Skip to main content

Four Intel 10th Gen Comet Lake CPUs Benchmarked Before Release

(Image credit: Shutterstock)

Comet Lake desktop CPUs have been selling ilegally in China since February. A couple of days ago, a user on the Chinese Sohu news portal apparently managed to buy not one, but four Comet Lake-S processors and put them through various benchmarks.

The Intel Core i5-10400, i5-10500, i5-10600K and i7-10700 are the stars of today's Comet Lake show. The tester said the i5-10500 and i5-10600K are engineering samples (ES) , while the i5-10500 and i7-10700 are qualification samples (QS). The clock speeds for the latter should be very close to what we can expect from the retail samples.

The i5-10400, i5-10500, i5-10600K are six-core, 12-thread processors with 12MB of L3 cache. The i5-10400 and i5-10500, which are 65W parts, reportedly feature a 4 GHz all-core boost clock, while the unlocked i5-10600K is expected to be rated for 125W and has a 4.1 GHz all-core boost.

The i7-10700 will reportedly come with eight cores, 16 threads and 16MB of L3 cache. This model seemingly operates within the 65W envelope and has a 4.6 GHz all-core boost.

Intel 10th Gen Comet Lake-S Specifications

ProcessorCores / ThreadsSingle-Core Boost Clock (GHz)All-Core Boost Clock (GHz)L3 Cache (MB)TDP (w)
AMD Ryzen 7 3700X8 / 164.4?3264
Intel Core i7-10700*8 / 164.74.61665
AMD Ryzen 5 3600X6 / 124.4?3295
Intel Core i5-10600K*6 / 124.34.112125
Intel Core i5-10500*6 / 124.24.01265
Intel Core i5-10400*6 / 124.34.01265

*Specifications are unconfirmed.

As we've known for a long time, Comet Lake-S processors will live on the new LGA1200 socket. The photographs on Sohu show that the Comet Lake's socket alignment notches are on the bottom of the processor instead of on the top, like they are on Coffee Lake chips. Intel likely revised the design to prevent users from trying to plug a Comet Lake part into a non-LGA1200 socket.

The Sohu user noted that the PCB for the engineering samples are a bit thinner than the retail chip. The qualification samples, on the other hand, share the same thickness as the retail PCB.

The close-up shots of the Comet Lake processors' backside show a redesign in the capacitor layout as well. Oddly enough, the capacitors on the qualification samples are divided into two groups, while the capacitors on the engineering samples are clustered together.

Intel 10th Gen Comet Lake-S Benchmarks

Normally, it's useless to get your hands on an unreleased processor, especially one that requires a new CPU socket. However, the Sohu user reportedly sourced a Dell B460 motherboard. As you would expect from an OEM motherboard, the power delivery subsystem is pretty weak, so the tester slapped a heatsink and fan over the area to prevent overheating.

The B460 motherboard only allows RAM speeds up to 2,666 MHz, which was the speed used. The tester also played with the BIOS settings a bit to unlock the power limit so the processor's frequency wouldn't scale down when it exceeded the TDP.  Finally, the test system is based on Windows 10 with the November 2019 update (Version 1909).

Being unreleased hardware, games didn't run properly with the Comet Lake chips. The Sohu user suspected that it was a problem with the motherboard. 

ProcessorCinebench R15Cinebench R20CPU-Z3DMark Time Spy (CPU Score)3DMark Fire Strike ExtremeMaster Lu Benchmark
AMD Ryzen 7 3700X2,116 / 2074,788 / 5025,588 / 5369,07523,102170,135
Intel Core i9-9900K2021 / 2184,937 / 5135,416 / 57411,23223,661178,971
Intel Core i7-107001,974 / 2074,828 / 4935,625 / 5689,12923,353171,861
AMD Ryzen 5 3600X1,652 / 2063,718 / 5044,249 / 5317,45620,552150,774
Intel Core i7-9700K1,439 / 1933,505 / 4544,244 / 5508,33218.859129,666
Intel Core i7-8700K1,398 / 1953,400 / 4723,886 / 5367,09617,013108,777
Intel Core i5-10600K1,368 / 1883,268 / 4443,769 / 5066,21117,786112,341
Intel Core i5-104001,319 / 1833,203 / 4293,678 / 5016,29117,337109,255
Intel Core i5-105001,199 / 1732,956 / 3973,363 / 4635,57115,88995,641
Intel Core i5-9600K1,022 / 1872,566 / 4712,886 / 5256,11813,80996,858

According to the results from the review, the i7-10700 is slightly faster than the AMD Ryzen 7 3700X, another 65W chip. But the difference is less than 2%, so it might not even be perceptible in a real-word scenario. The Ryzen 7 3700X did outperform the i7-10700 in the Cinebench R15 benchmark.

