In a latest {hardware} deep-dive, we took a take a look at how CPU cores and cache impacted gaming performance. To do this, we used three Intel Core processors and in contrast their core scaling efficiency, revealing that in in the present day’s video games many of the efficiency good points you see when going from a Core i5-10600K to a Core i7-10700K and even Core i9-10900K are largely as a result of improve L3 cache capability.
This was an fascinating evaluation as a result of most individuals who improve from an older era Intel half such because the Core i7-8700K to the newer Core i9-10900K and noticed a robust efficiency uplift in video games, are inclined to consider this comes as the results of the 67% improve in cores, however for probably the most half it’s truly as a result of 67% improve in L3 cache — a minimum of that is the case for in the present day’s most demanding video games.
Coming away from that testing, lots of you needed to know the way a lot distinction the L3 cache capability makes with simply 4 cores energetic, questioning if the margins could be even higher. So we’ve gone again and retested a 4-core, 8-thread configuration whereas including three extra video games to the checklist together with a fourth processor, the 4C/8T Core i3-10105F.
As a result of the Core i3 fashions are locked, we’re unable to make use of a 4.5 GHz clock frequency, as an alternative the 10105F ran at 4.2 GHz which is the spec all-core frequency, so it’s going to be working at a 7% decrease frequency than the Ok-SKU components. The Core i3-10325 would have been extra appropriate for this check, however we had been unable to safe one in time for this check. Regardless that the Core i3 half is working at an obstacle, it ought to stay an fascinating addition given the way more restricted 6MB L3 cache.
The quickest quad-core CPU that Intel has produced so far is the Core i7-7700K, or one of many greater clocked Comet Lake Core i3s, such because the Core i3-10325, each of which characteristic an 8MB L3 cache, or 2MB greater than the 10105F. So it’s going to be fascinating to see what sort of good points could be attainable utilizing a quad-core CPU with 20MB of L3 cache, one thing we are able to obtain with the 10900K by disabling half a dozen cores on that half.
To place this check collectively we used the Gigabyte Z590 Aorus Xtreme motherboard, clocking the three Intel Ok-SKU CPUs at 4.5 GHz with a 45x multiplier for the ring bus and used DDR4-3200 CL14 dual-rank, dual-channel reminiscence with all main, secondary and tertiary timings manually configured. The Core i3-10105F used the identical spec reminiscence.
The majority of the testing was ran with the Radeon RX 6900 XT because it’s the quickest 1080p gaming graphics card you should buy, though we have included some outcomes with the RTX 3090 for a take a look at Nvidia’s overhead, which we suspect will have an effect on the quad-core configurations.
Benchmarks
We’ll begin with Rainbow Six Siege the place we beforehand noticed the most significant performance variations between cache capacities. For instance, with all CPUs locked to six cores, we noticed an enormous 18% improve from 12MB to twenty MB of L3 cache.
Then with solely 4 cores enabled, the margin from the 10600K to the 10900K is definitely diminished to 13%, fairly a bit lower than the margin seen with 6 cores enabled. This offers us a touch that the bigger L3 cache turns into much less efficient at boosting efficiency with fewer cores obtainable to put it to use.
That is greatest illustrated by the Core i3-10105F, which was simply 9% slower than the 10600K, and keep in mind it’s clocked 7% decrease, so presumably a minimum of half that margin is all the way down to the distinction in clock velocity. It is nonetheless fascinating to notice that it was attainable to enhance efficiency from the 4-core Core i3 by 24%, when utilizing the 10900K with simply 4 cores energetic. That’s an enormous improve given each CPUs use the identical Comet Lake structure with the one distinction being a 7% clock frequency variation and the L3 cache capability, which is over 3x bigger for the i9 half.
The takeaway right here is that not all cores are equal, even when the cores are bodily the identical, a distinction in cache capability could make all of the distinction. That and cramming extra cache right into a CPU continues to be helpful with simply 4 cores energetic, nevertheless it’s much less efficient than what we noticed with 6 cores.
