Intel's new 200K Plus chips support ultra-fast DRAM out of the box but as my tests show, there's little benefit for most PC gamers in using warp-speed stuff

If you've been keeping up with Intel's latest processors, you'll know that with its Core Ultra 200S chips, it set a new standard for speedy DRAM support. Starting at 4,800 MT/s with the 12th Gen Core range, each new series of processors has steadily increased the speed of the memory controller, reaching 6,400 MT/s with Arrow Lake. Now it's gone a step further with the Core Ultra 200S Plus CPUs: a frankly ridiculous 7,200 MT/s.

These are all 'default' data transfer rates, i.e. without overclocking the memory controller, but as any PC gamer knows, most chips these days happily support DRAM modules that have XMP or EXPO profiles. These are charts of settings for the memory chip and integrated controller that let you push things even further.

However, as I've shown earlier this year, the clock speed of DDR5 doesn't make a huge difference in games, and when the Core Ultra 200S chips first appeared, ultra-fast DRAM barely made any difference at all. Even DDR5-8400.

So, is that still the case with the new Core Ultra 200S Plus processors? To find out, while collecting data for my review of the Core Ultra 7 270K Plus, I used two different memory kits: a 32 GB set of Corsair Vengeance RGB DDR5-6000 CL30 and a set of G.Skill Trident Z5 DDR5-7200 CL34, also 32 GB in capacity.

As for the rest of the test hardware that matters, the motherboard was an MSI MEG Z890 Ace, and the graphics card was a Zotac GeForce RTX 4070. Oh, and I retested the Core Ultra 9 285K, just to see if anything had changed since it appeared in October 2024.

It's only five games' worth of data, but I think it's fair to say that faster DRAM doesn't really make a whole lot of difference, as anything under 5% can be considered to be nothing more than a margin of error in the test runs.

Well, certainly not in four of the games I tested, and absolutely not for 285K. However, the 270K Plus did run Baldur's Gate 3 up to 4% faster on average, DDR5-7200 vs DDR5-6000, and the 1% low frames rate picked up a healthy 8% boost.

The thing is, DDR5-7200 is clocked 20% higher than DDR5-6000, so why aren't we seeing that kind of performance improvement? The answer to that lies in the fact that there are a lot of stages in the processing of a single frame, from beginning to end, and the speed at which all of this is done is primarily limited by the slowest element in the sequence.

If that's not related to the system memory in any way, then no amount of extra DRAM speed will make a jot of difference. However, some elements are related to memory, even if it's not immediately obvious.

For example, if a thread instruction demands a piece of data that isn't present in any of the CPU's caches, then the chip will have to request it from system memory. This is why AMD's X3D processors are generally not sensitive to DRAM speed, because they have whopping amounts of L3 cache.

However, the 270K Plus and 285K have exactly the same amount of L1, L2, and L3 caches, so why is the new chip running better than its very closely related cousin in Baldur's Gate 3?

The answer almost certainly lies in the fact that Intel increased the internal clock speeds for its Arrow Lake refresh. The 270K Plus' Die-to-Die (D2D) and Next Generation Uncore (NGU) clocks are 43% and 15% higher, respectively, compared to the 285K's.

That means the 270K Plus is able to shift data between and inside the tiles, which make up the whole processor, a little bit faster and with a reduction in latencies.

Out of everything I tested, Baldur's Gate 3 is by far the most cache/memory dependent, which is why the Ryzen 7 9800X3D massively outperforms any of Intel's processors in this game. Homeworld 3 is too, though to a far lesser extent. The other games care more for a processor's IPS rate (instructions per second) than anything else.

The exception to this would be when you've got something like an RTX 5090 but you're using it to render low-resolution, low-graphics esports games to get the highest possible frame rate. In those situations, speedy memory is likely to make a difference, but then again, you're probably using a 9800X3D too. If so, then warp-speed DDR5 won't make a difference.

Buying ultra-fast DRAM for your gaming PC only makes sense if the rest of the setup is such that a game's performance is entirely limited by system memory performance. If it's not, then just save your pennies and stick to the cheapest set you can find that isn't rock-bottom slow. Then again, DDR5-7200 isn't vastly more expensive than DDR5-6000.

With regards to the two DRAM kits I used, you'll currently need to hand over $510 at Amazon for the Corsair set and $550 at Amazon for the G.Skill stuff. If you don't mind a slightly higher CAS latency (and it really won't affect games that much), you can get 32 GB of DDR5-6000 CL36 for $370 at Newegg, or $430 at Newegg for a set of RGB-less DDR5-7200.

You'd expect DDR5-7200 to be monstrously expensive in today's global memory crisis, but since AMD's Ryzen chips are happiest with DDR5-6000, there's far less demand for the super-speedy stuff. So much so that you can pick up a 32 GB kit of DDR5-8000 for $540 at Newegg. Not that you should, of course.

Anyway, I digress. Summary time. With each new generation of processors, Intel has improved the support for ever-faster DRAM, without recourse to overclocking the controller. Top-speed memory modules aren't necessarily that much more expensive than 'normal' speed stuff. However, faster DRAM barely makes any difference in games, unless they're very DRAM-dependent, so it's not worth buying.

As the Talking Heads song goes, same as it ever was.