Early adopters will find the 7950X3D to be fast, efficient, and a little complex
Remember when AMD first introduced the Ryzen 7 5800X3D, and there was speculation about a superchip with ALL the cores plus that fantastic 3D V-Cache? We never got that on AM4, with the Ryzen 7 5800X3D the only entry in the Zen 3 family.
Thus, the very existence of the new Ryzen 9 7950X3D seems like the answer to many a dream upgrade wish, but it’s a rather complex part that requires a bit of care to really exploit its potential – at least in these early days.
Before moving on to the current 7000X3D on Windows situation, here’s a look at the current lineup of “X” – now joined by “X3D” – processors in the Ryzen 7000 Series:
|AMD Ryzen 7000 Series Desktop Processors
|Cores / Threads
|Base / Boost
|Ryzen 9 7950X3D
|16 / 32
|4.2 Ghz / 5.7 GHz
|Ryzen 9 7950X
|16 / 32
|4.5 Ghz / 5.7 GHz
|Ryzen 9 7900X3D
|12 / 24
|4.4 Ghz / 5.6 GHz
|Ryzen 9 7900X
|12 / 24
|4.7 Ghz / 5.6 GHz
|Ryzen 7 7700X
|8 / 16
|4.5 Ghz / 5.4 GHz
|Ryzen 5 7600X
|6 / 12
|4.7 Ghz / 5.3 GHz
Please note that prices on the non-X3D parts have changed since launch, with even AMD’s direct store offering pretty significant discounts – such as a $100 off for both the Ryzen 9 7950X (now $599) and Ryzen 9 7900X (now $449).
The 7950X3D Setup Process is a Bit Involved
In the past, upgrading to a new CPU required no more prep than verification that your BIOS offered support – and simple flash if it didn’t. Next, make sure you have some fresh thermal compound on hand for the swap, and re-mount your cooler. Easy! Well, this time there are a few more steps. It’s an involved enough process that AMD offers a setup guide (link). Failure to configure the new part properly won’t prevent the CPU from working, but you won’t get the performance boost expected with this new 3D V-Cache part without jumping through a few hoops.
Basic setup instructions (the only painful one is the recommendation for a clean install of Windows) include:
- An Updated Version of Microsoft Windows 10 or 11
AMD recommends a fresh image of Windows 11 version 21H2 build 22000.1455 or Windows 10 version 1903 build 19044.2546. Windows 10 works with Virtualization-based security (VBS) running but the L3 cache reporting may be incorrect.
- The Newest BIOS for Your Motherboard
Go to your motherboard manufacturer’s website, and download the newest available BIOS for your system. It makes sense to confirm that the Ryzen 9 7950X3D and 7900X3D are supported in the motherboard CPU compatibility list prior to purchase, but you can expect that the newest BIOS available on launch day of February 28th, 2023 or later will support these processors.
- The Newest AMD Ryzen Chipset Driver
AMD Ryzen 7000 processors with AMD 3D V-Cache technology require AMD Chipset Driver 5.02.16.347 (or newer) for optimal system performance and stability. Using an older version of this driver will not enable X3D / Ryzen 9 game performance optimizations. Once the chipset driver is installed, you must restart the system.
Ok, no problems here – if you enjoy clean installing your OS. Which I think as many as 1% of people might. But you aren’t finished yet. Further instructions include:
- Make sure the Xbox Game Bar App in installed and updated
This application must be updated to the newest version (5.823.1271.0 at the time of writing). If your application is not updated to the latest version, you can update it through the following steps:
- Open the Microsoft Store. You’ll find this in your Taskbar or Start Menu tiles. Log in if necessary
- Click the three-dot menu icon. This icon is in the top right corner of the app window
- Click Downloads and updates. This should be the first option in the menu
- Click Get Updates
- Wait for First Boot Note: The very first system boot after platform assembly may take 2-3 minutes. The system may appear unresponsive at this time. This period is used to perform first-boot memory training. After this initial memory training, boot/POST/reboot times will reduce to the typical 5-10 seconds you’re accustomed to with AMD Socket AM4. Further BIOS releases will reduce this first/initial boot time.
Updating BIOS and chipset drivers are easy enough tasks, but the requirement to be on the very latest version of the Xbox Game Bar is a little unusual. Also, you will have to double-check to ensure that the chipset driver actually installed everything, and if you do any CPU swaps (if you are testing different AM5 processors) it is recommended that you re-run the chipset software. In my experience a crucial driver could be missed even after running the latest chipset installer, so double-check your Device Manager for any unknown devices.