Subsequent up we’ve got Murderer’s Creed Valhalla which is a closely GPU-bound title, particularly when utilizing the ‘Extremely Excessive’ high quality preset, even with a Radeon RX 6900 XT at 1080p. In consequence the benchmark depends nearly completely on the GPU and solely the Core i3 sees a small dip in 1% low efficiency, suggesting that body time efficiency isn’t pretty much as good with the 6MB L3 cache, however not considerably worse both.
A brand new addition to this testing is Battlefield V. Right here we are able to see that the 20 MB L3 cache of the 10900K may be very helpful, even when restricted to six cores. Dialing all the way down to 4 cores creates a CPU bottleneck that may’t be solved with extra cache. 1% low efficiency was improved by 9% when going from 16 to twenty MB which isn’t nothing, nevertheless it’s lower than the 13% achieve we noticed with 6 cores enabled.
Now, with solely 4 cores enabled, the 10900K was similar to the 10600K inventory. In reality, the body time efficiency was higher, enhancing 1% lows by a 12% margin. When it got here to 1% low efficiency the Core i3 struggled a bit, dipping all the way down to 71 fps making it 19% slower than the 10600K and an enormous 30% slower than the 10900K. Once more that’s an unbelievable distinction that’s nearly completely all the way down to the distinction in L3 cache capability.
The F1 2020 outcomes are just like what we’ve seen to this point, the place the L3 cache paucity made extra of a distinction with 6 cores, although we do see some distinction with 4 cores energetic. The 10700K was simply 3% slower than the 10900K, whereas the 10600K was an additional 3% slower than the 10700K. So fairly constant scaling there. The Core i3 half although was 8% slower than the 10600K, or 11% slower if we take a look at the 1% low outcomes.
We are able to estimate that at most ~5% of that margin could possibly be as a result of distinction in clock velocity, and if that is the case, scaling continues to be roughly inline with that we noticed for the Ok-SKUs.
Hitman is a CPU intensive title, and we are able to see a 9% improve in efficiency when going from a 12MB to 20MB L3 cache with 6 cores enabled. That margin was diminished with simply 4 cores enabled, this time to six%. Then we noticed a 9% drop in efficiency from the 10600K to the 10105F.
Horizon Zero Daybreak is so much like Murderer’s Creed Valhalla within the sense that it’s primarily a GPU certain title, even at 1080p with a 6900 XT. In consequence, the 4-core configurations weren’t a fantastic deal slower than what we noticed with 6, 8 and 10 cores enabled. The Core i3 half additionally carried out nicely relative to the higher-end Core i5, i7 and i9 fashions.
Cyberpunk 2077 is a really CPU intensive sport and you will notice an enormous profit when upgrading to a contemporary 6-core processor, such because the Core i5-10600K. That stated, had the 10600K been armed with a 20MB L3 cache it might be 13% quicker on this title when wanting on the 1% low efficiency. That margin remained pretty in step with simply 4 cores enabled because the 10900K configuration was 12% quicker than the 10600K.
Apparently, the 4-core 10600K configuration was simply 6% quicker than the Core i3-10105F on common, however 19% quicker when evaluating the 1% low efficiency and that is the place the Core i3 actually struggled with simply 64 fps.
When wanting on the Core i3 half we see an enormous 56% efficiency disparity between the 1% low and common body price, whereas the 10600K sees only a 39% margin and this means way more constant body time efficiency.
It’s price noting that the 10900K with simply 4 cores energetic performed very nicely, providing easy and constant efficiency, regardless of being 25% slower than the 6, 8 and 10-core configurations. The 10105F although suffered from noticeable stuttering and once more that was mirrored within the a lot weaker 1% low efficiency. So whereas a Tenth-gen quad-core with a fats L3 cache can play the sport simply wonderful, it’s nonetheless a lot slower than a 6-core equal, whereas going to eight cores provides no enchancment in common body price efficiency or body time efficiency.
