalem's 3-level cache hierarchy.

It only has 80 EUs which is the reason for the relatively 3dmark scores. The other Acer goes down to sub 20W.

The 17W Swift is also faster in games. What I'd like to see is a 25-28W Swift using the i7.

The first to market devices tend to be disappointing. The ICL XPS 13 was a premium device but it had issues, aside from needing lots of power for its performance. The title goes to Asus now.

Tigerlake seemingly clocks fine to 4.8 in a ULV product and isn't that just Sunny Cove built on the 10SF node?

Not in the strictest sense.

L2 is considered a core cache in modern CPUs, and they had to increase that significantly. L3 cache size also changes and so does the caching scheme to exclusive. They both have to be balanced for latency in terms of cycles which will affect clocks.

Also Tigerlake as a CPU gets the dual ring bus.
 
Last edited:

mikk

Diamond Member
May 15, 2012
4,133
2,136
136
Redwood Cove being like Willow Cove gains is going to be laughably bad. Unless Golden Cove is a titanic jump and its 35-40% faster than Willow Cove to make up for the fact they have a year of minimal gains versus AMD that is doing 15-20% every year.

I do think the cache hierarchy change in WC could be preparing for future uarch expansions. Similar to Nehalem's 3-level cache hierarchy.



The 17W Swift is also faster.


Yes in 3dmark, as I said 3dmark is masking the power limit. Most power limited devices are not that bad in 3dmark and I guess most reviewer are using the first run, look how the scores fall down in Cinebench over several runs on many of these ultrathin devices. The problem starts in real world gaming. Regarding cooling, the best cooling from a Tigerlake device I have seen so far comes from the Vivobook S533. It runs constantly at 28W and in Cinebench R20 goes up to only 75-80 degrees:


Compared to MSI Summit B15 with a two fans solution at a constant 28W: 82-84 degrees:


Still very good. MSI Prestige 14 runs hotter with its 1 fan solution. The MSI Summit is using x8 modules there, it achieves better Firestrike results than the Asus. The S533 has only 1 slot and the second is soldered which means it's either running on mixed x16+x8 or both x16. That's why it's slower. DDR4 devices with 2 dimm slots have an advantage there, they can use x8 or even 2R x8 modules (requires 2x16GB).

The cooling performance bodes well for the Vivobook TP470 with Iris Xe Max. All of the new Vivobook appears to get the same fan/heatpipe.


ASUS Intelligent Performance Technology (AIPT)

Aerodynamic IceBlade for accelerated cooling

Each model in the new VivoBook series provides exceptional heat dissipation via an upgraded 8 mm heat pipe and a new IceBlade fan that efficiently accelerates heat transfer. The 87-blade IceBlade fan and impeller are made of a liquid-crystal polymer that allows them to be lighter and thinner than ordinary fans. Each IceBlade fan blade has a 3D-curved aerodynamic design, allowing the fan to achieve better flow rate and lower noise. In addition, the IceBlade fans also use a fluid-dynamics bearing, which can achieve better vibration reduction and lower noise when compared to ordinary fans.

If the air intakes are not closed on this the cooling on this should be better than the Swift 3x.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
@mikk The gaming performance is not good on both. The MSI device compared to a 4800U would be maybe 5-10% faster. You can see it underperforms the 4700U in few games. I want to see 30-40% gain over the 4800U.
 

mikk

Diamond Member
May 15, 2012
4,133
2,136
136
@mikk The gaming performance is not good on both. The MSI device compared to a 4800U would be maybe 5-10% faster. You can see it underperforms the 4700U in few games. I want to see 30-40% gain over the 4800U.

I know, it becomes clear from the Firestrike scores that they are lacking. The point is that among these DDR4 devices GPU performance can differ, 2 slot devices can be better. Unless Intel can work on driver related memory compression improvements (maybe something is broken atm) this won't change unfortunately. They really need DDR5 with Xe LP which we will see on ADL-P.
 

mikk

Diamond Member
May 15, 2012
4,133
2,136
136
28W DDR4 vs 15W LPDDR4x



What a difference in Ungine Heaven/Unigine Valley LPDDR4x makes, insane. However when it comes to real world gaming the 15W device struggles (as expected), lots of spikes up and down.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
They really need DDR5 with Xe LP which we will see on ADL-P.

I hope we see way more than just Xe with LPDDR5. Like an update for additional performance. Xe is barely competitive with Vega graphics. 30-40% best case win over your competitor that uses an older architecture does not bode well at all. Van Gogh is a real threat. Oh and the state of their drivers at this point. Tsk tsk.

At least its doing way better than their CPU line haha.

Both Arrandale/Clarkdale's Gen 5.75 and Haswell's Gen 7.5 were significant updates to their predecessors. Gen 5.75 added improved occlusion techniques thus along with on-package GMCH ended up being 2.5x better than the one in GMA X4000.
 

ThatBuzzkiller

Golden Member
Nov 14, 2014
1,120
260
136
Intel should fear the future as much as possible ...

