Intel i5-12600K Review – TWO CPUS IN ONE TESTED FULLY!
Intel’s Alder Lake CPUs officially launched last week, with the rising star being this unlocked i5, the 12600K. This is a 6P + 4E chip meaning it has 6 “proper” performance cores with two threads each, and 4 efficiency cores with no hyperthreading making for 16 total threads. The P cores can clock up to 4.9GHz, with the E cores capped at up to 3.6GHz. This, much like all the Alder Lake chips launched so far, has a 125W Base Power rating, and a much more modest 150W Maximum Turbo Power rating compared to the i9’s 241W.
So, how does this 6 and a bit core hold up against last gen’s 6 core, and AMD’s 6 and 8 core options? Well starting with Cinebench R23 single threaded, much like the i9 it holds a healthy lead over everything. It’s 20% faster than both the last gen 11600K and Ryzen 5600X, and 15% faster than the Ryzen 5800X. It’s score is down from the 12900K thanks to no Turbo Boost 3.0 support and lower peak clock speeds, but it’s a convincing lead so far.
In multi-threaded, both thanks to the faster single core performance and those extra E cores, the 12600K trounces, well, everything. It’s 800 points clear of even the 5800X, or about 15% faster, and a whopping 60% faster than the 6 core parts. That’s an excellent lead for sure, but what’s more impressive is adding in the rest of the chips I’ve tested and you’ll notice the 12600K actually sits on top of both the i9 11900K and i9 10850K! It’s only bested by the 12 core Ryzen 5900X and of course the bigger brother 12900K, but to have it besting both last-gen’s 8 core i9 AND the generation before that’s 10 core i9… That’s incredible.
In Blender and the BMW scene, the new i5 is still the fastest around, although the gap to the 5800X is a touch slimmer at 10%, but gen on gen… the 11600K runs 58% slower, with the 5600X only just ahead at 56% slower. That’s a significant upgrade over the Rocket Lake chip – the one launched in March this year.
In the Gooseberry render, that’s a different story. Thanks to Intel’s boost algorithm punching hard at the start only to sink slowly – and possibly some bugs with the pre-release version of the BIOS I’m testing with – the 5800X actually overtakes the 12600K, with the new i5 running a shade under 10% slower. It is still faster than both last – I was going to say last year but that’s not true, last month? The 11600K. It’s faster than both the 11600K and Ryzen 5600X, although the gap to the 5600X is down to just 25%.
In Premiere Pro, it’s king of the mid-field, netting a 5% lead over the 5800X and around a 20% advantage over the 6 core options. In After Effects, rather interestingly the last gen 11600K actually holds a convincing lead over everything. I’m not sure I can explain this one, especially since that’s a higher score than both the 12900K and 11900K got, but it’s confirmed in Photoshop where again the 11600K takes the lead taking practically the fastest score I’ve seen this week. If you’ve got any ideas feel free to leave them in the comments below!
Of course, most of you aren’t too concerned with rendering or editing, you care about gaming and so do I so let’s look at those numbers. Testing at 1080p with an RTX 3080, in CSGO the 12600K takes the backseat as both 6 and 8 core Ryzen chips offer 16% more performance. We are talking about 500 vs closing on 600 FPS though, so hardly a “noticeable” difference, but a win for AMD nonetheless. What is a win for Intel though is the generational improvement – 34% more performance over the 11600K which is a fantastic step up.
Cyberpunk is the opposite, with the 12600K taking a more noticeable victory, netting around 138 FPS compared to the 11600K at around 117 FPS, and the Ryzen chips at 108 and 106 FPS. Also worth noting the improvement in 1% low figures too, from around 65 to 88 FPS instead.
Watchdogs is also a clear victory for the new i5, running 149 FPS average and a stunning 112 FPS in the 1% lows, versus 116 FPS average and 86 FPS in the 1% lows for, well, everything else. Microsoft flight swings back the other way, with AMD leading the pack by 18%, running around 20 FPS faster on average. And finally in Fortnite it’s pretty much a tie. The last gen i5 does run a touch slower, but it’s not by enough to be notable at all.
I should make it clear, all of the benchmarks in this video have been using Windows 10. I’ve run multiple sets of complete benchmarks on both Windows 10 and 11, but thanks to countless bugs and issues I wasn’t able to get a consistent set of results on Windows 11. I did verify my 12th gen results were accurate – or as accurate as I can get them anyway – with 10 to 11. Windows 11, at least from my experience, has a long way to go before I’d be comfortable testing on it and confident the results I’d get would be accurate.
Most reviews of these chips would end here, but I think that’s a little short-sighted and leaves out a lot of interesting details since these aren’t just another chip. This is two CPUs pretending to be one, so let’s test them separately. Now testing the P cores on their own is easy, disable all E cores in the BIOS and test away. Getting the E cores on their own is harder – you can disable all but one P core, regardless of how many E cores are running. So, with a mixture of using task manager to set CPU affinity – basically telling the scheduler what cores the programs are allowed to use – and running games like Fortnite with a cmd command to pre-set affinity as it’s anti-cheat engine blocks any attempts to change affinity while it’s running, then you can test the E cores independently.
