The impressive part is that it has IPC on par with Raptor Cove, what’s holding it to Skylake performance is the low max clocks.

It literally has higher IPC than Zen4, which is an impressive showing for Loongson. I’m guessing it’s a very wide design that has to stay at low clocks because it is hampered by the relatively ancient 14nm process it’s manufactured on.

You’ve gotta give credit where credit is due.

The display supports it’s full resolution and refresh rate over HDMI 2.1. It’s just not working right now at full rate over HDMI with either AMD or NvIdia graphics cards for some reason.

The table included in this article is misleading, as they’ve cropped off the original Korean text which states that it’s currently not working at 240Hz over HDMI 2.1 with AMD either but that they contacted AMD who said it would work with a future driver update. The only reason that NVIDIA is listed at 120Hz in that table is because Quasarzone didn’t get a reply from NVIDIA in time for publication.

It’s unclear why no cards can do 240Hz over HDMI with that monitor when it’s in spec.

That monitor theoretically supports it’s full resolution and refresh rate over HDMI 2.1 too, the extra bandwidth isn’t the difference maker.

Right now no cards seem to work with it at 240Hz over HDMI, but it’s listed as supported on the monitors end.

Nothing has changed about 7950X3D scheduling since launch.

It’s just a matter of how many games are included in the average where the gamebar parking doesn’t work.

By “affinity controls” I just meant some way of assigning core affinity (cores that the process is allowed to run on) to processes. The most popular way to do this is process lasso, but you can even do it with the inbuilt Windows Task Manager so I used a generic term so it was clear the requirement wasn’t a specific piece of software.

Process Lasso is good, but it is also very bloated. It has a lot of unnecessary functions that you won’t need to use, and it also automatically changes a bunch of power settings to “optimise your PC”.

For example, by default it will disable core parking whenever steam launches a game because on other platforms that is a potential performance boost. But with these CPUs it means you have to either change that setting or always use affinity, because it will break the gamebar parking.

If you don’t pay for process lasso it will start nagging you every time it opens with a popup that can’t be dismissed for 30 seconds, so it’s really only viable if you’re willing to shell out the cash for it (or find it through alternate means).

The biggest upside for process lasso is that it is persistent, if you set a rule for a process it will stay that way every time you start it. It is annoying to set the rule for the first time, you have to search up the process and then manually tick the cores you want to bind it to, but it sticks. Process Lasso also would let you bind your background productivity processes to the non-vcache cores to free up the vcache cores more.

By comparison, CapframeX’s affinity control is simple but needs to be set every time you run the game. It’s just a simple button bind, I press “Numpad *” to cycle between all cores/cache cores/frequency cores for the currently active process.

​

It’s important to remember that even if you forget to set affinity and gamebar isn’t working, the 7950X (which you basically have in that situation) is still a good gaming CPU. It’s not quite as fast in games as 13th Gen Intel or Zen4X3D, but it’s still the next best thing.

Extra RAM would be if they used out-of-band ECC, which is where you use extra physical dies and a wider bus to handle the storage and transmission of parity data.

These use in-band ECC, which uses the existing memory chips/bus for parity. You have to toggle it in the driver, and with ECC enabled you don’t get the full memory size nor the full bandwidth (you lose 1/8th of each). As far as I’m aware, no GDDR6 GPUs use out-of-band ECC.

It’s purely a market segmentation thing, there is nothing physically stopping the RX7700 from running the exact same ECC mode as the W7700.

It has the same physical memory as the consumer cards, it just allows in-band ECC which is a software solution. There is no additional manufacturing cost to including in-band ECC. Nvidia even enables the option on the 4090.

You’re right that you’re paying for the certifications though.

1. You don’t necessarily need to worry about affinity, but if you want to get X3D performance in every game then yes some sort of affinity software is required. Not all games work with gamebar, even manually adding them to the game detection, and some games that do will still have issues because it turns off the parking if the load on CCD0 gets too high.

I personally use CapframeX for affinity as process lasso causes some games to crash (like BG3).

2. High performance mode disables core parking, which breaks the software. You need to use balanced.

3. The built in AMD solution is only really sufficient if you don’t want to use intensive software while gaming, and it won’t work for all games even then.

4. If you want to be using the extra 8 cores for productivity while gaming, you’ll need to use affinity instead of gamebar. Parked cores run no code, and running multithreaded workloads while gaming will immediately trigger the “disable parking if load is high” rule and allow the game to leave the VCache die which hurts performance. If your workload is gaming plus MT in the background, you will need to use affinity controls 100% of the time.

5. It’s an amazing CPU, I really love it. It is flawed though, the gamebar solution is incredibly half-arsed from AMD. I don’t think they could’ve gone for a lower effort, more fragile method than just “when gamebar says a game is open, turn off half the cores”. I’m fine with manually doing affinity because I honestly enjoy the tinkering, it’s fun to benchmark whether games (and indeed other applications) prefer clocks or cache. When using affinity, it’s the best CPU in the world for mixed gaming/productivity.

6. For your workloads, the 7950X3D makes more sense. It’s a monstrous CPU for MT, and if you get benefit from more MT I wouldn’t settle for the 7800X3D.

One other thing, I’d recommend getting 64GB now if you can afford it. 4 dimm support is atrocious, so if you ever need to upgrade ram you’ll want to instead take out the 32GB and put in 2x32GB. The price jump to 64GB is relatively pretty small.

Anandtechs testing is completely wrong because they used TDP on Ryzen and PL2 on Intel. PL2 is the actual power limit, but TDP isn’t. PPT on AMD is 1.35x TDP and is the actual power limit, which is why the measured power is much higher than the limit they set.

It’s absurd that they’ve never issued a proper correction for an article that has the AMD CPUs allowed to draw 35% more power at every comparison point. It’s been misleading people for years.