Discussion Intel current and future Lakes & Rapids thread

Page 278 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Isn't ADL more of a H2 2021 - H1 2022 kinda' product? I was under the impression the planning soup contained RKL + TGL for H1 2021.

H2 2021 was what it was "supposed" to be. But if you're launching Tiger Lake-H in H1 2021, it'll have a very short shelf life. Ditto for Rocket Lake-S as well. Assuming everything was moving along smoothly at Intel (hah!) you would think Tiger Lake would have launched in July 2020 (rather than September, ugh) and Alder Lake would show up in July 2021.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
I have to prove what? Look at MebiuW's profile on Weibo, you'll probably find an excel spreadsheet containing V/f curve points at some point. Look for one with a replies that say something akin to the following:

Then feel free to ask Bepo about that table yourself. Just ask him over Twitter. He'll tell you exactly what that table is. Do a little bit of research and you won't have to believe me but rather those directly involved.


And where is your proof that this is related to MebiuW? And he didn't predict a 4.7+ Ghz boost there, so how is this related? There is not even an exact date visible lol. I think you are really confused.


Isn't ADL more of a H2 2021 - H1 2022 kinda' product? I was under the impression the planning soup contained RKL + TGL for H1 2021.


ADL-P might start for the ULV segment just like Tigerlake. And yes ADL-P is H2 2021.
 

tomatosummit

Member
Mar 21, 2019
184
177
116

6-8-core 35-45W Tiger Lake for next year, though the GPU is slower.

EDIT: maybe we will see some SFF PCs with these chips in them.
Am I missing something? Everyone is saying it's officially tiger lake but the quote from intel man only says "willow cove".
The ghost of rocket lake still hangs in the air and alder lake is a possibility too.

Is bga1787 a new socket? It being on a new "socket" is also suspicious in itself as it looks like it is similar the desktop leap to lga1700 assuming the new pins are used for the same kind of stuff.
 

Adonisds

Member
Oct 27, 2019
98
33
51
When are we going to see Tiger Lake reviews? When will a single person be able to buy a Tiger Lake product? The media should stop spreading these paper launch dates that AMD and Intel give as the launch dates, it's ridiculous
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
I don't recall adding pipeline stages to be a major issue complexity wise (especially with modern tools and sims).

The core transistors doubled between Coppermine and Willamette and tripled between Northwood and Prescott.

Sure, its not purely pipeline stages but what was needed to make up for the loss. And it still was an overall loss.

Modern AMD/Intel CPUs are in 19-20 stage range. Considering how well Apple and ARM CPU does, I think we're due for another substantial architecture change focusing on lower clocks. AMD is better than Intel here though.

You can also see that with Silvermont. They went from an In-order core to an OoOE one. Core size did not increase at all. Yes, Silvermont loses support for HT, but it also has 2 less stages. Yet its 50% faster per clock versus Saltwell. So it went from needing multi-threading to be 35-40% faster, to not having HT but is 50% faster in single thread too. Combined with the 22nm process bringing 37% higher clocks, its more than 2x faster in ST and 50% faster in MT at same power.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Looking forward to Alderlake: AVX-256 unit size comparison
1600134856285.png


As you can see, the 2x256-bit unit on Sunny Cove is a substantial fraction of the entire Tremont core! On the 10nm process, 2x256-bit units are about 0.5mm2, while Tremont is 0.85mm2.

However, the Sunny Cove portion is using much larger cells for its cores. The 2x512-bit AVX512 blocks are much more compact on Knights Landing:

On Knights Landing, despite being on the 14nm process, the twin AVX-512 units are only <1mm2 in size.

If you assume Gracemont gets 2x256-bit units just like on Sunny Cove, a very conservative size reduction of 50% from 14nm would result in 0.25mm2.

Since Tremont already supports 2x128-bit SSE, AVX2 support will only need an additional <0.125mm2 in core area. Gracemont core even with substantially better performing uarch still might end up being under 1.5mm2 in size.
 

dullard

Elite Member
May 21, 2001
24,998
3,327
126
When are we going to see Tiger Lake reviews? When will a single person be able to buy a Tiger Lake product? The media should stop spreading these paper launch dates that AMD and Intel give as the launch dates, it's ridiculous
For laptop processors, there really isn't a good other option. Intel and AMD announce the processors when they are shipped to the laptop manufacturers and the processors are now out of their hands and into the wild. If the mobile processors weren't announced before they were purchasable, then any one or any company can rush to be the first to announce all the details. It would be a mad dash by laptop manufacturers to rush the data out far before the laptops have been tuned. The result is low quality reporting of different details by everyone in unfinished products. Or, AMD/Intel can announce the details in a controlled fashion with their own branding and let the benchmarks come out later.