So how do they both fair? In Cinebench R23 the P cores score around 13,500 points, putting it a sizable chunk below the 8 core 5800X, although still 25% faster than the 11600K and 5600X. The E cores meanwhile, well they get outperformed by pretty much anything you can think of. Even an i3 10100 offers better performance, around 50% more actually, in multi-threaded work.
In single threaded, the P cores actually gain a fraction more performance than with both cores enabled. That’s potentially thanks to the extra power budget, or just the scheduler being able to accidentally drop the thread on an E core for any amount of time. As for the E cores, well they are about 12% slower than a 10100 on single threaded work. What’s really interesting for me about them is the power usage. I was seeing between 26 and 30 watts being drawn from the 4 E cores themselves, or around 36W for the whole chip. When you compare that to the 10100, which in my testing ran at just 46W peak, well that’s only around 30% higher, but the i3 offers around 50% more performance. Conceivably then, an underclocked 10100 could match the power of the E cores, and still outperform them. Of course, space efficiency is also key with this design, but it’s an interesting comparison nonetheless!
Where it gets really interesting is in the gaming performance. In CSGO the P cores match the stock results – although actually outpace it in the 1% low figures. The E cores though? That was a rough experience. While it technically garnered 230 FPS average, there was a significant amount of hitching, frame drops and stuttering that made playing incredibly difficult and far from enjoyable.
In Cyberpunk the P cores on their own actually outperformed the stock results, both on average and in the 1% lows, by a reasonable margin. It’s not night and day or anything, but I think shows both the issues with the scheduler having to deal with the hybrid design, and that the E cores can actually be a hindrance to performance in gaming.
Speaking of a hindrance, in Watchdogs Legion, the anti-cheat software they use refused to let the game even launch with only the E cores and one P core active, actually both on this i5 and on the i9 with 8 E cores, it still refused to run with that configuration. The P cores actually dropped a touch compared to stock here though, although hardly a large fall and still well clear of anything else tested.
Flight, much like the i9 actually, saw a significant increase in performance on the P cores alone. It’s still not quite at the same level as both Ryzen chips, running 9 FPS lower on average, but it is a full 12 FPS higher than stock which is pretty significant – not far off 10% more performance for DISABLING the E cores. Speaking of them, they actually didn’t do as badly as expected. There was still some hitching and frame drops but not as bad, and still managed 80FPS average which is honestly impressive.
Finally in Fortnite the P cores stayed the same. Technically they actually rose a little, but not enough to take note of – although the almost halved performance from the E cores is worth noting. Remarkably the playing experience here, despite the halved FPS, wasn’t bad. It was certainly playable which is really very impressive.
One thing I haven’t mentioned yet is pricing – Intel has gone hard here, setting the 12600K at around £290 here in the UK. For context, the last gen 11600K is now selling for around £220, AMD’s Ryzen 5600X can be found for around £270, and their 5800X is a whopping £370. Considering the new i5 can outpace even the 5800X in CPU specific workloads, and in theory should also be faster in games, £290 for that seems like a bargain.
But, the key words there are “seems like”. That’s because while the chip itself is very competitively priced, the second you look to actually use it the costs start piling up. First is a motherboard. Z690 boards seem to start around £200, where AMD’s B550 platform can be had for £120-150 for a very competent board, or hell get an updated B450 board on ebay for under £100. Then there is the RAM. While you can buy a motherboard that supports DDR4, performance is meant to be less than ideal, so to get the most out of your new chip you need to spend over double the price of a reasonable DDR4 kit to get yourself some shiny new DDR5, like the Corsair Vengeance kit I’ve been using in a lot of my testing. Luckily power draw isn’t quite as bad on this i5, and while it can draw 70% more power than a 5600X, it’s still low enough that most 240mm AIOs should handle it fine. So, taking the added cost of the motherboard alone and you are comparing to the 5800X instead, and if you factor DDR5 in too you could afford a 5900X instead and that’s a no brainer if CPU horsepower is what you are after.
Don’t get me wrong, this is an impressive chip. The P core performance is fantastic, it’s a sizable improvement over last gen and a comfortable lead over AMD’s 6 core offering, and should perform better in games too – albeit hardly a night and day difference to most. The addition of the E cores, for me, added unneeded complexity and a slew of bugs that I expect will take a number of months to properly iron out. Personally, I’d hold off on buying this one and pick up the rumoured 6P + 0E design and since that’ll likely be locked anyway, a cheaper B or H series motherboard. That I think would be a real killer, it should be exceptional value, it shouldn’t bring too many bugs and have pretty incredible performance to boot. That’s the chip I think AMD should be afraid of, and what you should be excited about.
-
TechteamGB Score
