3 September 2024

Is the PS5 Pro really equivalent to an RTX 4090? (No!)


Playstation 5 Pro fever is nearing its zenith, with speculations becoming rife and consumers beginning to fantasize about what, exactly, the new console will be capable of. This is a fun look at some of the recent nonsense coming out of the Twittersphere on the subject:



Captured for posterity before "Pyo" comes to his senses...



As POWERFUL as the RTX 4090...


As easy as it might be to discredit such a claim or "imasgibation"* where a graphics processing unit is compared to a fully functional Death Star I'm not that sort of commentator. Let it never be said that Duoae was a man who minced words but not numbers whilst navel-gazing!
*I just invented this word, I hope it becomes a new meme...
Previously, I have already simulated the PS5 Pro, using a Ryzen 5 4600G, 16GB of DDR4 3200 RAM, an SSD, and an RX 7800 XT. Both the CPU and GPU were clock-locked to the specifications of the PS5 (and rumoured Pro) - which I should note are both below the stock specifications for the PC parts!! - and the analysis can be found in that link just above.

What I can see from my testing is that the clock-locked RX 7800 XT performs approximately equivalently to the stock RX 6800 in both Ratchet and Clank and Starfield. So, using some newly-generated data I have at hand, I will use this card as a comparison point...

Unfortunately, I don't have an RX 6700 10GB to compare against (for the base Playstation 5 system), which would be nice... and buying one has become a very expensive proposition - so, given I'm self-funded, that's likely to never happen, now. I also don't have an RTX 4090 to be able to pair with the 4600G to test this out but I DO have an RX 4070 Super, which, while not as powerful as an RTX 4090, is more powerful than the RX 7800 XT (at stock settings) in both raster and when using ray tracing. 

So, let's take a look at what this looks like!


Boom! Tested at 1080p with high/ultra settings in each game (except Alan Wake 2 which used the low RT preset)...


That's not great...

In almost all scenarios, even though both GPUs will likely be suffering from some form of CPU bottleneck (see below) the games running on the RTX 4070 outperform the same sections tested on the RX 6800.

Looking at the GPUs in a selection of titles being run on a number of different CPUs, it becomes clear that the RX 6800 actually isn't even that CPU-bound by the lowly 4600G at the high settings tested and is, instead, limited by its graphical abilities! 




Even here, the RX 6800 is less CPU-bound than the RTX 4070 Super...

Now, sure, you could argue that these settings would never be used on a console - and you'd be right! However, that just raises the possibility that the RTX 4070 Super would pull even further ahead, anyway!

While I don't have that specific data, we can simulate such a situation by comparing the percentage increase for each graphics card when going between the anaemic Ryzen 5 4600G to the relatively powerful i5-12400.


Even a CPU-boost doesn't really help the RX 6800 perform much better in the majority of titles...


We can see that the RTX 4070 Super, on average, benefits much more from the increase in CPU power, meaning it has a lot left in the tank, graphically speaking... and these are not clock-limited CPUs, we're talking about - so, scale back that 4600G performance by another small margin!

So, no, the PS5 Pro will not have the graphical horsepower of the RTX 4090. 


Imasgibation debunked!



Captured for posterity...


The PC Equivalent to the PS5 Pro...


Now, it seems clear to me that Hugh is not a sincere person and may, in fact, be having a little fun given his profile is labelled as "parody". However, enough people re-Tweeted his reaction to Wccftech's article that I felt I could include it here.

I think the second thing to note about this PC build is that it assumes a few key things:
  1. You can squeeze an AM4 CPU into an AM5 socket.
  2. You don't know how to budget.
  3. You know nothing about component performance.
Getting those out of the way, you realise that you can do far better than ol' Hugh in terms of cost! Here's my 5 minute white build, which would blow both the PS5 and PS5 Pro out of the water! (And also looks better, IMO!!)


All of these are around $850... so, that's a pretty big saving! (Actually, around the cost of a base PS5!)

But let's get even more specific! The 5700X is a beast of a CPU compared to the measly 4600G/4700G. It's not only a better, more performant core architecture, it also has a larger L3 cache (important for gaming!) and a faster boost clock frequency. The CPU also operates at a much higher TDP than the silicon in the PS5/Pro will have available to it... so, even ignoring the frequency differences, the APU will be throttling in order to keep itself working within the tight thermal and power limits applied to it!

The RX 7700 XT also has a similar issue, in that its frequency and power/thermal limits are higher than that of the GPU in the PS5/Pro. It also has an infinity cache (which the APU doesn't), larger L0 and L1 caches but a smaller L2 (half the size). The desktop GPU also has a smaller VRAM capacity and lower bandwidth to that VRAM - so, that's a bit of a wash.

The box cooler, given away with the CPU, will be sufficient enough to cool this beast - so no need for an expensive tower cooler - the same logic applying to the case, RAM, SSD, and power supply, these are also not needed unless you specifically want to make your CPU as expensive and small as possible, in which case, there are better options!

Finally, everyone and their mother knows that there is pre-applied thermal paste on all box coolers (or it comes for free with a cooler you purchase).

Really and truly, this was less of an imasgibation and more of an outright troll which got everyone talking. In which case, well-played!


Fin...


If you've seen any other such PS5 Pro type posts, feel free to link them below and I might do a follow-up post dissecting them!

No comments: