Ticker

6/recent/ticker-posts

Overclocking Intel Non-K CPUs: Core i3-12100, Core i5-12400 and Core i7-12700

We're revisiting Intel's locked Alder Lake CPUs with a special sneak peak of an upcoming MSI motherboard that's capable of making some of our favourite current generation processors much faster. For example, it's possible to make the Core i5-12400 up to 50% faster in games, and we're going to show you how.One of the only issues with parts such as the Core i3-12100, Core i5-12400 and Core i7-12700 is that Intel locks them down, meaning there's no way to tinker with them to boost performance by overclocking them.



Out of the box performance is still excellent, and if you're building a new PC it's hard to ignore what these locked Alder Lake CPUs have to offer. It'd be hard to ignore if the boost was in excess of 20% faster, and even more so when it can exceed 50% in some instances.
How Does It Work?

Overclocking locked Non-K Alder Lake processors was first discovered by renowned overclocker der8auer. Just a few months after the release of Intel 12th-gen Core series der8auer managed to overclock a number of locked CPUs using the BCLK method, pushing them to frequencies well above what they normally operate at. For example, he pushed a Celeron G6900 from the default clock frequency of 3.4 GHz to 5.3 GHz, a massive 56% frequency increase.



This was exciting news and we went on to investigate the findings, but we never did as there was a catch, a rather significant one, that made this pointless for the vast majority of users. Not to diminish der8auer's findings as they were accurate and hats off to him for being the first one to discover this, but the issue boiled down to motherboard support.



You see, der8auer discovered that BCLK overclocking was possible on the Asus ROG Maximus Z690 Apex, a nearly $1,000 Z690 motherboard that requires DDR5 memory. While a cool discovery regardless, it's not exactly a viable option for consumers. There were cheaper boards that supported BCLK overclocking such as the Asus ROG Strix B660-G Gaming WiFi and B660-F Gaming WiFi, but both required DDR5 memory and are still priced over $200 -- not the ideal pairing for a $160 Core i5-12400F processor.



The reason these LGA 1700 boards can overclock locked CPUs is because they feature a PCIe 5.0 clock generator. This is why the ROG Strix B660-G Gaming and B660-F Gaming offer PCIe 5.0 x16 support for the primary PCIe x16 slot, whereas almost all other boards are limited to PCIe 4.0.
OC That Makes Sense

With the Asus B660 boards both priced at $310, we've ignored locked 12th-gen overclocking, until now. News broke recently that MSI was working on a special BCLK OC version of their B660M Mortar to be called the "MAG B660M Mortar Max WiFi DDR4". Given the B660M Mortar is much cheaper than the Asus -- like half the price -- I was very interested.



We asked MSI about the rumors and they said the information was strictly confidential, but also that yes, this is something they're working on. Naturally we requested they send us a board and to our surprise they agreed and sent an early pre-production unit, so here we are.



When compared to the original B660M Mortar, this Max model is the exact same with the exception of the Renesas RC26008 external clock generator and a minor tweak to the VRM which sees the Auxiliary MOSFET upgraded from a 70A model to an 80A model.
The Test

Time to overclock, and the purpose of this feature isn't to max out the locked CPUs that we have, but show you that: 1) the BCLK function actually works, and 2) the minimum level of performance you should be able to obtain.

We've overclocked the Core i3-12100, i5-12400 and i7-12700 to an all-core frequency of 5.1 GHz with a Ring frequency of 4.1 GHz. The P-Core ratio was set to x39 using the fixed mode, the Non K OC microcode was enabled, the Ring ratio was adjusted to x31, and I went for a CPU base clock of 131 MHz.



Then for the memory, we used DDR4-3600 CL14 which was set to a x27 multiplier, which resulted in a DRAM frequency of 3537 MHz, which isn't the real frequency as that's not how DDR memory works, but for simplicity sake let's go with that.

Finally, the load-line calibration control was set to Mode 2 and the CPU Core Voltage mode was set to override mode with a Core Voltage of 1.37v. You could probably wind this down to 1.28 ~ 1.30v without compromising stability, but we wanted to make sure the CPUs were stable for all testing.



We're certain this overclock could be better tuned for greater efficiency or a higher frequency depending on silicon quality, but the goal was to find an overclock that worked for all chips. We have three Core i3-12100 chips on hand, one is an F-SKU and all worked. We also have two Core i5-12400 chips, both worked, and a single 12700, which also worked.



For benchmarking we're using DDR4-3600 CL14 memory which has been set to DDR4-3537 using the CL14 timings. Cooling the hardware was the Corsair iCUE H115i RGB Pro XT, and the graphics card of choice for this testing was the RTX 3090 Ti. Let's jump into the data…
Benchmarks

Testing with Cinebench R23 we find some great multi-core results, though this test isn't particularly memory sensitive, so we're mostly looking at clock frequency gains in this scenario. The 12100 overclock resulted in a 25% performance boost, while the 12400 became 18% faster. The 12700 overclock was a lot less impressive as that part is already clocked quite high out of the box -- just a 15% improvement -- but again, you could probably get some chips to run at a higher all-core frequency.





When looking at single core performance, we see a 19% uplift for the 12100, and a 17% boost for the 12400, while the 12700 was improved by just 6%. It's obvious the lower-end models benefit the most from the overclock.



The 7-Zip File Manager test also shows a 23% increase for the 12100's compression performance and a massive 32% increase for the 12400. The 12700 is less impressive with a 7% improvement.



The decompression performance is much the same: a 23% uplift for the 12100, 28% for the 12400, and just 7% for the 12700.



The 12100 overclock reduced the Blender render time by 19%, meaning the overclock was 24% faster. The boost was 21% for the 12400 and 14% faster for the 12700.
Power Consumption



Because we didn't tune the voltages for our overclocks, the power efficiency is terrible, almost doubling from stock figures. Our i5-12400 wasn't stable with 1.29v, and this still saw total system consumption hit 266W, or just an 11% reduction compared to what's shown here at 1.39v, which is why we didn't bother fine-tuning voltages for this testing.

Basically, power consumption increases massively because so does frequency. In short, welcome to overclocking!
Gaming Performance

The Factorio benchmark only uses a single core and is heavily influenced by cache performance. The 12100 saw a 26% performance increase, the 12400 a 27% increase, and the 12700 a minor 6% increase. Big performance gains are seen for the Core i3 and i5 models for the most part as those come clocked lower out of the box.



F1 2021 is another good example of why overclocking these clocked Core i3 and i5 parts is so exciting. In the case of the 12100 we're looking at a massive 40% performance boost, while the 12400 enjoyed a 31% increase. Those are massive gains and are certainly far more impressive than the 4% seen with the 12700.



These results show how for many of today's games, the difference between a part like the Core i3-12100 and Core i7-12700 isn't about the core count, but rather the massive difference in clock speed, that and the difference in L3 cache capacity.



The Riftbreaker is yet another example of this. Here the average frame rate of the 12100 was boosted by 45%, while the 12400 saw a 34% increase. Meanwhile the 12700 dropped a few frames when manually overclocked, resulting in a 2% performance decrease.

What's crazy to see is that when both the 12700 and 12100 are running at 5.1 GHz, the Core i7 processor is just 4% faster and we'd say the bulk of that margin can be attributed to the much larger L3 cache of the 12700.



The Core i3-12100 already performed well in Horizon Zero Dawn, but with the overclock frame rates were boosted by a further 19% to 178 fps, which is just 5% slower than the stock 12700. The 12400 also saw a 19% performance increase hitting 190 fps and now we're looking at Core i7-like performance.



Far Cry 6 is a lightly threaded game that relies heavily on single core performance. This means overclocked the 12100, 12400 and 12700 all delivered very similar performance with just 5% separating the Core i3 and i7 models.



We know Shadow of the Tomb Raider likes core and cache, so this time the 12100 can't catch the higher-end CPUs, but still a 28% performance bump is nice and it meant frame rates were up well over 100 fps at all times in our test.

The 12400 also enjoyed a strong improvement via the overclock, boosting performance by 30%. But again we find an example where the 12700 was slightly slower, dropping just a few frames when overclocked.



Cyberpunk 2077 performance was more limited with the 12100 and it appears as though clock frequency wasn't the primary bottleneck as the overclock only improved performance by 11%. Meanwhile, the 12400 saw a 20% uplift, though nothing was gained from the overclock with the 12700.



Watch Dogs: Legion is a CPU demanding title and as you can see prior to the overclock the Core i3-12100 was only just able to keep 1% lows above 60 fps, but when overclocked performance was boosted by 30%, resoluting in 82 fps for the 1% lows and an average frame rate of 110 fps.

The 12400 enjoyed similar performance gains, 1% lows were increased by 34%, while the average frame rate was boosted by 32%. But once again we find the 12700 gains very little from the overclock, we're looking at basically no performance improvement.



Last up we have Rainbow Six Siege and all three CPUs had no issue with this older title. Still, if you are after more frames our 12100 overclock boosted performance by 33% and the 12400 overclock resulted in a 16% improvement.

The Core i7-12700 results are unexpected as we only found about a 15% improvement in productivity performance from our overclock. Most of the games showed little to no uplift, but then RSS saw a 24% performance uplift and we suspect this is largely due to the increased cache and memory bandwidth as this game is very memory sensitive, so the results make sense.
What We Learned

Those results were certainly impressive and we think the Core i5-12400 is the sweet spot for this board, though the Core i7-12700 also makes sense. We say this because the B660M Mortar currently retails for $160 and MSI is expecting the new Max version to come in only a little above that, so we're hoping for $170 at most.



The catch is then timing. MSI is aiming for an August release date, which isn't far away but by that point we'll be getting closer to next gen CPUs. In short, the Mortar Max is an exciting product, but we fear it's going to arrive a bit too late for many consumers.

Surely, if you happen to find yourself in need of a new PC around that time, the 12100, 12400 and 12700 are great options and sticking one on the Mortar Max, assuming the board costs less than $200, it would be a ripper of a deal.



In comparison, the Ryzen 5 5600 on the MSI B550M Mortar, which is a motherboard of comparative quality, that combo would set you back $300. We suspect the 12400F on the B660M Mortar Max will be about $40+ more, but the ~15% price premium will be worth it after seeing the results today.

Granted, we've yet to compare an overclocked 12400F against an overclocked 5600, but we expect the results to be favorable for Intel. One last thanks to the MSI motherboard team for providing us with an early pre-production board to play around with, we just wish you could buy it right now, but fingers crossed it won't take too long to hit shelves.

Post a Comment

0 Comments