Our first Ice Lake Core i7-1065G7 benchmarks are in and they outgun the AMD Ryzen 7 3700U
Intel is about to launch its 10th generation Ice Lake processors later this year to succeed the current 9th generation Whiskey Lake-U series. Unlike the minor tiptoe in performance from 8th gen Kaby Lake-R to 9th gen Whiskey Lake-U, the gen-to-gen performance gain with Ice Lake promises to be a larger jump.
Intel recently gave us the opportunity to test the upcoming 10 nm Core i7-1065G7 in both 15 W and 25 W TDP levels so we can see what kind of performance gains 25 W can bring to the table and how the new CPU compares to last generation options. OEMs will offer two versions of the Core i7-1065G7 at retail depending on the chassis design: a 15 W SKU and a 25 W SKU. This particular processor is expected to supplant the 14 nm Core i7-8565U and compete directly with AMD's 14 nm Zen+ Ryzen 7 3700U especially in graphics performance where Intel had been stagnant.
Before we begin, it's important to note that these test samples are preliminary Intel machines not meant to represent any final OEM designs. Retail laptops with the same Core i7-1065G7 CPU from Lenovo, Dell, HP, Acer, and others will indubitably have different levels of performance due to differences in chassis designs and cooling. Thus, our data and findings below only paint a broad picture of what we can expect from the Core i7-1065G7. It won't be until we can personally test a handful of retail units that we can comfortably give a more detailed verdict on this particular Core i7 Ice Lake processor.
In addition to the above forewarning, Intel had some limitations as to what kinds of tests we were allowed to run on these early Ice Lake machines. We are therefore unable to provide a few of our usual measurements like power consumption, temperature, and battery life until proper retail units become available.
For a technical overview of Intel Ice Lake and the technology behind it, see here. We will be jumping straight into the raw performance levels of the 15 W and 25 W Core i7-1065G7 to see how the new CPU stacks up against the Core i7-8565U, i7-8550U, and AMD's Ryzen 7 3700U.
SDS SYSTEM Configuration |
|
CPU | Intel® Core™ i7-1065G7 (15 W and 25 W) |
Memory | 8GB LPDDR4X-3733 (dual channel) |
SSD | Intel® SSD Pro 7600p Series 256GB |
Graphics | Intel® Iris® Plus Graphics (64 EUs) |
Graphics Driver | 26.20.100.7010 |
Display | 13.3" 4K Panel 3840x2160 |
Wireless | Intel® Wi-Fi 6 AX201 2x2 |
OS | Windows 10 Pro Build 1903 |
Processor Performance - Nearly tied with the Core i5-9300H
Starting with CineBench R15, we can observe that the 25 W Core i7-1065G7 offers a 44 percent performance boost over its 15 W Core i7 version in multi-threaded loads. The 25 W Core i7 results are still about 11 to 25 percent higher than the Ryzen 7 3700U and about 32 percent higher than the average Core i7-8550U in our database taken from a sample size of 76 laptops.
The 15 W Core i7-1065G7, however, is less impressive as results are slightly below the average Core i7-8550U and about on par with the older Ryzen 7 2700U in multi-threaded loads. Single-threaded performance is actually quite close to the 25 W i7-1065G7 which is where the AMD alternatives really struggle.
More impressively, the 25 W i7-1065G7 is neck-to-neck with the Core i5-9300H according to CineBench R20. When considering that the latter is a 45 W CPU commonly found on gaming laptops, this is a strong showing for the 10th gen 25 W Core i7.