For these initial tests – and I stress initial here as this new processor was benchmarked using very recent AGESA code that definitely feels recent (EXPO stopped working with our RAM after the flash) – we are running a new BIOS on our original AM5 test platform. This is an MSI X670E ACE motherboard, paired with a G.Skill NEO Z5 DDR5-6000 CL30 kit originally provided with the board by AMD for the Ryzen 7000 Series launch.
|PC Perspective GPU Test Platforms
|AMD AM5 Motherboard
|MSI MEG X670E ACE
BIOS v1.25, v1.65 (R9 7950X3D)
Resizable BAR Enabled
|AMD AM4 Motherboard
|ROG Crosshair VIII Hero (Wi-Fi)
Resizable BAR Enabled
|Intel Z690 Motherboard
|ROG STRIX Z690-E GAMING WIFI
Resizable BAR Enabled
|32GB G.Skill Trident Z5 NEO DDR5-6000 CL30 (AM5)
G.Skill Trident Z NEO DDR4-3600 CL14 (AM4)
|32GB G.Skill Trident Z5 NEO DDR5-6000 CL30
|SK Hynix Platinum P41 2TB NVMe SSD (AM5)
Solidigm P44 Pro 1TB NVMe SSD (AM4)
Samsung 990 PRO 2TB NVMe SSD (Z690)
|be quiet! Dark Power Pro 12 1500W
|Windows 11 Pro
|Drivers (GPU Testing)
|GeForce Game Ready Driver 528.02
Some CPU Performance Results
AMD has positioned this X3D variant of the current 16-core Ryzen processor as a gaming product, and surely that is the best use-case for such a product. However, as it will sit in one’s system as the CPU, some probing of processing performance is probably prudent.
We will get into power consumption later on, but an intriguing aspect of the Ryzen 9 7950X3D is its (significantly) lower TDP, as this is just a 120 watt SKU, with the original 7950X rated at 170 watts (and more than capable of boosting to much higher power draw under load). Anyway, bear efficiency in mind when examining these charts.
As expected, given its lower TDP if nothing else, the Ryzen 9 7950X3D falls short of the 7950X, but it’s still a 16-core part and thus well ahead of the Ryzen 9 7900X and Intel Core i9-12900KS.
The next two charts will tell a similar story:
One compute benchmark that I hoped would show some benefit to the the Ryzen 9 7950X3D’s 3D V-Cache was 7-Zip. Here, indeed, the X3D part is faster at compression than the X version – but the 7950X still wins in decompression:
Benchmarking CPUs with a GPU
You, dear reader, have probably already read (or watched video of) test results from other hardware review outlets, so I hope offer something a bit different. Maybe. Yes, I’m using the fastest GPU around right now, the RTX 4090 (it the Founders Edition), and while the few tests to follow are all at 1920×1080, I have an obsessive need for repeatable benchmarks that requires the use of built-in benchmarks – which some outlets proudly eschew in favor of custom in-game runs that are only as consistent as the game engine (and the person operating said run) allows.
I won’t argue against outlets that refuse to use these so-called “canned” benchmarks, and there are certainly some that could present certain inconsistencies (random elements that are not consistent run-to-run). Still, when comparing CPUs with the same GPU, the margin of error needs to be as small as possible. Who wants to read that one CPU is about 4 FPS faster than another, with a +/- 5 FPS margin? Not very useful. But the ideal circumstance – having the outlet test your chosen game at your preferred screen resolution at your preferred detail settings – is never going to happen, so this is all academic anyway.
Synthetic, yes. But UL’s GPU benchmark suite is the industry standard. Here’s a look at how the flagship CPUs from Intel and AMD (both AM4 and AM5) scale with our reference (sorry, Founders Edition) RTX 4090:
Uh oh. Based on this first test I might be looking at some sort of configuration issue with our AM5 platform and/or Windows 11 with the new X3D part. Mind you, these results are based on the SECOND complete round of GPU testing which was performed after seeing lower than expected results in these benchmarks, re-installing the chipset drivers, and verifying that all setup steps were completed. The only thing I didn’t do from AMD’s list was the recommended clean install of Windows. I might have to eat crow over that choice.
Here’s another example of my setup’s performance:
It’s looking more and more like the X3D results did not make proper use of the V-Cache, if the outstanding performance of the Ryzen 7 5800X3D is any indication.
To ensure that there could not possibly be ANY sort of GPU bottleneck, I also went into the archive and installed 3DMark 11, running it at both the Performance (1280×720 resolution) and Entry (1024×600 resolution) presets. And if a 1024×600 benchmark isn’t purely CPU bound, I’ll eat my hat.
That’s more like it! I guess. Anyway the Performance score still looks off (though the 7950X3D does edge out the 5800X3D), but in the rather comical by 2023 standards Entry test (again, we are talking 1024×600 here) the Ryzen 9 7950X3D is the clear winner, besting Intel’s flagship Core i9-13900KS by more than 2800 points. Now if that isn’t reason enough to spend $699 on this processor, I don’t know what is.
