Here I’ll give you a slightly different take on the recently released Ryzen 7000 X3D chips. Yes, they are fast and in many places compete for the first spot in performance but there is much more to it which I feel is not getting the attention it deserves. Over the past few years we have been getting huge performance jumps from newest generation chips, but there’s also been a massive shift in power consumption of these chips. It’s all good and well to be the fastest but at what cost.
With the energy prices being beyond affordable, and concerns around the need for higher-end motherboards that can support so much power - all of this drives up the costs as well as cooling requirements. So for this video I want to not only look at raw performance, but rather what it is like to run X3D and also highlight who these chips are for.
Quick note - for now AMD has only released their 12 and 16 core chips, their more budget friendly chip is due to come out early April so we will have to wait and see how that one stacks up. For this video we will be working with the top of the line 7950X and its new brother 7950X3D.
Let me quickly explain the design of the new X3D chips and what’s so special about them. Since the previous generation, AMD has been able to increase the Level 3 cache on the chip by attaching extra memory directly on top of the die, hence the name 3D. In the previous generation, this was only implemented on the 8-core 5800X3D chip, which features a single CCD. However, there are some limitations to this approach, mainly due to power restrictions. As a result, the chips run at lower voltage, which consequently leads to lower frequency than their non-3D counterparts, and there is no option for manual overclocking.
With the new CPU’s AMD also installed the cache on one CCD and left the second one with a more conventional design. This arrangement allows one CCD to have more cache and lower frequency while the other has higher frequency but lower cache, resulting in a potentially optimal configuration. AMD also provides users with some additional adjustments for these chips, including the ability to enable PBO and utilise their curve optimizer.
Due to both CCD’s being different, AMD has also released a bunch of tools within the chipset drivers to help Windows select the right CCD for the optimum performance. Without going too deep into the details, there are optimisations concerning parking and enabling cores for different workloads and so on - if you get these chips, make sure to install the full list of chipset drivers and keep it updated as there are likely to be improvements and bug fixes coming for a little while. It is recommended to set your Windows power plan to balanced mode rather than performance mode to allow the new scheduling tool to function effectively.
With the summary out of the way - let’s jump into some benchmarks. For these tests, we'll be using the Gigabyte X670E Aorus Master Motherboard, Trident Z Neo memory clocked at 6000 M/T, RTX 4090 GPU and version 2 of the ASUS Strix 360mm AIO. In most of the tests, we will be comparing the stock 7950X to the 7950X3D with and without PBO and curve optimizer enabled.
Let's begin with Borderlands 3. At 1080p resolution, we observed a 3% improvement in average FPS when upgrading to the 7950X3D, and an additional 2% improvement when the chip was optimised. As for 1 percentile FPS, there was a 3% improvement without optimization and a further 6% improvement with optimization.
One interesting aspect to note is the performance per 100 Watts. Usually, we would observe a low value for this metric in CPU benchmarks, which is why we compute it per 100 rather than per 1. However, in this case, we observed incredible performance from the 7950X3D chips in terms of power efficiency. In fact, in this scenario, the 7950X3D was 86% more power-efficient at stock and 90% more power-efficient with optimizations. That is a huge deal!
When we go up to 1440p we see a similar story, but with 11% improvement on 1 percentile performance at stock and 13% with optimisation. Here we have about 92% improvement in power efficiency.
At 4K resolution, we found that the 7950X actually had the highest average FPS, taking the lead by 6.6%. However, it still lagged behind on 1 percentiles and power efficiency. It is worth mentioning that this is the only game where the 7950X3D did not outperform the 7950X.
Moving on to Horizon Zero Dawn, at 1080p, we observed only a small 3-6% improvement in FPS, but a significant 61% improvement in power efficiency. Similarly, at 1440p, we saw even less improvement in FPS, but still approximately 61% improvement in power efficiency. And lastly at 4K there is essentially no FPS improvement as we are completely bottlenecked by the GPU but power efficiency is still present at 64% improvement.
This is really important not just because of power efficiency but also cooling requirements - less power means a cooler chip, but we will get into that a bit later on.
Next game is Shadow of The Tomb Raider and here at 1080p we have probably the biggest performance uplift of 23% while at stock and 26% with some optimisation. When it comes to power efficiency we are at double the frame rate per Watt when compared to 7950x and that is while having higher FPS.
At 1440p we have much of the same with 22-24% improvement on average FPS and 10-12% improvement on 1 percentiles, while still having double FPS per Watt. At 4K resolution, we noticed a slight shift, with a 16-17% improvement in average FPS and a 16-19% improvement in 1 percentiles with 92-94% power efficiency improvement.
We also benchmarked World War Z and there it was between 5 and 10% improvement on average FPS and 5 to 13% improvement on 1 percentiles across different resolutions with 60-75% power efficiency improvement.
We can go on like this, but this is getting old pretty quickly - it seems like the x3D chip is generally performing better in most cases, with significant improvements in power efficiency across the board. There are some exceptions, such as in Borderlands 3 at 4K resolution, where the 7950x performs better. However, it's clear that the x3D chip is delivering impressive results overall, particularly in terms of power efficiency.
Let’s now jump into the production workloads where the previous 5800x3D chip did not really perform much better than its non 3D counterpart and right off the bat in Vray we see x3D chip losing out to its now cheaper alternative.
On the other hand when we run a more complex set of tests in Davinci resolve from PugetBench, there is performance to be gained from 7950x3D with PBO enabled, but do note that if we throw in a curve optimiser it loses out to the non 3D variant.
In our short Blender benchmark, we see that having more cache does not necessarily translate to better performance as the CPU runs only slightly slower.
Same goes for our longer custom Blender render. Here the more standard chip renders the scene 6.5% faster but that is not the full story, let’s dig a bit deeper.
When we look at CPU power during this test, we see a pretty drastic difference between 7950x and 7950x3D. That is over 80 Watt difference which is about 55 to 60% more power for only 6.5% improvement. For me personally that is not worth it.
One area where the new Ryzen 7000 chips have faced criticism is their cooling performance. These chips are designed to maximise their performance by running at high temperatures of up to 95 degrees when under heavy load. The extent of their boost depends on the cooling provided. However, the new 3D chips consume less power due to their lower voltage, and as a result, they don't boost as high. Consequently, they run cooler. In this example we hit around 80 degrees and stay there.
This brings us well to the conclusion. The latest release from AMD offers a highly competitive chip for high-end gaming, with exceptional power efficiency. Despite its emphasis on gaming, this chip's large number of cores enables it to handle production-focused applications with relatively small performance loss. So if you are in the market for a chip that can handle both gaming and can do some projects on the side - this is a good, but expensive option. Alternatively if you're strictly interested in gaming, waiting for the 7800x3D release in April might be a better option. That chip is expected to perform similarly but at a significantly lower cost.
Lastly if you're looking to save some money, consider looking for an older 5800x3D chip. These chips are still available in some stores and are often heavily discounted. By pairing them with last generation motherboards and RAM, you can build an impressive system at a much lower cost. I hope you found this information helpful!
Comments