By: Tiger Lake Leaks (tiger.delete@this.lake.leak), April 14, 2020 9:30 am
Room: Moderated Discussions
As many of you know, Intel's 10nm process suffered from two big problems. First the yield was poor. Secondly, the frequency was very underwhelming compared to the improved 14nm process.
The long delayed Cannonlake was the first evidence of these problems.
Even Icelake was a disappointing release, with significantly worse clock frequency for the CPU cores, although impressive graphics performance.
To illustrate these problems, here is the best 14nm and 10nm processors from Intel. Both are 4-cores, 8-threads, and 15W, and also priced at about $400.
Bottom line, the 10nm processor has about 20-30% worse CPU frequency, but about 2-2.5X better throughput on the GPU.
Intel didn't even release a 45W Icelake-H model, since it wouldn't have been remotely competitive.
In the last few weeks, there have been a variety of leaks of Tiger Lake engineering samples suggesting clocks ranging from 2.3-3GHz. If true, this would mean that Intel has tweaked their 10nm process to finally deliver good clock frequency. We might finally see competition again!
The long delayed Cannonlake was the first evidence of these problems.
Even Icelake was a disappointing release, with significantly worse clock frequency for the CPU cores, although impressive graphics performance.
To illustrate these problems, here is the best 14nm and 10nm processors from Intel. Both are 4-cores, 8-threads, and 15W, and also priced at about $400.
- Process: 14nm vs. 10nm
- Base CPU clocks: 1.8GHz vs. 1.3GHz
- Peak CPU clocks: 4.6GHz vs. 3.9GHz
- GPU EUs: 24 EUs vs. 64 EUs
- GPU clocks: 0.3/1.15GHz vs. 0.3/1.1GHz
- Memory: DDR4-2400/LPDDR4-2133 vs. DDR4-3200/LPDDR-3733
Bottom line, the 10nm processor has about 20-30% worse CPU frequency, but about 2-2.5X better throughput on the GPU.
Intel didn't even release a 45W Icelake-H model, since it wouldn't have been remotely competitive.
In the last few weeks, there have been a variety of leaks of Tiger Lake engineering samples suggesting clocks ranging from 2.3-3GHz. If true, this would mean that Intel has tweaked their 10nm process to finally deliver good clock frequency. We might finally see competition again!
| Topic | Posted By | Date |
|---|---|---|
| Tiger lake leak, Intel 10nm fixed? | Tiger Lake Leaks | 2020/04/14 09:30 AM |
| Tiger lake leak, Intel 10nm fixed? | anon | 2020/04/14 08:08 PM |
| Tiger lake leak, Intel 10nm fixed? | Wes Felter | 2020/04/15 02:46 PM |
| Keep in mind Intel supposedly fixed their 10nm process at the same time TSMC is ramping up 5nm (NT) | anon | 2020/04/23 10:14 PM |
| 28+ W | me | 2020/04/15 07:42 PM |
| 28+ W | David Kanter | 2020/04/17 09:12 AM |
| 28+ W | another anon | 2020/04/17 11:13 PM |
| 28+ W | Dummond D. Slow | 2020/04/18 09:27 AM |
| 28+ W | David Kanter | 2020/04/19 07:27 PM |
| 28+ W | Anon | 2020/04/19 08:47 PM |
| 28+ W | David Kanter | 2020/04/19 09:50 PM |
| 28+ W | anny | 2020/04/20 05:56 AM |
| 28+ W | Anon | 2020/04/20 08:03 AM |
| 28+ W | wumpus | 2020/04/20 02:39 PM |
| EPYC F line | Anon | 2020/04/20 05:35 PM |
| 28+ W | gallier2 | 2020/04/19 11:44 PM |
| 28+ W | Alberto | 2020/04/20 09:07 AM |
| 28+ W | Adrian | 2020/04/20 09:55 AM |
| 28+ W | Anon | 2020/04/20 02:18 PM |
| 28+ W | Anon3 | 2020/04/20 04:02 PM |
| 28+ W | Dummond D. Slow | 2020/04/20 09:05 PM |
| 28+ W | Adrian | 2020/04/20 03:54 AM |
| This was a comparison to put the "high clock" in context | Dummond D. Slow | 2020/04/20 08:55 AM |