Cinebench R15 | |
CPU Single 64Bit | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
HP Omen 15-dc1020ng | |
Microsoft Surface Pro (2017) i7 | |
Lenovo ThinkPad T480-20L6S01V00 | |
Average Intel Core i7-8550U (108 - 172, n=76) | |
Lenovo ThinkPad 25 | |
Lenovo IdeaPad S540-14API | |
Lenovo Yoga C930-13IKB | |
HP Envy x360 15-bq102ng | |
Lenovo Ideapad 720S-13ARR | |
CPU Multi 64Bit | |
Intel NUC8i7BE | |
HP Omen 15-dc1020ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Lenovo IdeaPad S540-14API | |
Lenovo ThinkPad T480-20L6S01V00 | |
HP Envy x360 15-bq102ng | |
Lenovo Yoga C930-13IKB | |
Average Intel Core i7-8550U (301 - 761, n=78) | |
Lenovo Ideapad 720S-13ARR | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Microsoft Surface Pro (2017) i7 | |
Lenovo ThinkPad 25 |
Cinebench R11.5 | |
CPU Single 64Bit | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Microsoft Surface Pro (2017) i7 | |
Average Intel Core i7-8550U (1.43 - 1.95, n=32) | |
Lenovo IdeaPad S540-14API | |
Lenovo Yoga C930-13IKB | |
Lenovo ThinkPad 25 | |
HP Envy x360 15-bq102ng | |
Lenovo ThinkPad T480-20L6S01V00 | |
CPU Multi 64Bit | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Lenovo IdeaPad S540-14API | |
Intel NUC8i7BE | |
Lenovo ThinkPad T480-20L6S01V00 | |
HP Envy x360 15-bq102ng | |
Average Intel Core i7-8550U (4.38 - 8.56, n=32) | |
Lenovo Yoga C930-13IKB | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Microsoft Surface Pro (2017) i7 | |
Lenovo ThinkPad 25 |
Cinebench R10 | |
Rendering Multiple CPUs 32Bit | |
HP Omen 15-dc1020ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel NUC8i7BE | |
Lenovo Yoga C930-13IKB | |
Lenovo IdeaPad S540-14API | |
Average Intel Core i7-8550U (11694 - 24180, n=31) | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
HP Envy x360 15-bq102ng | |
Microsoft Surface Pro (2017) i7 | |
Lenovo ThinkPad 25 | |
Lenovo Ideapad 720S-13ARR | |
Rendering Single 32Bit | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Intel NUC8i7BE | |
HP Omen 15-dc1020ng | |
Average Intel Core i7-8550U (4202 - 8211, n=31) | |
Microsoft Surface Pro (2017) i7 | |
Lenovo Yoga C930-13IKB | |
Lenovo ThinkPad 25 | |
Lenovo IdeaPad S540-14API | |
HP Envy x360 15-bq102ng | |
Lenovo Ideapad 720S-13ARR |
Cinebench R20 | |
CPU (Single Core) | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
HP Omen 15-dc1020ng | |
Average Intel Core i7-8550U (383 - 394, n=2) | |
Lenovo IdeaPad S540-14API | |
CPU (Multi Core) | |
HP Omen 15-dc1020ng | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Lenovo IdeaPad S540-14API | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (1118 - 1138, n=2) |
Benchmarks like CineBench and GeekBench are relatively poor indicators of CPU performance since the tests are short in duration. Longer benchmarks like wPrime, Blender, and HWBOT are better at gauging CPU performance over time to account for any throttling. The 25 W Core i7-1065G7 is still able to consistently outperform the 15 W version by about 30 to 40 percent in each of the three aforementioned benchmarks. Results are about 10 percent slower than the Coffee Lake-U Core i7-8559U in each case.
Interestingly, x265 4K encoding on the 15 W Core i7 is on par with the Ryzen 7 3700U while rendering on Blender is 40 percent slower on the Intel.
Running CineBench R15 Multi-Thread in a loop shows how well the processors can perform over time as illustrated by our graph below. The 25 W Core i7 is able to maintain its lead over the Ryzen 7 3700U even after accounting for throttling. However, throttling results will always vary between chassis designs and so OEMs could limit Core i7 performance even further if they wish for smaller chassis designs.
wPrime 2.10 - 1024m | |
Lenovo ThinkPad 25 | |
HP 14-dk0008ng | |
Microsoft Surface Pro (2017) i7 | |
HP Spectre x360 13t-ae000 | |
HP Envy x360 15-bq102ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (258 - 435, n=11) | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel NUC8i7BE |
TrueCrypt | |
Serpent Mean 100MB | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (0.357 - 0.469, n=6) | |
HP Spectre x360 13t-ae000 | |
HP 14-dk0008ng | |
HP Envy x360 15-bq102ng | |
Microsoft Surface Pro (2017) i7 | |
Lenovo ThinkPad 25 | |
Twofish Mean 100MB | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Average Intel Core i7-8550U (0.583 - 0.793, n=6) | |
HP Spectre x360 13t-ae000 | |
HP 14-dk0008ng | |
HP Envy x360 15-bq102ng | |
Microsoft Surface Pro (2017) i7 | |
Lenovo ThinkPad 25 | |
AES Mean 100MB | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Average Intel Core i7-8550U (3.3 - 4.6, n=6) | |
HP Spectre x360 13t-ae000 | |
HP 14-dk0008ng | |
HP Envy x360 15-bq102ng | |
Microsoft Surface Pro (2017) i7 | |
Lenovo ThinkPad 25 |
Blender - v2.79 BMW27 CPU | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (841 - 937, n=4) | |
Apple MacBook Pro 13 2018 Touchbar i5 | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Lenovo IdeaPad S540-14API | |
Intel NUC8i7BE | |
Intel NUC8i7BE | |
HP Omen 15-dc1020ng |
WinRAR - Result | |
Intel NUC8i7BE | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (3847 - 5287, n=7) | |
HP Spectre x360 13t-ae000 | |
HP Envy x360 15-bq102ng | |
HP 14-dk0008ng |
7-Zip 18.03 | |
7z b 4 | |
Intel NUC8i7BE | |
Intel NUC8i7BE | |
Apple MacBook Pro 13 2018 Touchbar i5 | |
HP Omen 15-dc1020ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Lenovo IdeaPad S540-14API | |
Average Intel Core i7-8550U | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
HP 14-dk0008ng | |
7z b 4 -mmt1 | |
Intel NUC8i7BE | |
Intel NUC8i7BE | |
HP Omen 15-dc1020ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Average Intel Core i7-8550U | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Apple MacBook Pro 13 2018 Touchbar i5 | |
Lenovo IdeaPad S540-14API | |
HP 14-dk0008ng |
HWBOT x265 Benchmark v2.2 - 4k Preset | |
Intel NUC8i7BE | |
Intel NUC8i7BE | |
HP Omen 15-dc1020ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Apple MacBook Pro 13 2018 Touchbar i5 | |
Average Intel Core i7-8550U (3.77 - 4.43, n=3) | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Lenovo IdeaPad S540-14API | |
HP 14-dk0008ng |
ComputeMark v2.1 | |
1024x600 Normal, QJuliaRayTrace | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (302 - 1416, n=9) | |
HP 14-dk0008ng | |
Lenovo ThinkPad 25 | |
Microsoft Surface Pro (2017) i7 | |
Intel NUC8i7BE | |
HP Envy x360 15-bq102ng | |
1024x600 Normal, Mandel Scalar | |
HP 14-dk0008ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
HP Envy x360 15-bq102ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (181 - 908, n=9) | |
Lenovo ThinkPad 25 | |
Intel NUC8i7BE | |
Microsoft Surface Pro (2017) i7 | |
1024x600 Normal, Mandel Vector | |
HP Envy x360 15-bq102ng | |
HP 14-dk0008ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Average Intel Core i7-8550U (163 - 948, n=9) | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Lenovo ThinkPad 25 | |
Microsoft Surface Pro (2017) i7 | |
Intel NUC8i7BE | |
1024x600 Normal, Fluid 2DTexArr | |
Average Intel Core i7-8550U (105 - 524, n=9) | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Lenovo ThinkPad 25 | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
HP 14-dk0008ng | |
HP Envy x360 15-bq102ng | |
Intel NUC8i7BE | |
Microsoft Surface Pro (2017) i7 | |
1024x600 Normal, Fluid 3DTex | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (86 - 518, n=9) | |
Lenovo ThinkPad 25 | |
Intel NUC8i7BE | |
HP 14-dk0008ng | |
HP Envy x360 15-bq102ng | |
Microsoft Surface Pro (2017) i7 | |
1024x600 Normal, Score | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
HP 14-dk0008ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Average Intel Core i7-8550U (836 - 4314, n=9) | |
HP Envy x360 15-bq102ng | |
Lenovo ThinkPad 25 | |
Intel NUC8i7BE | |
Microsoft Surface Pro (2017) i7 |
* ... smaller is better
Prime95 Stress - Faster Clock Rates for Longer on the 25 W SKU
We stress the laptop with Prime95 to observe clock rate behavior when at 100 percent CPU utilization. Intel says that the 25 W SKU will not necessarily run at higher clock rates than the 15 W SKU, but it will be able to maintain higher clock rates for longer than the 15 W SKU. We're able to confirm this with Prime95; our screenshots below reveal that the 25 W SKU is able to maintain a clock rate of 3.3 GHz for about two times longer (~6 seconds vs. 14 seconds) than the 15 W SKU during the beginning of the stress test. Thereafter, clock rates stabilize at 2.5 GHz compared to just 2 GHz or lower on the 15 W SKU. These observations confirm our CineBench R15 loop test results from above.
While our test units lack the feature, Intel says retail SKUs will have AI-driven Adaptix Dynamic Tuning that should result in more gradual changes to clock rates over time depending on the application and overlying cooling solution. When encoding video, for example, users should see more gradual drops in clock rates over time instead of the steep slope we've recorded here. The technology would work similarly to Dell's existing Dynamic Power Policy for the XPS laptop series.
Graphics Performance - More of a Mixed Bag
Next up is GPU performance. Intel has been phoning it in for the past few years when it comes to the UHD Graphics series as it has barely changed since the early Skylake days. It wasn't until the advent of the mobile Ryzen Vega series that finally pushed Intel back to the drawing board for its integrated Ice Lake GPU. The result is a huge gain in graphics performance over the current UHD Graphics 620 series by as much as 110 percent and 160 percent for the 15 W Core i7-1065G7 and 25 W Core i7-1065G7, respectively, according to 3DMark 11. This puts the 15 W Intel Core i7 on par with the AMD RX Vega 8 and 25 W Core i7 in between the GeForce MX130 and GeForce MX150.