Desktops are plug/play and you can get some benchmarks in under an hour. You get good data on the same day the chips ship. Laptops need to be specifically tailored to the exact hot spots of the processors. It takes weeks or more to do this right. So laptop benchmarks are delayed. Since the juicy details have already been announced, a laptop manufacturer can take the time to make the best product they can and let others benchmark that result.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
17-18W sustained load for this model (because the cooling isn't good enough?). This device seems to use DDR4-3200, there is only a very little performance degradation in Firestrike and timespy.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
Those graphics numbers look quite good for the segment. They are notably better than the equivalent 10th Gen i7G7 config and the competing 4700 and 4800u numbers at 15w. I suggest that its possible that AMD may decide to carry more enabled CUs in lower SKUs in the follow on products to Renoir, especially in the 6 and 8 core ones, while also using process refinements to push clocks on those units some.
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
looks to me that that book (swift) can't even hold 17W sustainable
2,7GHz is low
the before 3,2GHz is more believable, I expect 3,4GHz sustainable within 15W in cinebench, 3,1 in handbrake
but let's wait for the reviews
overall tigerlake=pentium m dothan? anyone remember?

tiger lake H is IMO ready, rocket lake is not
tiger lake H with proper tuned m-atx board with proper tuned RAM can be fast as 10900K and in single threaded it looks like it has a power of 5,8GHz skylake core, which is scary
Intel can't afford that their 45WH notebook chip is faster in CAD, excel, matlab....etc than their top workstation and desktop CPUs
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
17-18W sustained load for this model (because the cooling isn't good enough?). This device seems to use DDR4-3200, there is only a very little performance degradation in Firestrike and timespy.

It's not thermal throttling. PL1 is 17W.

Also the previous Icelake generation used LPDDR4x. The reports saying it uses DDR4 are likely wrong. The other reviews that use Tigerlake is also using the Swift 5 and its LPDDR4x.

@TheGiant Tigerlake is no Banias/Dothan.

2.7GHz at 17W is much better than Icelake and about on par with it at 25W setting even though its only about a 25% gain. On Icelake, its 2-2.2GHz.
 
Last edited:

itsmydamnation

Platinum Member
Feb 6, 2011
2,744
3,078
136
Another Tigerlake preview: https://translate.google.com/transl...or-intel-core-11-generace&prev=search&pto=aue

4.6GHz single core @ 17W. I compared with other laptops on that site. The GPU is comparable to an MX 350 with 64-bit memory. The 128-bit variants will need a 25W setting.
17watt single core is better then i expected honestly , was expecting mid 20's. Will be interesting to see how AMD react with Cezanne will they push single core max wattage higher to increase clocks or keep it at the lower values like renior.
 
  • Like
Reactions: lightmanek

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
17watt single core is better then i expected honestly , was expecting mid 20's. Will be interesting to see how AMD react with Cezanne will they push single core max wattage higher to increase clocks or keep it at the lower values like renior.

20W @ 4.7GHz.

Even with Willow Cove, they are acting like they have mini Pentium 4. Really should have seen the chip in 2018, so little wonder there. At 4.3GHz like in earlier leaks it might only need 10-11W.

@LightningZ71 Intel should be able to push Iris Xe further too. At 1.3GHz its way under what Icelake uses for power.

According to SharkBay, there's a Tigerlake-H35. That's the same chip used for U chips but set at 35W. Maybe that'll use Iris Xe Max with 1.5-1.65GHz GPU clocks?
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
CPU+GPU load and long term comparisons:

The Acer Swift 5 using Tigerlake sustains 1.5GHz CPU and 1GHz+ GPU at 17W. The best-in-class Icelake device is the Microsoft Surface Laptop 3. That device does 1.3GHz CPU and 850MHz GPU but at 25W.

At 15W, Icelake drops to 0.9-1GHz CPU and 600-700MHz GPU. Remember, that's with a CPU benchmark and GPU benchmark running together.

In Cinebench R20 MT, Tigerlake @ 17W can maintain 1680 points. Surface Laptop 3 @ 25W with 1065G7 is at 1400 points.
 
Last edited:

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
I agree that there is likely thermal and power headroom for Xe on TigerLake, but, we do not know how well Xe on 10sf can actually clock. It may have some interesting process relatd clock limits in this application. It may not matter in the end, it looks like Intel made Xe wide and lower frequency here.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
but, we do not know how well Xe on 10sf can actually clock.

Actually we kinda do. The Xe-LP presentation had the peak frequency going above 1.7GHz.

Of course it uses higher power, but if they wish to, it can be raised. At the same power as Gen 11(slightly lower voltage), it should be able to reach 1.5GHz.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
The HWinfo screen says 2130 Mhz CL 36-39-39-90, so it should use LPDDR4 4266. The 3dmark scores are very high though considering this is not a 28W device with proper cooling.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Notebookcheck's Tigerlake Iris Xe gets 1778 points in 3DMark Time Spy. 85% higher than Gen 11, and 33% higher than 4800U at 25W with LPDDR4x.

Did some analysis based on the review I last linked. Time Spy seems to be a good indicator of performance at this level. The Swift 1165G7 system is underperforming by 10-15% in Alien vs. Predator game compared to MX350 considering Time Spy scores, but there will be implementation differences.

Considering how balanced the Microsoft Surface Laptop 3 is, I'd love to see what they got out of Tigerlake.

