Today we're re-reviewing our original Core i9-9900K test and updating it with 95-watt TDP results, which are largely based on the official Intel specification. For a better context, please read our opinion article earlier this week entitled "Do we need to review the Core i9-9900K again?"
The short version of this is that motherboard manufacturers are currently held responsible for the fact that the 9900K does not meet the specifications. In reality, we firmly believe that Intel is cheating on its own specs and pushing board partners to run the 9900K on the standard clock multiplier table rather than the official performance spec.
However, the 9900K does not run on the Intel specification, it is essentially overclocked and this has resulted in power and heat results going through the roof. In today's re-test, we will show how the Core i9-9900K behaves if the Intel specification is complied with and this data is compared with the current out-of-the-box experience.
It doesn't matter where you stand. In our opinion, a resource that shows how these configurations are compared under the same test conditions is useful information. The MSI MEG Z390 Godlike was used for the unlimited tests, and for the limited tests with 95 watts I used the Asus ROG Maximus XI Hero, loaded the memory profile "xtreme" and decided to use the Intel settings that the 95 -Watt force TDP. So let's get to the results …
The Cinebench R15 multithreaded scores come first. We previously found the 9900K to exceed the 2000 point limit. With the applicable TDP limit, however, the value is reduced by 14%. It is roughly on par with the Core i7-7820X and was crucial that it was a few percent slower than the 2700X. You may already get a sense of why Intel is happy when the board partners run out of specifications.
Next we have the Blender short-term test, and here the 95-watt TDP-restricted configuration can burst up to 120 watts for only 10 seconds. In other words, for about half of the test, it's almost completely unleashed, and we only saw a 9% drop in performance. This is still a reasonable decline, but it's not the whole story.
Professionals who want to invest in a rendering rig run workloads that take much longer than 20 to 30 seconds. We generally talk about hours of rendering work. While we saw a 9% reduction in the short-term test, we see a 14% reduction in the more realistic rendering workload. That's a pretty big drop, and means the 9900K only keeps up with the Ryzen 7 2700X.
The Corona benchmark runs over a minute and here we see a 13% drop in performance when performance is limited. The 9900K was 25% faster than the 2700X when it was allowed to operate without power limitation, but with the forced 95 watt TDP it is only 9% faster. It's still faster, but the margin is not that impressive anymore.
Here we see a 15% power reduction for the 9900K using the 95 watt limit, which means that it was only 4% faster than the 2700X, while we previously found it to be 23% faster.
The 7-Zip compression performance is not significantly affected by the 95 watt TDP limit as we see a minimal drop of 3%.
When decompressed, the 9900K suffers a little more and shows 7% slower results when the TDP limit is enforced. This is enough to make it slower than the Ryzen 7 2700X. However, there is no significant difference between the two.
Excel is the perfect example of a short workload. In less than 10 seconds, the 9900K is not affected by the official specification, and we see almost the same performance with and without the applicable TDP limit.
When testing with HandBrake, a performance reduction of 14% is determined if the TDP limit is enforced. Therefore, the 9900K is only 4% faster than the 8700K and 13% faster than the Ryzen 7 2700X. The 1st and 2nd generation Ryzen CPUs are not so well suited for AVX workloads. So let's look at the edges in the H.264 test.
While we saw a 14% drop in performance when executing the H.265 encoding, we only see half of those affected by a workload without AVX encoding. However, Ryzen is much better with H.264 and looks more competitive here.
The 9900K with the TDP limit dropped to an all-core value of 4 GHz in the H.265 test, while maintaining 4.2 GHz in the H.264 test.
In the further course we see another example, in which the 95-watt TDP sees the 9900K with a small lead behind the 2700X. It is also interesting that with this Premiere Pro CC export, the unlimited 9900K matched the 7820X, a 140 W part in the same process. I have already said that the 9900K should have at least 140 W TDP power and that seems to fit what we see here.
The Premiere Warp stabilizer test does not always maximize all cores, but is a typical machining load. Here we see a 6% power reduction when the TDP limit is set. Still, that was enough to see the 9900K behind the 8700K.
The following graphic results explain a lot … During the Core i9-9900K's launch week and the first series of tests, we saw some test reports claiming that the 9900K consumes less power than the 8700K, which doesn't really do the common sense test consists .
However, if you test the 9900K with a 95 watt TDP limit and the 8700K without a TDP limit, you will get it. Even if the TDP limit applies, the 8700K and its 6 cores are not nearly as badly affected as the 9900K and its 8 cores, so that even the 9900K requires a TDP limit of at least 125 watts.
What we see here is a 31% drop in total system consumption as all 8 cores are shut down from 4.7 GHz to 4.0 GHz and you can see that a 15% reduction has a profound impact on system consumption since we also take a lot of voltage off the chip at the lower clock speed.
In Blender we see a 27% reduction in total system consumption, and now the 9900K looks like a powerfully efficient CPU. It was a few percent faster than the 2700X and here we see that it reduced overall system consumption by 12%. Previously it was 19% faster than the 2700X, but consumption also increased by 21%.