The i5-10400, i5-10500, i5-10600K were nowhere close to catching the AMD Ryzen 5 3600X. To be fair, the Ryzen 5 3600X is a 95W processor, while the i5-10400 and i5-10500 are 65W parts. Nevertheless, the two Core i5 chips might not pose a threat to the 65W AMD Ryzen 5 3600 either, since its performance is very close to the X variant.

ProcessorCinebench R15Cinebench R20CPU-ZMaster Lu Benchmark
AMD Ryzen 9 3900X727 / 1891,784 / 4611,954.2 / 49460,758
Intel Core i7-10700675 / 1771,656 / 4211,908.2 / 484.456,331
Intel Core i7-8700K657 / 1731,624 / 4161,804.9 / 48358,399
Intel Core i9-9900K657 / 1721,636 / 4191,831.1 / 483.460,482

AMD's dominance doesn't only come from offering more cores, but the Zen 2 microarchitecture and 7nm process node play an essential part in the equation as well. To level the playing field, the Sohu user locked the AMD Ryzen 9 3900X, i7-10700, i9-9900K and i7-8700K to a four-core, four-thread configuration at 4 GHz.

The charts show the Ryzen 9 3900X crushing three generations of Intel processors. In Intel's defense, Comet Lake seemingly performs better than Coffee Lake at the same core count and frequency. The improvement, as small as it may be, is there, based on these results. 

A report last week suggested that Comet Lake-S will finally see a reveal on April 30, meaning we might not see the products in shelves until late May or early June.

  • Gurg
    So the 10700 non-K beats the 3700x in Time Spy and Extreme the two measures closer to gaming. Also almost 10% increase over 9700k. The 10700k when overclocked with decent motherboard and memory should be a great CPU for gaming.
    Reply
  • [dev]
    Gurg said:
    So the 10700 non-K beats the 3700x in Time Spy and Extreme the two measures closer to gaming. Also almost 10% increase over 9700k. The 10700k when overclocked with decent motherboard and memory should be a great CPU for gaming.

    10% ?? Incredible difference LOL
    Don't you need another motherboard for that 10% ??
    Reply
  • mdd1963
    [dev] said:
    10% ?? Incredible difference LOL
    Don't you need another motherboard for that 10% ??

    I rather doubt many 9700K owners are participating in any CPU/mainboard upgrades this year....(and any that do, certainly are not doing so for any sort of 'my Battlefield 5 frame rates are not fast enough' cause) :)
    Reply
  • st379
    Gurg said:
    So the 10700 non-K beats the 3700x in Time Spy and Extreme the two measures closer to gaming. Also almost 10% increase over 9700k. The 10700k when overclocked with decent motherboard and memory should be a great CPU for gaming.
    10700 is 16 threads. It is suppose to be better than the 9900k without overclock not the 9700k.
    Probably the 10700k = 9900ks.
    i5 10600=8700k performance with minor ipc improvement if at all.
    lga 1200.
    I really don't see it selling well because of the new motherboard and high price.
    People with i3 or pentium can't upgrade to these new cpus without a new motherboard.
    Intel should not have change motherboard. It is skylake cpu they could make it work on lga 1151.
    Reply
  • al_capwn
    I think the biggest thing 10th-gen Intel CPUs will bring to the table is a lower price (I hope). Here's to waiting for 10nm from Intel, or smaller, for desktop. I'm glad AMD is putting up fantastic competition.

    I don't see upgrading my CPU until DDR5 is mainstream, and to see what Intel and AMD offers, in the foreseeable future.
    Reply
  • bit_user
    st379 said:
    Intel should not have change motherboard. It is skylake cpu they could make it work on lga 1151.
    Core uArch and socket have almost nothing to do with each other.

    Your complaint would certainly apply to Coffee Lake (where Intel could've retained the same socket for an unprecedented 4 generations), but Intel is now finally making I/O changes that necessitated a new socket.

