The test data of MediaTek Tianji 9000 was finally lifted on December 20. Now let's take a look at the performance of the first TSMC 4nm process, the X2+A710+A510 with more full specifications and the ten core Mali-G710.

First, reread the specifications of Tianji 9000, TSMC 4nm, 3.05G X2 (1M L2)+3x2.85G A710 (512K L2)+4x1.8G A510 (256K L2), 8M L3 and 6M SLC cache, Mali-G710 MC10， The first 7500Mbps LPDDR5x memory support.

The "enemy in life" Snapdragon 8 Gen 1 is Samsung 4nm, 3G X2+2.5G A710, and L2 cache on both sides is consistent. The L2 cache of Qualcomm A510 small core is unknown, and it uses the core 2+2 structure design with less area, 6M L3+4M SLC, 6400Mbps LPDDR5 memory.

The Tianji 9000 has a higher frequency of super core and large core, higher L3 and SLC caches, and higher memory specifications (although there is no mass production machine using LPDDR5x memory), but the biggest difference may be the process difference between TSMC and Samsung.

The Tianji 9000 is an engineering machine that doesn't even have a graphite heat sink (unfortunately, this brick has a large heat sink space). It has a 12GB LPDDR5 6400 memory+256GB UFS 3.1+LCD screen (claimed to have a 0.6W higher power consumption than the A screen)+native Android 12. GeekBench and Anthare are both offline special versions, but there is no big difference between the subproject and the latest version, which can be compared horizontally.

[GeekBench 5] Arm's X2 super core this time is a model of standing still. Even TSMC+Fage, the single core has not made much progress. The single core has 1264 points, so far, it can not win the A13 on the iPhone 11 Pro Max.

However, in terms of multi-core performance, Tianji 9000 has three 2.85GHz high frequency A710s (higher than most SoCs' super core frequencies), as well as the first big cache in the Android camp. Its multi-core performance is slightly better than A14. While other media teachers' Tianji 9000 testing machine, which has better heat dissipation and physical fitness, has broken the 4400 point mark, but this one in our hands only hovers below 4300 points.

As a sub project of GeekBench 5, you can join the team running of Snapdragon 888 and 888+. Tianji 9000 and Snapdragon 8 Gen 1 are both X2 super cores at 3GHz, and their scores are very close. However, in the sub project, the navigation HTML5、 Image repair and motion structure are weak, ranging from 1% to 6%, while PDF rendering, Gaussian blur, image compression and other items are 10% or more strong.

In the multi-core part, when the total score of Tianji 9000 is 14% strong, the floating point leads by 21%, and the most exaggerated ray tracing and machine learning projects are more than 30% strong. The position of Android CPU boss is solid. However, judging from the score changes, the 2.85GHz UHF A710 of Tianji 9000 (2.5G for Qualcomm) seems to be significantly weaker on HTML5 and Clang.

It should be noted that the early Snapdragon 8 Gen 1 has no advantage over Snapdragon 888+in terms of multi-core performance, and the subsequent manufacturers will come to help.

[GFXBench] This part introduces the Snapdragon 8 Gen 1 QDR data unique to Teacher Feiwei. This year, Qualcomm's Adreno GPU has regained its former glory and finally can compete with Apple's A15.

The GPU of Tianji 9000 is one gear weaker than Snapdragon 8 Gen 1, but far stronger than expected. In the high pressure Aztec test, it even slightly beats the youth version A15 of iPhone 13, close to the Max-Q version A15 of iPad mini 6, 4% less than the iPad mini 6, 11.4% less than the Snapdragon 8 Gen 1 QRD, and 18.1% less than A15; In the low pressure Manhattan and Manhattan 3.1, we can catch up more closely.

[Antulu] Here we borrow the data from the 8 Gen 1 refrigerator of WHYLAB, and Fage is still passive cooling. The CPU and MEM (memory+flash memory) of Tianji 9000 are 13% and 12% stronger respectively, the UX (data security/data processing/strategy games/image processing and I/O performance and other projects) is 2% stronger, and the GPU is 12% weaker, but the total score is 102W. The GPU weight of Entertainment Rabbit is not right (manual funny. jpg).

The remaining PCMark (18162 points) and ETHZ AI benchmark V5 test (1058K) will not be discussed in detail.

According to the AndSPEC06 test that teacher Fei Wei stole, the X2 super nuclear energy efficiency of Tianji 9000 is 49% ahead, and the A710 super nuclear energy efficiency is 40% ahead (even at a higher frequency).

A thousand words come together into two sentences: TSMC Yes! MTK Yes！

Follow our Weibo @ Love Computer

Follow our WeChat official account: playphone

Of course, we also pay attention to our Bilibili account: love computer