AMD's new Zen 4 flagship, the Ryzen 9 7950X is a 16-core, 32-thread CPU that clocks between 4.5 GHz and 5.7 GHz depending on the load, with two Core Complex Dies (CCD) along with a 6nm I/O die. The additional CCD compared to more affordable Ryzen models means the total L3 cache capacity is increased from 32MB to 64MB, and now there's a total of 16MB L2 cache.
In case you missed it, we already checked out the 6-core Ryzen 5 7600X yesterday, with the 7700X and 7900X reviews coming over the next few days. Compared to the mainstream 7600X, the TDP is increased from 105w from a single CCD to a whopping 170w, though do note because the same I/O die has been used the same 28 PCIe 5.0 lanes are supported.
The Ryzen 9 7950X is set to replace the well-regarded 5950X, but quite surprisingly the launch price or MSRP has been dropped from $800 the previous generation to $700 for the 7950X -- that's a healthy 13% discount for what should be a much faster processor.
Bargain shoppers and current AM4 owners will want to know though that at present the 5950X can be had for as little as $550, so the newer Zen 4 16-core model is coming in at premium, of course.
There's a lot more to go over and a ton of testing and benchmarks as usual, we've been busy over the past few weeks updating all our CPU data and as a result we have 17 CPUs for comparison.
For the AM4 platform we've got a range of Zen 2 and Zen 3 CPUs with popular models such as the Ryzen 9 3950X, Ryzen 7 5800X, and of course, the 5950X is also included, but there are half a dozen more models also included. We have included important Intel 10th-gen Core series models but due to time constraints we decided to skip the 11th-gen. Then for the 12th-gen Alder Lake CPUs we've tested using DDR4-3200 dual-rank CL14 memory and DDR5-6400 single-rank CL32 memory and we have from the Core i3-12100 up to the Core i9-12900K for comparison.
The new AM5 test system is based on the MSI MEG X670E ACE, for which we followed AMD's guidelines to use the supplied DDR5-6000 CL30 memory. All testing was conducted using the GeForce RTX 3090 Ti, Windows 11, and Resizable BAR enabled for all configurations.
Here's a look at the Ryzen 9 7950X's clock behavior in Cinebench R23. After an hour of load testing, the CPU maintained an all-core frequency of 5.1 GHz on the primary CCD1 and 4.9 GHz on the secondary CCD2, so there's a 4% frequency discrepancy between the two core complex dies.
These results were achieved using the be quiet! Pure Loop 2 FX 360mm liquid cooler installed inside the be quiet! Silent Base 802.
For single core workloads the 7950X appeared to maintain a clock frequency of 5.725 GHz, or 25 MHz above the advertised clock frequency. The maximum reported frequency was 5.85 GHz, though this didn't appear to be achieved under load.
Starting with Cinebench R23 multi-core results, we… probably should have included some Threadripper CPUs, damn this thing is fast.
For reference the 24-core/48-thread Threadripper 3960X scores about 34,000 points in this test, making the 7950X faster. Just as impressive is the fact that despite packing half the cores of the 3970X, it's just shy of 20% slower, so 50% fewer cores, less than 20% slower, amazing.
Single core is also strong, meaning the Ryzen 9 7950X should have no glaring weakness. Here we're looking at a 4% uplift from the 12900K and a 30% improvement from the 5950X.
Moving on to 7-Zip compression results we see that the 7950X is also very strong in this test, beating the 12900K armed with DDR5 by a 22% margin and the 5950X by a 27% margin. For those interested, it was 73% faster than the 7600X and scaling is not great as compression doesn't utilize AMD's SMT very well.
Where SMT scales well is when looking at decompression performance, where the 7950X is 150% faster than the 7600X -- super impressive scaling given there's 166% more cores. When compared to the 5950X we're looking at a 28% uplift and a massive 90% boost from Intel's Core i9-12900K.
The 7950X is also a beast in Blender, delivering 30% greater performance than that of the 5950X and 60% more than the 12900K, so AMD looks to be back to dominating Intel in productivity benchmarks, though Raptor Lake could change that very soon.
The Corona benchmark results are also mighty impressive, a 40% boost from the 5950X is seen along with a 51% improvement over the 12900K.
The Premiere Pro 2022 Puget Systems benchmark surprised us, though it probably shouldn't have given how much faster the Zen 3-based 5950X is when compared to the 5600X.
We're looking at a similar performance trend here making the 7950X 40% faster than the 7600X, which also means it's 25% faster than the 5950X and 42% faster than the 12900K.
Moving on to Adobe Photoshop, never before have we seen a Ryzen CPU dominate both Premiere and Photoshop. This is because Photoshop relies heavily on single core performance and typically Intel has had the upper hand. Zen 4 changes that as the 7950X narrowly edges out the 12900K by a mere 3% margin, making it 24% faster than the 5950X.
It's the same story when we look at After Effects. Previously even the 12600K was able to match the 5950X, but with the 7950X offering 33% more performance it again nudged ahead of the 12900K.