    Comet lake tried to move to PCIe 4.0. Rocket Lake doubled the DMI datapath, upgraded its CPU-direct PCIe lanes to 4.0, and added another x4 for NVMe storage.
    Reply
  • bit_user
    al_capwn said:
    I think the biggest thing 10th-gen Intel CPUs will bring to the table is a lower price (I hope).
    Nope. These are 14 nm CPUs, meaning the cost structure won't change relative to Coffee Lake Refresh CPUs. That's not to say they won't be cheaper, but Intel could more easily just drop prices on their current gen.

    What's worse is that having more cores per CPU should actually make them more expensive (assuming Intel is able to hold its margins constant).

    al_capwn said:
    Here's to waiting for 10nm from Intel, or smaller, for desktop.
    Intel's CFO recently said that 10 nm would be less profitable, for them. You can bet they'll push the higher production costs onto consumers, to the extent they can. Only stiff competition from AMD can hope to keep that in check.
    Reply
  • TerryLaze
    bit_user said:
    Nope. These are 14 nm CPUs, meaning the cost structure won't change relative to Coffee Lake Refresh CPUs. That's not to say they won't be cheaper, but Intel could more easily just drop prices on their current gen.

    What's worse is that having more cores per CPU should actually make them more expensive (assuming Intel is able to hold its margins constant).
    More cores per money=cheaper,at least for the consumer.
    There has never been a CPU gen from intel that had prices based on production costs...
    bit_user said:
    Intel's CFO recently said that 10 nm would be less profitable, for them. You can bet they'll push the higher production costs onto consumers, to the extent they can. Only stiff competition from AMD can hope to keep that in check.
    Less profitable does not mean higher production costs,the stiff competition from AMD made intel have twice the profit as usual with 14nm so it's only logical to assume that 10nm is not going to make as much again,intel has to assume that profit will go down again to the usual ~10bil.
    https://www.statista.com/statistics/263565/intels-net-income-since-2004/
    Reply
  • Gurg
    bit_user said:
    Nope. These are 14 nm CPUs, meaning the cost structure won't change relative to Coffee Lake Refresh CPUs. That's not to say they won't be cheaper, but Intel could more easily just drop prices on their current gen. What's worse is that having more cores per CPU should actually make them more expensive (assuming Intel is able to hold its margins constant). Intel's CFO recently said that 10 nm would be less profitable, for them. You can bet they'll push the higher production costs onto consumers, to the extent they can. Only stiff competition from AMD can hope to keep that in check.

    Actually I believe you are wrong. Keeping on 14nm means that Intel has been able to continue using much of the same equipment this whole time. So in an accounting sense it is now fully depreciated so Intel's manufacturing costs are going down significantly. A semiconductor chip plant basically needs to replace lots of existing equipment each time they move to a smaller production size for improved tolerances. Additionally yields per wafer improve and can be maximized the longer the process is in production.

    In a financial sense the shift to 10nm will result in higher production costs due to the accelerated depreciation on the new replacement fab equipment needed to improve tolerances. Additionally the yield of good chips per wafer is much lower when moving to a new production size initially improving over time. Both of those factors outweigh the increased number of chips that can fit on the wafer. Since Intel is still looking to reduce to 7nm or lower in the coming years means that depreciation cost of new equipment will remain high for the entire shorter run at 10 nm compared to the long run at 14nm. Intel will also have trouble achieving the same high yield percentages it is currently getting from its long run at 14nm.

    That wafer that Intel was letting the media handle and take pictures of a month or two ago was most likely a reject with only around 20% or fewer good chips that didn't justify cutting the low number of good chips out for use and would have been scrapped. Usually with low yield wafer, the fewer good chips do not perform that well either.
    Reply
  • st379
    bit_user said:
    Core uArch and socket have almost nothing to do with each other.

    Your complaint would certainly apply to Coffee Lake (where Intel could've retained the same socket for an unprecedented 4 generations), but Intel is now finally making I/O changes that necessitated a new socket.

    Comet lake tried to move to PCIe 4.0. Rocket Lake doubled the DMI datapath, upgraded its CPU-direct PCIe lanes to 4.0, and added another x4 for NVMe storage.

    If Amd can do it Intel can do it. Not to mention 1151v2 cpu not supported on 1151v1.
    I remember that when the new cpu arrived someone release a bios update for 1151v1 that make those cpu work on it.
    Reply