For the last decade, the vast majority of the developers specifically optimized their code for Intel microarchitectures. If developers start designing their software around their competitor's microarchitecture then the performance gap could potentially become wider or be maintained ...

Even with the developers on Intel's side, they still couldn't keep their performance leadership in any capacity ...
 

mikk

Diamond Member
May 15, 2012
4,133
2,136
136
I hope we see way more than just Xe with LPDDR5. Like an update for additional performance. Xe is barely competitive with Vega graphics. 30-40% best case win over your competitor that uses an older architecture does not bode well at all. Van Gogh is a real threat. Oh and the state of their drivers at this point. Tsk tsk.

At least its doing way better than their CPU line haha.

Both Arrandale/Clarkdale's Gen 5.75 and Haswell's Gen 7.5 were significant updates to their predecessors. Gen 5.75 added improved occlusion techniques thus along with on-package GMCH ended up being 2.5x better than the one in GMA X4000.


The peak performance is really good from Xe LP but it needs a lot more bandwidth for some reason and also a 25W+ Tigerlake version. Not sure if the memory compression is subpar or if it's partially a driver thing. There is a new MSI Prestige 14 user test on youtube, it reaches a graphics score of 5904 in Firestrike and 1673 in timespy. With enough bandwidth and 28W it can be really fast.
 

yuri69

Senior member
Jul 16, 2013
387
617
136
Intel should fear the future as much as possible ...

For the last decade, the vast majority of the developers specifically optimized their code for Intel microarchitectures. If developers start designing their software around their competitor's microarchitecture then the performance gap could potentially become wider or be maintained ...

Even with the developers on Intel's side, they still couldn't keep their performance leadership in any capacity ...
Intel is still the undisputed leader in volume & SW toolchain. This won't change any time soon. So the targeted "average CPU" will still be an Intel.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Intel is still the undisputed leader in volume & SW toolchain. This won't change any time soon. So the targeted "average CPU" will still be an Intel.

They have advantage in some areas, and people will still go for them, but most of the market is eventually going to go to AMD if they never get competitive again.

It just takes time that's all. Over many years they've established trust from consumers because they had a better product. There's a delay due to the buffer created, but it also takes time to create the said buffer.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
The peak performance is really good from Xe LP but it needs a lot more bandwidth for some reason and also a 25W+ Tigerlake version. Not sure if the memory compression is subpar or if it's partially a driver thing. There is a new MSI Prestige 14 user test on youtube, it reaches a graphics score of 5904 in Firestrike and 1673 in timespy. With enough bandwidth and 28W it can be really fast.

The reviews are all lacking in some respects. I'd do some tests and eventually I want a 2-in-1 again but not spending money on what seems like a product that misses its mark. Tigerlake has been an increasing disappointment for me starting from July. Really started when they did their information-less PR launch.

Based on frequency and power numbers, I'm wondering if there's significant variances and binning in play? If they are struggling with 10nm, and 10nm SF being a new process, there has to be a noticeable difference.

The Chinese reviewer that did rank testing also did frequency/power testing for the CPU. The frequency seems too low and the power too high compared to what's out there.
 

Hulk

Diamond Member
Oct 9, 1999
4,214
2,006
136
I am thinking that Rocket Lake is using the Sunny Lake architecture instead of Willow Cove due to limitation in die space at 14nm. That's also probably why we are only seeing a maximum of 8-core parts.

It's a shame that Intel can't produce their fastest architecture (Willow Cove) at 10nm for the desktop for Rocket Lake. It would be a good competitor for Zen 3.
 

Ajay

Lifer
Jan 8, 2001
15,430
7,849
136
I am thinking that Rocket Lake is using the Sunny Lake architecture instead of Willow Cove due to limitation in die space at 14nm. That's also probably why we are only seeing a maximum of 8-core parts.

It's a shame that Intel can't produce their fastest architecture (Willow Cove) at 10nm for the desktop for Rocket Lake. It would be a good competitor for Zen 3.
Well, no. 10nm, given the marketing name '10SF' (super fin) is still broken. There's a decent probability that ADL will regress in performance. Maybe Intel will be able to produce enough to heavily bin for a top performer, but then Zen4 will be here.
 
  • Like
Reactions: spursindonesia

ondma

Platinum Member
Mar 18, 2018
2,720
1,280
136
They have advantage in some areas, and people will still go for them, but most of the market is eventually going to go to AMD if they never get competitive again.

It just takes time that's all. Over many years they've established trust from consumers because they had a better product. There's a delay due to the buffer created, but it also takes time to create the said buffer.
There is also the matter of capacity.
 

Hulk

Diamond Member
Oct 9, 1999
4,214
2,006
136
Well, no. 10nm, given the marketing name '10SF' (super fin) is still broken. There's a decent probability that ADL will regress in performance. Maybe Intel will be able to produce enough to heavily bin for a top performer, but then Zen4 will be here.

