Intel released a large amount of Ice Lake news today, including an increase in performance data on the new CPU. ExtremeTech did not realize that Intel had conducted a test event in which reporters from various sites were invited to evaluate Ice Lake under controlled conditions and did not include this data in our initial coverage today. We have contacted Intel to clarify the situation, taking into account that we were present on Architecture Day last winter to inform about Ice Lake's CPU architecture and we have covered the research and development of Intel's foundry during The last eight years.
Now that we really know what CPU performance is like, we have a much better basis for discussing it in relation to Intel Whiskey Lake. Our sister site, PCMag has made an exhaustive comparison of Ice Lake with Whiskey Lake, with Core i7-8565U represented in various form factors and systems from several original equipment manufacturers. Actually, that is very useful because it shows how much distance there is between the laptop and how important is the proper (and thermal) test.
We will cite benchmarks of the PCMag article and we strongly recommend that you read the full story for an in-depth analysis of the publication. Let's start with Cinebench R15:
From good to early, we can see that the ice lake has some problems in the 15W envelope. The fact that the performance of the single-threaded CPU increases by 1.22x when it is given room to breathe in a 25W design is proof that the power consumption of the CPU is very heavy for the core. There is only 5 percentage spread between Core i7-8565U machines in regards to a single thread. When we switch to multiple threads, we give CPU 1.66x more thermal margin resulting in improved performance of 1.33x. Comparing 15W to 15W, older Intel CPUs are all faster, especially HP Envy 13.
In a 25 W configuration, ICL won the overall benchmark, but only 5 percent faster than HP Envy 13. Ryzen 5 2500U exceeded (PCMag does not have Ryzen 7 to compare with the updated APU 3000 series in cellular systems). )
POV-RAY shows some very interesting performance figures, partly because they are completely different from the Handbrake distribution. Envy cell 13, which is the fastest Core i7-8565U in Cinebench R15, is the slowest system in Handbrake (apart from Pentium Gold, which is not really calculated for our needs). Zenbook 13 is 20 percent faster than the 15W ICL test bench, although system performance is equivalent to Specter X360 and Envy 13. Giving a 25 W CPU to play instead of 15 W increases the performance by approximately 24 percent, which allows ICL to beat its competitors.
There are other results for the CPU side available in PCMag and I will look at them for a more complete picture. What we see together is that the 15 W power envelope is suitable for the 10th generation CPU family. Sometimes, the ICL is slightly faster than the 14 nm Whiskey Lake CPUs, sometimes slower, but we don't see much evidence of an increase in envelopes of lower power.
However, at the same time, we also see substantial variations in the results of the Core i7-8565U of 14 nm. This is not surprising; Intel began giving OEMs more freedom to design SKU when Core M debuted, but not all systems were created in the same way. Certain laptops may feel faster than others in certain circumstances. Recently we talked about how an increase in the variation in silicon performance explains many of AMD's decisions about 7 nm products and Ryzen 7 the company This is a clearly different variation classification, but that's really the point. Silicone companies have begun to design several variants in many ways because simply removing them has proved very expensive or very impossible.
It addresses the components of the Ice Lake CPU. What about the GPUs? Here, the news is much more positive.
On the rise of La Tomb Raider Low, ICL can keep 40 fps at 1366 × 768 and 26 fps at 1920 × 1080. Interestingly, it gives the system more room to reduce the score, does not increase to 768p and makes it constant at 1080p. AMD Vega 8 the lower classes do not compete well here, and meanwhile Vega 11 will provide an additional margin of GPU, it is likely that AMD does not completely close the gap. Only laptops equipped with MX150 and MX250, with Nvidia GPU, beat Intel integrated graphics.
Tom Clancy's Rainbow Six: Siege is very impressive, with the same performance between a 15W and 25W CPU. Once again, only MX150 and MX250 outperform Intel's integrated performance. AMD Vega 8 produces a performance that can be played at 1366 × 768 but does not reach the minimum threshold of 30 fps that we consider the minimum for 1080p games.
All GPU numbers basically follow this pattern. You will see that AMD survives better in some than in others, but Intel is ahead of everything. Vega 11 will improve this result, but it may not be enough to change the results in most games.
Implications and conclusions
In our previous report written today, I am implying that one of the reasons for the smaller Intel CPU clock could be because Intel uses a greater amount of TDP to deliver GPU performance. While it may be true, the performance pattern of 15W-25W is different for CPUs compared to GPUs. Moving from 15W to 25W almost always be Improve CPU performance. Switching from 15W to 25W improves the performance of synthetic GPU benchmarks in ICL, but has a weaker impact on the actual game. Only World of Tanks in Core seems to be responding strongly to the additional room for maneuver of TDP, which indicates that in most cases, additional power does not enter the GPU; It is used to speed up the CPU.
The benefits for Ice Lake in relation to Whiskey Lake are quite anemic, although this can vary dramatically depending on the Whiskey Lake system you have now. When there is a 10-15 percent difference between several systems that are equipped with the same processor, it will obviously have an impact on the comparison of ICL. In general, we say that Lake of Ice is comparable to Lake of Whiskey: sometimes faster, sometimes slower, but rarely dramatically distinguished in one way or another.
The GPU improvements, on the other hand, are huge. Assuming Ryzen 7 3700U and 3500U It is a relatively modest improvement over its predecessor, AMD must have a 7 nm APU in the market to use ICL. We have no time limit when that can happen. Of course, AMD is currently focusing on desktop computers and server rooms, which means we don't even know when Intel's 10 nm silicon will face AMD 7 nm in the market.
The third pillar is energy consumption and battery life, and we still don't know how ICL compares this metric; Intel prohibits testing laptop samples for such things. Currently, Ice Lake offers massive increases in one area, receives small gains from small losses in another area and offers an unknown level of increase in the third area. Players who want some ability to play above the lights should be the main beneficiaries of the improvements we have seen so far. If this performance survives, AMD needs to push Intel to 7nm or see that its long dominance in the integrated cellular GPU market eventually falls, which is not really a sentence I thought I once wrote.