The last productivity benchmark we have today looks at code compilation performance, and here the 7950X delivered 25% greater performance than Intel's current flagship, the 12900K. The margin from the 5950X was also similar, for another easy win from AMD's new 16-core processor.
On the gaming front we'll start with Factorio which is very cache sensitive and only looks at single core performance. Despite packing twice the L3 cache of the 7600X, this is split into two CCD's and it would seem capacity isn't the key to best performance seeing the 5800X3D's result.
The R9 7950X was a mere 2% faster than the 7600X, making it 4% slower than the 12900K when paired with DDR5 memory.
The extra cores of the 7950X helped it jump ahead of the 7600X in Watch Dogs Legion, pushing the average frame rate 6% higher, to get within 3% of the 12900K armed with DDR5-6400 memory.
That's a 22% increase from the 5950X, so an impressive generational leap when it comes to gaming performance.
Rainbow Six Extraction performance was average, though perfectly fine given the 7950X was just 4% slower than the fastest CPU configuration tested, the 12900K running DDR5-6400 memory.
Moving on to Hitman 3, we find that the 7950X is no faster than the 7600X, so both were roughly on par with the 12900K using DDR5 memory. A good overall result even though Hitman doesn't require more than six Zen 4 cores.
Tiny Tina's Wonderlands is a good representation of how CPU demanding many modern games are. This game can be useful for testing low-end CPUs, but for the 7950X it just shows us that we're able to max out the RTX 3090 Ti, matching other high-end CPUs.
The 7950X was comparable to the 7600X in F1 22 where it was just 2% faster hitting 340 fps on average. That's a 6% uplift over the 5800X3D and an 11% improvement from the 12900K.
Spider-Man Remastered doesn't take any advantage of all those extra cores which results in the 7950X delivering 7600X-like performance, meaning it's significantly slower than the 12900K running DDR5 memory, trailing by a 13% margin.
Shadow of the Tomb Raider saw a reasonable 6% improvement to 1% lows for the 7950X over the 7600 and this placed it alongside the 12900K as the fastest CPU tested in this title.
Like the 7600X, the R9 7950X is impressive in Horizon Zero Dawn, hitting 212 fps which saw it lead the 5800X3D by a 7% margin and the 12900K running DDR5 memory by a 13% margin.
In Cyberpunk 2077 the 16-core flagship only beats the 7600X by a 1% margin in average frame rates, but the 7950X was able to boost 1% lows by a more impressive 9% margin. That said, performance overall was still underwhelming as the 12900K was 17% faster.
ACC also saw little improvement for the 7950X over the 7600X, meaning it was a good bit slower than the 5800X3D, but 5% faster than the 12900K.
The Riftbreaker data is interesting as the 7950X was actually 6% slower than the 7600X here. This appears to be an issue with 2 CCD Zen 4 processors as we've also seen this frame rate decline with the 7900X. AMD is aware of this and are in contact with the developer to see if there's a solution, so this small performance issue might be rectified shortly.
In Counter-Strike: Global Offensive the 7950X improves upon the 7600X's already impressive result by 5%, hitting an average frame rate of 511 fps. That's 14% faster than the 12900K running DDR5 memory and a massive 37% faster than the 5950X.
12 Game Average Performance
As we saw in the 7600X review, Zen 4 CPUs appear to handle DDR5-6400 well and enjoy a small performance boost when using the higher clocked memory, so there's a good chance both would nudge just ahead of the 12900K using the same spec memory, but either way performance overall is going to be much the same.
As impressive as the Ryzen 9 7950X's performance is, it comes at the cost of power. The previous-gen 5950X consumed a similar level of power to the 5800X, which was an incredible feat, the same can't be said for the 7950X which pushed total system power consumption up to 355 watts.
That's a massive 60% increase in total system draw when compared to the 5950X, so it's actually much worse than the 5950X when it comes to performance per watt, which isn't surprising given the 7600X only matched the 5800X.
When compared to Intel it's not a bad result, using slightly less power than the 12900K for ~60% more performance in this productivity test, so in terms of power efficiency Zen 4 wipes the floor with Alder Lake.
The Zen 4 CPUs intentionally give the impression that they're difficult to cool by delivering as much performance as possible by taking full advantage of the thermal and power headroom. AMD says with the new AM5 socket and higher TDP, Zen 4 processors will run into a thermal wall before they hit a power wall. This means under heavy load they'll sit at TJMax which is about 95 degrees Celsius for the Ryzen 7000 series, and this will be particularly true for the 12 and 16-core models.
AMD has stressed that this behavior is by design and add that it's important to note TJMax is the maximum safe operating temperature -- not the absolute maximum temperature.
In the case of Zen 4, the processors are designed to run at TJMax 24/7 without risk of damage or deterioration. AMD went on to say that 95C is not running hot, rather Zen 4 will intentionally go to this temperature under load because the power management system knows that this is the ideal way to squeeze the most performance out of the chip without damaging it.
