News Intel 11th Gen Tiger Lake CPU Obliterates Ice Lake Chip

gg83

Honorable
Jul 10, 2015
136
18
10,585
0
So that much L3 cache makes up for the lower core count? Or is it a combination of the new architecture plus extra cache? I always wondered why they don't use a "ton" of cache
 

JayNor

Prominent
May 31, 2019
138
26
610
0
Any sightings yet of a Tiger Lake-H 8 core part? I saw it described on wccftech, but haven't seen benchmark leaks as we see for the Tiger Lake-U parts.
 

JayNor

Prominent
May 31, 2019
138
26
610
0
One of the wccftech articles on Tiger Lake-H includes this statement:

"Tiger Lake-H CPUs would additionally feature Two-Level Memory (2LM) and SGX (Software Guard Extensions)."

I understand the 2LM feature to be part of the support required for Optane DIMMs. Has anyone seen explicit mention of Optane DIMM support for Tiger Lake-H? That would be an interesting feature for a low power desktop chip.
 

bit_user

Splendid
Ambassador
The design helps push the Renoir APU ahead of the Tiger Lake part by up to 31.8%. Some might argue that the margin is underwhelming when you consider the fact that the Ryzen 7 4800U has double the number of CPU cores as the Core i7-1065G7.
We don't know how linearly that benchmark scales.

More importantly, consider that the 4800U is also constrained by a 15 W envelope, which is why its base clock is so much lower.

In fact, if you multiply the cores by the base clock, the 4800U has 14.4 core GHz and the i7-1165G7 comes in at 11.2. Divide those and you get only a 28.6% advantage for the 4800U. So, if anything the 4800U seems to perform better than expected! Since it's probably not accomplishing that through superior IPC, I'd assume it's still managing to run above its base clocks, even with all cores/threads cranking.
 

bit_user

Splendid
Ambassador
I always wondered why they don't use a "ton" of cache
Cache costs money (via more die area) and burns power. The performance benefits it confers tend to be logarithmic, although there are certain to be discontinuities for some workloads.

Adding cache is a reliable but somewhat costly way to gin up performance by a couple %.
 
Reactions: gg83

PCWarrior

Distinguished
May 20, 2013
97
26
18,560
0
Geekbench 5 results:
Here is Tigerlake i7 1165G7 in Acer Swift SF314: It does ST:1486, MT: 5827
Here is a Ryzen 7 4700U in Acer Swift SF314: It does ST:1068, MT: 4637

So basically Tigerlake is running circles around Renoir: 39.1% higher single threaded performance and 25.6% higher multi-threaded performance (even with half the physical cores – yes with HT vs without SMT but it is still like comparing a 5.5c5.5t vs a 8c/8t and the former winning by 25%)
 

bit_user

Splendid
Ambassador
Geekbench 5 results:
Here is Tigerlake i7 1165G7 in Acer Swift SF314: It does ST:1486, MT: 5827
Here is a Ryzen 7 4700U in Acer Swift SF314: It does ST:1068, MT: 4637
Thanks for posting. That's indeed interesting.

With laptops, I guess it's reasonable to assume they were plugged-in, but I do worry about things like that, ambient temperature, unobstructed airflow, and whether both had dual-channel memory configurations (note the Intel laptop has 2x the RAM).
 

PCWarrior

Distinguished
May 20, 2013
97
26
18,560
0
I do worry about things like that, ambient temperature, unobstructed airflow, and whether both had dual-channel memory configurations (note the Intel laptop has 2x the RAM).
For the 4700U it was dual channel – it is stated in the benchmark report itself (look again here) under the memory information (channels: 2). If anything it is for the Intel system (look again here) where we don’t have this information – although since it is 16GB it probably was 2x8GB instead of 1x16GB.

I think it is reasonable to assume that these benchmarks were run on a table/desk and not wrapped up in a blanket or something obstructing airflow in different ways. They are in the same chassis so identical in terms of thermal/cooling solutions (ventilation, heatpipes, fan). As for ambient temperature we can reasonably assume they were both tested in a room with humanly tolerable temperatures.

Besides we have more similar results anyway. There are two more 1165G7 cpus in Acer Swift SF314 laptops (see here, ST:1508, MT: 5811 and here, ST:1451, MT: 5890), performing similarly/slightly better than the result I quote above (ST:1486, MT: 5827). Meanwhile we have one more test of a 4700U in an Acer Swift SF314 laptop (see here (ST:1070, MT: 4723) ) doing similarly to the scores I quoted above (ST: 1068, MT 4637).
 
Reactions: bit_user

bit_user

Splendid
Ambassador
I think it is reasonable to assume that these benchmarks were run on a table/desk and not wrapped up in a blanket or something obstructing airflow in different ways.
Or on somebody's bed. I'd look for a couple more Ryzen benchmarks, just to be sure there's nothing weird about that sample.

They are in the same chassis so identical in terms of thermal/cooling solutions (ventilation, heatpipes, fan).
Should be, but it's not a 100% guarantee. Especially if the Intel chip is the higher-end option. Though, I'd agree that the same chassis has the best chance of being identical.

As for ambient temperature we can reasonably assume they were both tested in a room with humanly tolerable temperatures.
My office (an upstairs bedroom) gets up to 85 F, in the summer (well tolerable, if the air is dry and I'm using a fan). In an air conditioned lab, where the Intel machine is likely being tested, it could be as cool as 65 F.

Besides we have more similar results anyway. There are two more 1165G7 cpus in Acer Swift SF314 laptops (see here, ST:1508, MT: 5811 and here, ST:1451, MT: 5890), performing similarly/slightly better than the result I quote above (ST:1486, MT: 5827). Meanwhile we have one more test of a 4700U in an Acer Swift SF314 laptop (see here (ST:1070, MT: 4723) ) doing similarly to the scores I quoted above (ST: 1068, MT 4637).
Thanks for that. Tiger Lake looks to be an exciting launch, then.
 

ASK THE COMMUNITY