Next, we will have a look at some 1080/low game benchmarks. And I REALLY HOPE that the Ryzen 9 79050X3D is being shown to its full potential, and that the charts to follow aren’t just embarrassing poorly-optimized nonsense from this editor.
Those Necessary 1080/Low Benchmarks
I will begin with the one test that perplexed me, and had me questioning my settings and test methodology. It is Shadow of the Tomb Raider, configured with the preset called “Lowest”, no DLSS (TAA was selected), and the DX12 API. The benchmark was run at 1920×1080 in exclusive fullscreen mode. Both AMD platforms and the Intel platform were running the same RTX 4090 FE card, and all systems were running GeForce Game Ready Driver 528.02 (the latest version when I began testing the Core i9-13900KS).
Does this result – the average of three identical benchmark runs – make any sense? Is the Ryzen 7 5800X3D the fastest gaming processor ever made for a title like SotTR? I figured that there was something off with this result – AGAIN, possibly a lack of proper optimization with the Ryzen 9 7950X3D – even though I checked every box from AMD’s setup guide…
As a sanity check, let’s look at the results from the venerable Metro Exodus benchmark tool (the original, not the Enhanced Edition), which was run using the Low preset and DX12 API. No display scaling of any kind was active, and we are again looking at 1920×1080 fullscreen results using the RTX 4090 FE and GRD 528.02:
An impressive showing for the AMD X3D processors from both generations, with the nod clearly going to the new Ryzen 9 7950X3D here (finally).
Next we look at Cyberpunk 2077, version 1.61 (yes, version matters with that game). This was run at the Low preset, no display scaling of any kind enabled, and we are once again running at 1920×1080 in fullscreen mode.
The Ryzen 9 7950X3D just slips past Intel’s Core i9-13900KS here (maybe the system IS configured correctly??), but Intel has the edge in frametime consistency in this test.
Finally, a look at a less demanding graphics workload via the Final Fantasy XIV Endwalker benchmark tool, which was configured with the “Standard Desktop” preset and run at 1920×1080 fullscreen.
We are well into mid-300s here in this lower-complexity test (we are using the “Standard Desktop” preset after all), with the top result going to Intel’s Core i9-13900KS at an average FPS of 363.6, wich is 11 FPS faster than the Ryzen 9 7950X3D and 20 FPS faster than the Ryzen 7 5800X3D.
As mentioned earlier, these results are simply academic – and very possibly flawed! Still, who in their right mind would spend $699 on the latest and greatest AMD gaming processor, buy a Founders Edition RTX 4090, and run at 1080/low? How many people even game at 1920×1080 on RTX 4090 systems (other than esports pros)? If you game at 2560×1440 or higher, it becomes far less important which one of these flagship gaming CPUs you choose, as higher resolutions and detail settings place far more weight on GPU performance – and less on any CPU bottlenecks.
The staggering difference between Intel and AMD power consumption at the high end this generation is nothing new, and power draw from this Ryzen 9 7950X3D is one of the obvious highlights of this new part. It has a lower TDP than the non-X3D model, and even peak power consumption was under 150 watts in my testing. To compare this to the power consumption I recorded using the Core i9-13900KS is just… well, it’s just not fair.
The disparity between this new Ryzen 9 7950X3D and the Core i9-13900KS is just staggering. The 7950X3D does consume nearly 150W under a full all-core load like this, but this is well below the non X3D version, which is right in the middle here.
Another benefit of the lower power draw? The same cooler that saw the 7950X consistently hit temps in the mid-90s was able to keep the 7950X3D to a max of 75 C through all testing. That’s 20 C cooler, and no throttling! YES.
Conclusion (for now)
AMD’s Ryzen 9 7950X3D marks the beginning of the X3D era for the AM5 platform, and it’s a platform that needs it. The sheer brilliance of AMD’s Ryzen 7 5800X3D processor, the mature AM4 platform, and inexpensive DDR4 prices conspire against AM5. That probably isn’t good news for AMD, considering how many existing AM4 board have (or will get) an X3D upgrade of their own, but that’s what happens when you make a generational part like the 5800X3D.
My own journey with the new Ryzen 9 7950X3D was a rocky one, and even after “perfectly” optimizing my test system I was still seeing results that, frankly, suggest a configuration problem. With better optimization that I was able to achieve, I suspect that the 7950X3D will end up being an even more impressive part than I already think it is. The results that I achieved were still quite good, and efficiency of this new 120W part is unparalleled at this performance level.
I look forward to re-testing this part on more mature firmware and with a clean install of the very latest build of Windows 11. The story is really just beginning – particularly as the Ryzen 7 7800X3D looms on the horizon. That single-CCD X3D part could be the one. I’m hopeful it will be easier to set up, at least.