Unfortunately, raw 3DMark benchmarks almost never tell the whole story especially for GPUs outside of Nvidia. We tested five games on the Core i7-1065G7 ranging from undemanding to very demanding: Rocket League, Bioshock Infinite, Overwatch, Witcher 3, and Shadow of the Tomb Raider. Results are on par with the RX Vega 10 at worse and, depending on the title and laptop model, about 20 percent faster or more at best. Expect gaming performance to vary considerably in retail SKUs as OEMs must balance the 25 W power envelope between the CPU, GPU, and cooling solution. In this case, a dedicated GeForce MX130 still offers more consistent and reliable gaming performance.
An interesting observation is that real-world gaming performance between the 15 W and 25 W Core i7 test samples is almost the same despite what 3DMark is telling us. The 25 W Core i7 is just a few percentage points faster than the 15 W Core i7 on average when running the few games we've tested. Aside from the early unoptimized graphics drivers, we suspect that the CPU may be a bottleneck to the GPU or vice versa when running real-world titles to explain the similar gaming results between them.
We also noticed that games would stutter more frequently than usual even after taking into account the uneven frame pacing. We're crossing our fingers that this will be addressed once retail models become available or that it was simply a fluke on the sample test unit.
Shadow of the Tomb Raider - 1280x720 Lowest Preset | |
Average NVIDIA GeForce MX150 (30 - 50, n=5) | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Acer Nitro 5 | |
Intel NUC8i7BE | |
Apple MacBook Pro 13 2018 Touchbar i5 | |
Dell XPS 13 9380 2019 |
Overwatch - 1280x720 Low (Render Scale 100 %) AA:FX AF:1x | |
Average NVIDIA GeForce MX150 (60 - 174, n=6) | |
Apple MacBook Pro 13 2018 Touchbar i5 | |
HP Envy x360 15-bq102ng | |
Intel SDS Core i7-1065G7 Preliminary Sample 25 W | |
Lenovo ThinkPad X1 Carbon 2018-20KGS03900 | |
Intel SDS Core i7-1065G7 Preliminary Sample 15 W | |
Microsoft Surface Pro (2017) i7 | |
Acer Nitro 5 |
Preliminary Verdict - Excellent Start With More Work To Go
The jump from Kaby Lake Core i7-7500U to Kaby Lake-R Core i7-8550U brought a large boost to CPU performance while GPU gains were minimal. In a similar manner, we now expect the jump from Whiskey Lake-U Core i7-8565U to 15 W Ice Lake-U Core i7-1065G7 to bring a large boost to GPU performance with only minimal raw CPU gains. Intel has managed to squeeze significantly more GPU power over the aging UHD Graphics 620 to be competitive against the <35 W AMD RX Vega 10 all while staying within a smaller 15 W TDP envelope — an impressive feat on its own right.
If you're looking for faster CPU performance, however, only the 25 W Core i7-1065G7 can get the job done. We're seeing a 40 percent gain in multi-threaded CPU performance over the 15 W version. Interestingly, real-world gaming performance between the 15 W and 25 W versions is almost the same in many cases meaning that the jump from 15 W Core i7-1065G7 to 25 W Core i7-1065G7 benefits the CPU much more than the integrated Iris GPU. The biggest reason to own the 25 W Core i7 SKU over the 15 W Core i7 SKU would be for the higher processor performance and not its GPU performance as a result.
Finally, it's worth noting that the upcoming 25 W Core i7-1065G7 performs very similarly to Intel's existing Coffee Lake-U 28 W Core i7-8559U. This uncommon processor can be found on the Intel NUC8i7BEH mini PC and it gives us a very close approximation to the raw horsepower of the impending 25 W Core i7-1065G7. Of course, Ice Lake will integrate more QoL features outside of just performance not found on Coffee Lake-U including Thunderbolt 3, Wi-Fi 6, Adaptix, and AI inference technology that we will test in more depth once retail units become available. The extent of how AI features and deep learning can improve everyday workloads on Windows remains to be seen.
Our first impressions of the Core i7-1065G7 are positive overall from a pure performance standpoint. The 15 W SKU offers similar CPU performance to the Core i7-8565U and slightly better GPU performance than the more demanding AMD RX Vega 10 while the 25 W SKU is within 10 percent of the Core i5-9300H in raw processing power. The GPU still needs some work for more consistent performance across games that will surely be improved over time with driver updates.
Expect the first Ice Lake laptops to launch in Q4 2019 as part of Intel's lineup of first generation Athena models.