AMD's Ryzen 7 9850X3D could save you hundreds on your new build during the RAM crisis — CPU tech nearly eliminates the performance difference between cheap and expensive RAM

News
By published

Basically identical performance, and potential cost savings.

FPS difference between DDR5-4800 and DDR5-6000 on AMD Ryzen 7 9850X3D
(Image credit: AMD via Videocardz)

AMD announced its new Ryzen 7 9850X3D chip at CES 2026, showcasing a 7% performance uplift over the iconic Ryzen 7 9800X3D, one of the best CPUs. What the company didn't show in that presentation is that you don't necessarily need high-speed DDR5 RAM to achieve those gains. In a marketing slide for the CPU obtained by Videocardz, the company says there's only a 1% difference in FPS between DDR5-4800 and DDR5-5600.

This has been true for X3D chips since the beginning because they leverage the extensive on-die L3 cache, which allows the CPU to access RAM less frequently. Managing latency is key in gaming workloads, so the L3 cache is preferred over system memory in most cases. Since the cache is shared across units of the same SKU, it becomes the bottleneck rather than the RAM you have.

Image 1 of 2
Ryzen 7 7800X3D gaming benchmarks
(Image credit: Future)

On average, there's a less than 1% difference in FPS, and the company even includes current pricing for 32GB kits at the bottom. According to AMD, the average price for a DDR5-4800 kit is $400, while the DDR5-6000 kit can run you up to $470. Both of those are very generous estimates, as we were able to find a 32GB DDR5-6000 kit for "just" $325.99 on Amazon, though it's CL36. A similarly spec'd CL30 kit is $354 on Newegg.

Given the AI boom and its consequences on the industry, it makes sense for the Red Team to market this point. The chipmaker has never explicitly mentioned it before, but we did in our 9800X3D review when it came out. So, if you have bone-stock DDR5 lying around without EXPO or XMP, the X3D chips will give you the same glorious performance as someone with an overclocked kit.

Google Preferred Source

Follow 3DTested on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Hassam Nasir
Hassam Nasir
Contributing Writer
23 Comments Comment from the forums
  • JayGau
    You can't evaluate the performance gain with RAM speed by only looking at the average fps. 1% and 0.1% lows have to be taken into account. Pretty much every test I have seen in the last few years shows very little improvement on the average fps, but the 1% low can benefit much more from the RAM speed.

    That's relevant even at higher resolution, where the average fps is mostly limited by the GPU, but the 1% low fps is still affected by the CPU performance. And fluctuating 1% low is something you absolutely want to avoid (if the word "stuttering" means anything to you).
    Reply
  • thestryker
    AMD marketing isn't doing the company any favors lately. First they made a hilariously awful slide comparing PTL to their offerings and now this. As mentioned above lows are where DRAM differences tend to be most notable. They also don't mention the latency for either kit which makes the comparison even more useless. Most DDR5-4800 kits have rather high latency out of the box.
    Reply
  • Gururu
    The pain is in the ddr5 leap.
    Reply
  • Kerbs07
    Sorry, but in reality this is negligible and you're likely to not get any benefit. Particularly when they're showing 1080p results and whoever is running one of these CPUs probably will be on 4K or 1440p. This just sounds like marketing from AMD.

    If we get decent results at modern resolutions then yeah but not point for running a 1080p monitor. Just being realistic here. We already knew what difference RAM makes. Not sure what AMD is trying to say.
    Reply
  • ohio_buckeye
    Can’t say I agree with this article. Received an email from Newegg the other day showing 16gb ddr5 on sale for only $250. Ill stay with my LGA 1700 ddr4 build for a bit.
    Reply
  • usertests
    This theoretically works, it's a known benefit of X3D... But you need to buy a $470-500 CPU (if you're looking for the best one).

    The savings highlighted by AMD is only $70. If there's a cheaper 6000 MT/s kit, what about 4800 MT/s garbage tier?

    Micro Center has been peddling bundles with single-channel memory, such as 7500X3D + motherboard + 1x16GB for $300. If X3D can overcome the bandwidth being slashed in half, that'd be impressive.
    Reply
  • dmitche31958
    JayGau said:
    You can't evaluate the performance gain with RAM speed by only looking at the average fps. 1% and 0.1% lows have to be taken into account. Pretty much every test I have seen in the last few years shows very little improvement on the average fps, but the 1% low can benefit much more from the RAM speed.

    That's relevant even at higher resolution, where the average fps is mostly limited by the GPU, but the 1% low fps is still affected by the CPU performance. And fluctuating 1% low is something you absolutely want to avoid (if the word "stuttering" means anything to you).
    Spot on. If your going to make a claim they should have done sons benchmarks on productivity apps that in most cases are not GPU bound.
    Reply
  • -Fran-
    Le sigh...

    For non VCache'd CPUs:
    Fr7Bfr-wPYw View: https://www.youtube.com/watch?v=Fr7Bfr-wPYw

    For VCache'd CPUs:
    aD-4ScpDSo8 View: https://www.youtube.com/watch?v=aD-4ScpDSo8

    It's been quite a while since it was proven the effects of lower speed RAM affect the VCache'd siblings A LOT LESS. Yes, they still get a hit, but the VCache helps with it. Only standout game from memory is SpiderMan, which really likes memory bandwidth to a weird degree.

    Also, this CPU is not aimed at productivity. For that you have either the 9950X or its VCache'd sibling.

    I couldn't find their test with the 9800X3D, but I'm guessing they haven't done it, since it's already KNOWN.

    This is to say, for once, AMD's marketing is not flat out wrong or deceptive.

    Regards.
    Reply
  • Gururu
    -Fran- said:
    Le sigh...

    For non VCache'd CPUs:
    Fr7Bfr-wPYw View: https://www.youtube.com/watch?v=Fr7Bfr-wPYw

    For VCache'd CPUs:
    aD-4ScpDSo8 View: https://www.youtube.com/watch?v=aD-4ScpDSo8

    It's been quite a while since it was proven the effects of lower speed RAM affect the VCache'd siblings A LOT LESS. Yes, they still get a hit, but the VCache helps with it. Only standout game from memory is SpiderMan, which really likes memory bandwidth to a weird degree.

    Also, this CPU is not aimed at productivity. For that you have either the 9950X or its VCache'd sibling.

    I couldn't find their test with the 9800X3D, but I'm guessing they haven't done it, since it's already KNOWN.

    This is to say, for once, AMD's marketing is not flat out wrong or deceptive.

    Regards.
    I love how the 14700k shows the same effect on memory speed as the 7800X3D. Only 4% difference for both.
    Reply
  • thestryker
    -Fran- said:
    This is to say, for once, AMD's marketing is not flat out wrong or deceptive.
    Everyone knows the impact isn't as big on 3D V-cache, however their marketing is absolutely deceptive. They're not showing lows and the timings used aren't reported.

    FWIW HUB's written form is a much simpler reference: https://www.techspot.com/review/2866-ddr5-ram-stock-vs-xmp-expo/
    -Fran- said:
    Only standout game from memory is SpiderMan, which really likes memory bandwidth to a weird degree.
    Hogwarts and memory latency.
    Reply