Most AM4 coolers should be compatible with AM5, the exception being those that are secured from the rear side of the motherboard as the backplates aren't removable from AM5 boards. This means that all coolers can now be installed without requiring rear access to the motherboard, as is the case with HEDT processors, which is good news.
For all our testing we used the be quiet! Pure Loop 2 FX 360mm liquid cooler which is 100% compatible with AM5. After an hour of looping the Cinebench R23 multi-core with the Pure Loop installed inside the be quiet! Silent Base 802 we recorded a peak CPU temperature of 99C for the primary CCD and 96C for the secondary CCD, or just above the 95c TJMax.
It will be interesting to revisit these Zen 4 processors using a range of coolers to see how cooling performance affects CPU performance. But that's an investigation for another time.
Cost vs. Performance
For our value analysis we'll start with the gaming numbers, looking at cost per frame while only taking into account the price of each CPU. Now, it's to be expected because the 7950X only matched the gaming performance of the 7600X which costs half as much money, the cost per frame is terrible and much worse than even the 12900K, the 7950X isn't a gaming CPU.
When factoring in the entire platform cost things don't improve for the 7950X. It's much worse value than Intel's flagship for gaming and you'd just go for the 12700K as a gamer, which is similar to the 7600X in terms of value. Again, this isn't a gaming CPU and the same is true of the 7900X which we'll fully review soon. These are primarily productivity CPUs that also happen to game really well, but also they are no better than the 6 and 8-core versions, so if you're just gaming get one of the lower core count models and save yourself a ton of money.
Where the Ryzen 9 7950X excels is at productivity. For professionals, price to performance isn't that important within reason: a product that saves time, saves money. As far as desktop CPUs on mainstream platforms go, the 7950X is currently unbeatable costing $33.33 per 1000 pts in the Cinebench R23 test when factoring in not just the CPU cost, but also the memory and motherboard.
This makes it 16% better value than the 5950X, though we'd say for most the almost 60% boost in performance is a far bigger draw card. The 12900K can't live with it either, not only is the 7950X 40% faster, but it's also 7% better in terms of value.
For video professionals it's a similar story, the 7950X will save you a good chunk of time compared to other desktop CPUs, and it's also great from a value perspective, roughly matching the cheaper 5950X while offering around 15% better value than the 12900K.
What We Learned
The Ryzen 9 7950X is the new performance king and the jack of all trades, apart from maybe power consumption and pricing, of course. The Zen 4 flagship can do everything exceptionally well with no real weaknesses. Even the price, while certainly expensive, for what you're getting it's actually reasonable and as we noted at the start of the review, it's a $100 drop from the previous generation at launch, so it's hard to complain about that.
In most productivity workloads it stomps the Core i9-12900K and it just happens to be better value, too. As a professional-grade product, the 7950X is sensational, but as a gaming CPU it's overkill and a bit pointless.
This is essentially the same situation we observed with Zen 3, so if you read our reviews and bought a CPU for pure gaming, we hope we saved you some money. For example, if you bought the 5900X or 5950X in an effort to future-proof your gaming experience, unfortunately you've just thrown money away. Outside of streaming, there is zero benefit to the 12 and 16-core models over the 8-core and for the most part even the 6-core models, and the same will be true of Zen 4, if not more so.
Of course, if you plan to work and play using the same PC, then no CPU has ever done both so well as the 7950X. We can't anticipate how competent Intel's upcoming Core i9-13900K will be against the 7950X, but we suspect we'll know exactly by this time next month, so it's probably best to wait before making your purchase.
For those of you interested in the Ryzen 9 7950X, we feel the premium you have to pay for X670E motherboards won't be a deterrent, nor will the price of DDR5 memory. You will also want to ensure proper cooling, and above all else, bring a big cooler to the table for maximum performance.
You may have noticed some of the additional testing we did for the 7600X review was not included here. The reason is that there's nothing new to show: the 7950X scales the same with memory, it can handle DDR5-6000 with four DIMMs populated, and Resizable BAR scales as expected. Overclocking is something we'll probably look at soon in a separate feature.
The Ryzen 9 7950X is an impressive kit of hardware, it's the ultimate no compromise solution and it's hands down the best at what it does, at least for now. For a few generations now AMD has dominated productivity performance with their 16-core desktop CPU. The 3950X made mockery of the Core i9-9900K back in the day, though it did lag behind for gaming. The 5950X retired Intel's entire HEDT lineup and demolished the 10900K in the process, even managing to give it a hard time in the gaming benchmarks.
The Ryzen 9 7950X has faced the stiffest competition from Intel in the 12900K, while still managing to come out on top by winning the productivity benchmarks comfortably and offering comparable gaming performance. We can highly recommend the Ryzen 9 7950X, with the only caveat that Intel's 13th-gen will be launched in a matter of weeks, so before pulling the trigger we suggest you assess your options when both chip giants will have their latest and greatest competing head to head.