A bad implementation is the Dell XPS 13. The MS SL3 is 20-30% faster in most games, and doesn't drop performance under battery, while XPS will drop it by 30%. XPS also requires a 46W PL1 setting under AC to achieve those figures, while SL3 remains at 25W.

Honestly I don't think a drop from 28W to 25W will affect it a whole lot. The GPU is quite power efficient. The higher TDP might impact it in CPU limited situations by 5-10%, but this is in games that will be getting 70-80 fps.
 
  • Like
Reactions: lightmanek

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
20W @ 4.7GHz.

Even with Willow Cove, they are acting like they have mini Pentium 4. Really should have seen the chip in 2018, so little wonder there. At 4.3GHz like in earlier leaks it might only need 10-11W.

@LightningZ71 Intel should be able to push Iris Xe further too. At 1.3GHz its way under what Icelake uses for power.

According to SharkBay, there's a Tigerlake-H35. That's the same chip used for U chips but set at 35W. Maybe that'll use Iris Xe Max with 1.5-1.65GHz GPU clocks?
I am sorry, can we just take a step back for a moment and appreciate how far tech has come since the Pentium 4?

A Pentium 4 at 4ghz would have a 115 watt TDP and it sucked a ton of juice. A “mini Pentium 4” at 4ghz made from a Tiger Lake core would use 10-15W. That is insane.

CPU+GPU load and long term comparisons:

The Acer Swift 5 using Tigerlake sustains 1.5GHz CPU and 1GHz+ GPU at 17W. The best-in-class Icelake device is the Microsoft Surface Laptop 3. That device does 1.3GHz CPU and 850MHz GPU but at 25W.

At 15W, Icelake drops to 0.9-1GHz CPU and 600-700MHz GPU. Remember, that's with a CPU benchmark and GPU benchmark running together.

In Cinebench R20 MT, Tigerlake @ 17W can maintain 1680 points. Surface Laptop 3 @ 25W with 1065G7 is at 1400 points.
I am interested in worst case scenarios. Prime95, small FFTs, locked to 15W. Renoir ends up between 1-1.2ghz. Does Tiger Lake fair better? I imagine so.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Pushing it way past reasonable limits and using clocks to compensate everything else was the ideology of Netburst.

I am interested in worst case scenarios. Prime95, small FFTs, locked to 15W. Renoir ends up between 1-1.2ghz. Does Tiger Lake fair better? I imagine so.

Is that using CPU only load?
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
Notebookcheck's Tigerlake Iris Xe gets 1778 points in 3DMark Time Spy. 85% higher than Gen 11, and 33% higher than 4800U at 25W with LPDDR4x.

Did some analysis based on the review I last linked. Time Spy seems to be a good indicator of performance at this level. The Swift 1165G7 system is underperforming by 10-15% in Alien vs. Predator game compared to MX350 considering Time Spy scores, but there will be implementation differences.

Considering how balanced the Microsoft Surface Laptop 3 is, I'd love to see what they got out of Tigerlake.

A bad implementation is the Dell XPS 13. The MS SL3 is 20-30% faster in most games, and doesn't drop performance under battery, while XPS will drop it by 30%. XPS also requires a 46W PL1 setting under AC to achieve those figures, while SL3 remains at 25W.

Honestly I don't think a drop from 28W to 25W will affect it a whole lot. The GPU is quite power efficient. The higher TDP might impact it in CPU limited situations by 5-10%, but this is in games that will be getting 70-80 fps.
Those number speak to the situation for AMD that even the 80EU parts from Tiger Lake will produce gpu performance that is at least as fast as their top performing 8CU Vega mobile solution at similarly configured power limits.

Unless Lucien has remarkably better thermals, and Cezanne clocks those VEGA CUs at much higher levels, the next year or so is Intel's mobile market to loose with respect to performance in this sector. From what we're seeing on the CPU side, Tiger Lake is significantly better in single thread than any of the 4xxx series mobile chips, and it takes the full house 4800u to significantly beat it in multi-thread scenarios as the 4600u and 4700u aren't appreciably better or worse than the Tiger Lake parts in their MT scores.

I guess, at this point, what remains to be seen is volume numbers. How many chips can both manufacturers get out the door ad how many can the device manufacturers get into their products and shipped? Also, when shopping the remaining Ice Lake inventory against the new Tiger Lake parts, it's interesting to note that the bottom tier Tiger Lake I3-1125G4 is essentially as good or better than any of the Ice Lake I5s and even rivals the I7-1060/65 in many ways.
 
  • Like
Reactions: clemsyn

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
@LightningZ71

As laptop OEMs have shown us in the past, they'll louse up units with dGPUs even when there are appreciable gains being made in iGPUs. Intel has more pull with OEMs for now, so they can get configs tweaked a little more than AMD could (especially in the past). But I guarantee you that we'll see TigerLake laptops with pointless dGPUs, you just wait.

I don't expect that much of the market to be defined by who has the best mobile iGPU.

As far as CPU performance goes, Renoir is already a bit long in the tooth. TigerLake has new competition on the horizon.