Given what we've seen from the results of overall system consumption, these thermal numbers, though shocking, aren't that surprising. With the Noctua NH-D15 and Corsair Hydro H100i Pro, I found that the 9900K reached temperatures in the mid-1980s when it was fully unleashed.
However, using the 95-watt TDP specification, the 9900K reached a maximum of only 64 degrees in our Blender stress test, and that number was reduced to only 58 degrees with our custom loop. So if all cores are operated at 4 GHz, the 9900K is as cool as a cucumber, but at 4.7 GHz it turns the CPU socket into a fiery pit of melting silicon. Ok … it's not that bad, but it's damn hot in comparison.
When testing with Assassins Creed Odyssey, we find that CPU-intensive titles show a measurable drop in performance, but this only occurs under unrealistic conditions, e.g. B. when playing at 1080p with an RTX 2080 Ti. At 1080p we see an 8% deterioration in frame time performance, then this margin is reduced to 3% at 1440p.
Playing GPU-bound titles like Forza Horizon 4 has no effect, and I'm assuming that the 1080p results will look like this in most tiles. So keep that in mind, as this review focuses primarily on CPU-bound games, with Forza being the exception.
Here we see a 7% drop in frame time performance when testing with Hitman at 1080p, a 6% drop in 1440p, and only when we reach 4K does the margin evaporate.
Interestingly, we don't see any real impact in Project Cars 2, and this title is a little strange in the sense that the 9900K is so much faster than the 8700K. I'm not sure why that is because the game doesn't need 8 cores. Other sources have confirmed these margins, so it's not a strange bug in our test system.
There is no real leeway when testing with Rainbow Six Siege. Every modern CPU with about 4 GHz seems to work well here.
We see Shadow of the Tomb Raider seeing a pretty hefty 15% drop in frame time performance at 1080p, although once we're at 1440p we're almost entirely tied to the GPU.
After all, we have Star Wars Battlefront II, where we see a small drop in performance at 1080p, nothing extreme, and when we reach 1440p, the edges are almost nothing.
First things first: Our original Core i9-9900K test is as valid today as it was on the day we released it. In this second look, we took a closer look at the question of why the 9900K is so hot out of the box and explain why some test reports showed cheaper performance figures, albeit with reduced raw power.
player I don't have to worry It doesn't make much of a difference with the 95-watt TDP specification, and overclocking on Z390 motherboards doesn't result in insane thermals and power consumption for games.
However, if you both work and play and your work involves intensive CPU workouts – it's the only type of workouts I seem to be doing these days – that's a different story. We noticed an increase of 20 degrees in our long-term blender workload temperatures, which rise through the roof when the 9900K is operated outside the TDP specification. It was of course similar with electricity consumption. We saw an increase in total system consumption of almost 40%. However, if the 9900K were forced to work with Intel's TDP specification and meet performance limits, it would be a powerfully efficient 8-core processor. You get 2700 times the power and save a little more than 10% electricity.
But that's the problem for Intel, the 9900K was already a tough sell in the overclocked configuration used by all motherboard manufacturers and now … Intel knew this would be the case. It is an $ 500 8-core desktop CPU that competes with an $ 300 8-core desktop CPU. As we just saw at the 95 watt limit, it is hardly faster than the Ryzen 7 2700X. In some tests, it's even slower, and that's a terrible result for a CPU that costs ~ 70% more.
This is a big problem for Intel and they have painted themselves in a corner here. For the 9900K to make sense, it must be operated with a high-quality aftermarket cooler at around 70 ° C for anyone who is not an extreme overclocker, and for this the TDP cannot be higher than around 105 watts.
Even at 105 watts, it's barely faster than the uncovered 8700K and just a whisker faster than the much cheaper 2700X, so you can't have that with a Core i9. For those of you wondering, a limit of 105 watts means the 9900K can withstand a clock speed of 4.15 GHz in our Blender workload and runs at 69 degrees with the Corsair H100i Pro. This corresponds to an increase of 150 MHz compared to the TDP limit of 95 watts and an increase in the operating temperature by 5 degrees.
Basically, the 9900K is a really good overclocker if you invest in the right cooling. The 2700X, on the other hand, is in a situation where what you see immediately is pretty much what you get. You can get a little more out of it and setting the memory sub-timings really helps, but you can't overclock the snot like cores from Intel CPUs.
With motherboards that technically overclock the 9900K to the standard clock multiplier table, for example 4.7 GHz as an all-core, there is not much left under these conditions. For most 5 GHz, the limit will be good luck keeping it cool without much time, effort and risk. So when you look at the unlimited results, you realistically speak of an increase of up to 6% over what is shown here, and we certainly found this when we tried to overclock the 9900K in our first test.
With a view to future reviews, we plan to show the typical out-of-the-box experience. If this means overclocked CPUs, we'll show it too. We will continue to post additional insights and features like this when necessary.