Discussion Intel current and future Lakes & Rapids thread

[dullard]

They seem to compare their 28W cpu to 15W 4800U - actual opponent seems to be 4900HS instead. Doesn't 4900HS bring some 30% graphics uplift.....

The intel system was running "AC / Balanced mode" so it was probably running at 28W.

1) The 4800U is configurable to 25W. Intel's benchmarks claim the AMD system was at "highest available performance profile". To me, that sounds like the 4800U may have been running at 25W, not 15W.
2) The 4900HS is 35W.

Thus, the 25W 4800U is slighly closer in power to the tiger lake chip than the 35W 4900HS. But, this is certainly not an apples-to-apples comparison.

 

[Hitman928]

Oh, there's also this from the leaked slide deck apparently:

View attachment 29096

The Renoir laptop appears to be 35W PPT short, 25W PPT long and unspecified STAPM power target.

All I see from that is that the Tigerlake CPU uses way more power on battery than the 4800u, lol. I am quite certain all of that is dependent on how the individual laptop is configured and is a (I'll use the word) strange slide for Intel to use to try and promote their CPU.

 

[eek2121]

The model they got was intended for the Chinese market. They had to do a fresh Windows install so that they could get it into English.

What about the power profile...Do you have any idea how much that skews the results?

 

[Hitman928]

This chip will beat a 6C renoir chip in blender, cinebench, etc.

I very much doubt that but we'll see what the reviews show.

 

[Hitman928]

What about the power profile...Do you have any idea how much that skews the results?

The two laptops are clearly running different power profiles, at least on battery as shown in that Intel AC/DC slide. We'll have to wait for reviewers to get actual models and try to review them against the same/similar models using Renoir or Cometlake to see how TGL really compares.

 

[eek2121]

The two laptops are clearly running different power profiles, at least on battery as shown in that Intel AC/DC slide. We'll have to wait for reviewers to get actual models and try to review them against the same/similar models using Renoir or Cometlake to see how TGL really compares.

Agreed. Using a high performance profile for either laptop means that all benchmarks should be called into question, as the benchmarks end up testing the cooling system and TDP limit more than anything.

 

[TESKATLIPOKA]

It beats a 6 core Renoir. What many people don’t get is that physical core counts don’t matter. At the end of the day single threaded performance is what counts. This chip will beat a 6C renoir chip in blender, cinebench, etc.

Both physical count and single threaded performance matters.
Beat a 6C12T Renoir in Cinebench? That's ~50% difference so how can Tiger Lake with 4C8T beat that with "only" 20-25% better IPC? It would need to have 20% higher clocks during the test to be even or a bit better. Can It do that?

 

[mikk]

This is Tigerlake at 17W: https://ithardware.pl/testyirecenzj...i7_1165g7_willow_cove_i_xe_w_akcji-13113.html

It uses PL1 28W the review says.

the computer used for testing makes use of it to the maximum (i.e. PL1 set to 28 W)

 

[IntelUser2000]

It uses PL1 28W the review says.

It also says the laptop goes to 17W on long loads, which is PL1.

 

[Olikan]

This is Tigerlake at 17W: https://ithardware.pl/testyirecenzj...i7_1165g7_willow_cove_i_xe_w_akcji-13113.html

Loses to desktop Picasso? (IGP gaming)

 

[naukkis]

Loses to desktop Picasso? (IGP gaming)

And loses badly to 4800H. And multicore performance is about 3750H level - at least in that review Tigerlake is not impressive.

 

[tamz_msc]

And loses badly to 4800H. And multicore performance is about 3750H level - at least in that review Tigerlake is not impressive.

Huh, multi-core is much better than 3750H.

 

[Hitman928]

Loses to desktop Picasso? (IGP gaming)

It's desktop Picasso (3400G). According to this video, it's roughly equivalent to a 4800HS with DDR-3200.

Give the 4800HS LPDDR4 and it probably gains a few fps. So, first brush very rough approximation it looks like TGL will probably edge out Renoir in iGPU in most cases, at least at 28W. In that review, TGL does fail hard in some titles though, not sure why or if Renoir will have the same issue (probably not, haven't seen that in any of the Renoir reviews). At 15W I'm not so sure, I haven't seen anything about TGL performance at 15W.

 

[Hitman928]

It also says the laptop goes to 17W on long loads, which is PL1.

Thermal throttling perhaps? I mean they specifically say it is configured for PL1 to be 28 W and then at the end say it reduces to 17W on long loads so effective PL1 is 17W. Given "standard" PL1 is 15W it seems to me more like it is thermal related throttling but I don't speak Polish and have no idea how trustworthy or competent this reviewer is.

 

[IntelUser2000]

=Thermal throttling perhaps? I mean they specifically say it is configured for PL1 to be 28 W and then at the end say it reduces to 17W on long loads so effective PL1 is 17W. Given "standard" PL1 is 15W it seems to me more like it is thermal related throttling but I don't speak Polish and have no idea how trustworthy or competent this reviewer is.