Subsequent up we’ve got Shadow of the Tomb Raider which is one other CPU demanding sport however once more, with simply 4 cores enabled, the Tenth-gen CPUs aren’t capable of reap the benefits of that further L3 cache like they’re with 6-cores. It’s not till the L3 cache is ready to 6MB with the Core i3 half that efficiency begins to fall away, dropping the 1% low efficiency by 16% when in comparison with the 10600K.
Testing with the RTX 3090
As we had been wrapping up the quad-core testing, we thought it is perhaps fascinating to re-run just a few of those utilizing the GeForce RTX 3090. With 6-cores enabled in SoTR, the 10600K was 13% slower than the 10900K utilizing the Radeon and 15% slower with the GeForce, so not an enormous change there.
Then with 4 cores enabled, the 10600K was 4% slower than the 10900K with the Radeon, whereas we see radically completely different outcomes with the RTX 3090. Right here the 10600K is 14% slower than the 10900K for the typical body price and 13% slower for the 1% low. That is the results of Nvidia’s added overhead through the use of the CPU for a lot of its GPU scheduling.
Nevertheless, it’s the Core i3 half that is actually crippled by the software program scheduling, tanking 1% low efficiency to only 54 fps, making it 28% slower than the 10600K, whereas it was 16% slower earlier than. Had we used the RTX 3090 for all testing, the 4-core leads to Hitman, Cyberpunk 2077, Battlefield V, and so forth may need been way more important.
We additionally examined Watch Canine Legion with each Radeon and GeForce high-end GPUs. Earlier than with the Radeon we didn’t see a lot of a distinction with the Ok-SKU components, with 6 cores enabled the outcomes had been a lot the identical after which 8 cores solely provided a small enhance. Nevertheless, 4 cores drop efficiency fairly significantly. The 10600K, for instance, was 17% slower with 4 cores enabled, and we noticed related margins with the 10700K and 10900K. That stated, efficiency between the assorted 4-core configurations is comparable and even the Core i3 half manages to hold in there.
Utilizing the RTX 3090 doesn’t do a lot to the 4-core outcomes. The 10600K was 19% slower with simply 4 cores enabled when in comparison with its inventory 6-core configuration. Unexpectedly although, the margins with 6 and eight cores enabled are fairly completely different and it appears that evidently the cache performs a bigger position with the RTX 3090 put in, presumably as a result of the CPU is having to do extra work.
It’s additionally fascinating how we don’t see the identical scaling with 4 cores enabled, nevertheless it does seem as if that is too few cores to reap the benefits of the elevated L3 cache capability.
What We Realized
This was an fascinating take a look at CPU efficiency however in all probability not what lots of you had been anticipating. We consider the expectation was that with fewer cores, the bigger L3 cache of the 10900K would play a fair higher position, however for probably the most half that doesn’t look like the case.
It’s clear that for mid to high-end gaming, quad-cores are formally out. We’ve identified this for a while, which is why AMD and Intel have stopped producing mid-range quad-core processors. For lower-end techniques nevertheless, quad-cores nonetheless work nicely, although it ought to be clarified that after we say quad cores, we imply 4-core/8-thread processors that do assist simultaneous multi-threading.
Body time efficiency might be seen struggling in demanding titles akin to Battlefield V, Shadow of the Tomb Raider and Cyberpunk 2077, for instance, however in most video games for those who’re utilizing a funds GPU, such because the Radeon RX 5500 XT, GeForce GTX 1650 Super, or something slower, an honest quad-core will allow an appropriate stage of efficiency.
Body capping to 60 fps will assist easy our body charges in a title like Cyberpunk 2077 because it reduces CPU load in addition to body to border variance. So for those who’re working into stuttering points, strive capping the body price to one thing extra sustainable as that may assist.