Aren't they currently shipping/selling 10SF Tiger Lake parts? I think the problem is two-fold. First, they don't yet have capacity for TG mobile and desktop, and second and probably more to the point is that while 10SF is working for TG (low leakage/low power), they don't have it ready for high performance (high clocks damn the TDP) desktop parts. And they know they need the frequency to compete with Zen 3.

It was probably something like do we put out 10SF Tiger Lake desktop parts with <5GHz max clocks much later and/or in lower supply, or go with 14nm+++ for max clocks using Sunny Cove now. Now being 6 months from now actually.
 
  • Like
Reactions: Executor_

DrMrLordX

Lifer
Apr 27, 2000
21,620
10,829
136
Aren't they currently shipping/selling 10SF Tiger Lake parts?

Yes, but curiously, they're still only 4c parts. Intel has yet to market anything on any iteration of 10nm with more than 4c. 8c Tiger Lake-H is coming sometimes next year, albeit in unknown quantity.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
If they were really interested in extra volume on desktop, they could have released "T" series processors based on Tiger Lake 4C die. They would have made fine i3, low power processors for corporate environments. There's no need though. That market segment is handled well enough by the Comet Lake i3 parts in low power modes.
 

Shivansps

Diamond Member
Sep 11, 2013
3,851
1,518
136
If they were really interested in extra volume on desktop, they could have released "T" series processors based on Tiger Lake 4C die. They would have made fine i3, low power processors for corporate environments. There's no need though. That market segment is handled well enough by the Comet Lake i3 parts in low power modes.

It would have worked vs Raven and Picasso. But that would have caused AMD to use Renoir far more agressively on desktop.
 

Dave2150

Senior member
Jan 20, 2015
639
178
116
Well, no. 10nm, given the marketing name '10SF' (super fin) is still broken. There's a decent probability that ADL will regress in performance. Maybe Intel will be able to produce enough to heavily bin for a top performer, but then Zen4 will be here.

AMD have likely picked all the low hanging fruit that Jim Keller left them for Ryzen. I highly doubt they'll manage more than a 5-10% IPC increase again, before going to a new architecture. Hopefully I'm wrong though, their pace of innovation to this point (using Jim Keller's designs) have been refreshing.
 
  • Haha
Reactions: spursindonesia

chrisjames61

Senior member
Dec 31, 2013
721
446
136
AMD have likely picked all the low hanging fruit that Jim Keller left them for Ryzen. I highly doubt they'll manage more than a 5-10% IPC increase again, before going to a new architecture. Hopefully I'm wrong though, their pace of innovation to this point (using Jim Keller's designs) have been refreshing.
But they just managed 19%? So where did you come up with this theory?
 

Hulk

Diamond Member
Oct 9, 1999
4,214
2,006
136
Slightly off the current thread topic, but still pertinent...

What is the minimum number of applications do you think would be required to be tested to arrive at a representative IPC metric? IPC is important but we still have a hard time identifying metric that most of this forum can agree on for making cpu core IPC comparisons. If we could come up with a small number, say 5 or 6 benchmarks that would be easily available to forum members we could easily create a database so discussions regarding IPC could be apples-to-apples, at least for this forum and based on those tests.

I'd be willing to compile results. We really only need a few results from each core. The hard part is getting accurate clocks while tests are being run and isolating to one core/2 threads during testing. Which can actually be accomplished using the Task Monitor by shutting down cores.
 

podspi

Golden Member
Jan 11, 2011
1,965
71
91
AMD have likely picked all the low hanging fruit that Jim Keller left them for Ryzen. I highly doubt they'll manage more than a 5-10% IPC increase again, before going to a new architecture. Hopefully I'm wrong though, their pace of innovation to this point (using Jim Keller's designs) have been refreshing.

But they just managed 19%? So where did you come up with this theory?


Zen 3 is a "clean slate" design. Zen 2 is the result of Keller's low-hanging fruit being picked.


Anandtech said:
Typically when designing a CPU core, the easiest thing to do is to take the previous design and upgrade certain parts of it – or what engineers call tackling ‘the low hanging fruit’ which enables the most speed-up for the least effort. Because CPU core designs are built to a deadline, there are always ideas that never make it into the final design, but those become the easiest targets for the next generation. This is what we saw with Zen 1/Zen+ moving on to Zen 2. So naturally, the easiest thing for AMD to do would be the same again, but with Zen 3.
However, AMD did not do this. In our interviews with AMD’s senior staff, we have known that AMD has two independent CPU core design teams that aim to leapfrog each other as they build newer, high performance cores. Zen 1 and Zen 2 were products from the first core design team, and now Zen 3 is the product from the second design team. Naturally we then expect Zen 4 to be the next generation of Zen 3, with ‘the low hanging fruit’ taken care of.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
It would have worked vs Raven and Picasso. But that would have caused AMD to use Renoir far more agressively on desktop.

AMD has nowhere near the capacity, nor the market segment level demand to use Renoir any more agresively than they already are on desktop, which is to say barely