There's no such thing as standard. The manufacturer is free to set the PL1 at any level in between the lower and the higher TDP level. Of course, I've not seen a single manufacturer set it at low TDP levels, since performance is desired.

The previous Swift was set at that mark so I can believe being at 17W.

 

[TESKATLIPOKA]

At notebookcheck there is also a performance comparison of Tiger Lake.
Link
It really does perform on par with 4700U in Cinebench MT, although that is a 28W version against 15W Renoir, but that is still pretty good.

 

[eek2121]

Both physical count and single threaded performance matters.
Beat a 6C12T Renoir in Cinebench? That's ~50% difference so how can Tiger Lake with 4C8T beat that with "only" 20-25% better IPC? It would need to have 20% higher clocks during the test to be even or a bit better. Can It do that?

The all core boost of the i7 is 4.3 Ghz, while the SC boost is 4.8 Ghz. The SC boost of Renoir 4.2 Ghz.

 

[Hitman928]

There's no such thing as standard. The manufacturer is free to set the PL1 at any level in between the lower and the higher TDP level. Of course, I've not seen a single manufacturer set it at low TDP levels, since performance is desired.

The previous Swift was set at that mark so I can believe being at 17W.

Traditionally the U processors were 15W models with cTDP up and down. I put standard in quotes because it seems like Intel is shifting away from this and is just using a cTDP range now but because of the traditional TDP of U based models, people widely still see 15W at the "standard" TDP.

 

[IntelUser2000]

Traditionally the U processors were 15W models with cTDP up and down. I put standard in quotes because it seems like Intel is shifting away from this and is just using a cTDP range now but because of the traditional TDP of U based models, people widely still see 15W at the "standard" TDP.

Right. I'm saying its no longer true anymore. For Icelake "15W" there is a 15W device, a 17W one, an 18W one, 22W, and 25W.

And they can also make it so it adjusts based on whether you want it Silent, or Performance. Even on the fly so when you have it folded as a Tablet its at low TDP, and higher with laptop mode.

 

[mikk]

It's desktop Picasso (3400G). According to this video, it's roughly equivalent to a 4800HS with DDR-3200.

Give the 4800HS LPDDR4 and it probably gains a few fps. So, first brush very rough approximation it looks like TGL will probably edge out Renoir in iGPU in most cases, at least at 28W. In that review, TGL does fail hard in some titles though, not sure why or if Renoir will have the same issue (probably not, haven't seen that in any of the Renoir reviews). At 15W I'm not so sure, I haven't seen anything about TGL performance at 15W.

And also the Tigerlake graphics driver used there is really old. i7-1165G7 should crush Ryzen 4700U, the difference should be bigger there.

 

[Hitman928]

On notebookcheck there is also a performance comparison of Tiger Lake.
Link
It really does perform on par with 4700U in Cinebench MT, although that is a 28W version against 15W Renoir, but that is still pretty good.

28 W vs 15 W is 87% higher TDP. A 4800u at 25W versus TGL at 28W, the 4800u is 54% faster. It's not really close when given similar TDPs.

 

[TESKATLIPOKA]

The all core boost of the i7 is 4.3 Ghz, while the SC boost is 4.8 Ghz. The SC boost of Renoir 4.2 Ghz.

But that 4.3Ghz all core boost is not for the 15W version, for 15W is only 3.7Ghz.

 

[Hitman928]

Right. I'm saying its no longer true anymore. For Icelake "15W" there is a 15W device, a 17W one, an 18W one, 22W, and 25W.

And they can also make it so it adjusts based on whether you want it Silent, or Performance. Even on the fly so when you have it folded as a Tablet its at low TDP, and higher with laptop mode.

Ok, we're on the same page. Still doesn't explain why the reviewer said it was configured for 28W PL1 but then it backed off to 17W on long loads. Either he was mistaken with how it was configured or the laptop was throttling.

 

[IntelUser2000]

Ian says to be branded as Xe, the laptop needs to use dual channel memory. Otherwise, its UHD I guess?

Ok, we're on the same page. Still doesn't explain why the reviewer said it was configured for 28W PL1 but then it backed off to 17W on long loads. Either he was mistaken with how it was configured or the laptop was throttling.

Eh. He made a mistake. I've seen worse mistakes, and from vendors themselves!

 

[TESKATLIPOKA]

28 W vs 15 W is 87% higher TDP. A 4800u at 25W versus TGL at 28W, the 4800u is 54% faster. It's not really close when given similar TDPs.

I was comparing It against 4700U 8C8T which has disabled SMT. If you enable SMT of course the difference will increase. The graph shows 43% difference between 4700U vs 4800U in Cinebench R15.