The new Ryzen lineup initially consists of four processors set to launch on September 27. Starting at the high-end, we have the Ryzen 9 7950X selling for $700, offering AMD’s now standard 16-core / 32-thread configuration at a new 170W TDP. It’s clocked up to a whopping 5.7 GHz with a 4.5 GHz base and 80MB of L3 cache, split into 64MB L3 and 16MB L2.

Next is the Ryzen 9 7900X bringing 12 cores and 24 threads at up to 5.6 GHz for $550. Then we have the Ryzen 7 7700X for $400, as the 8-core option with clock speeds up to 5.4 GHz, followed by the most affordable model for now, the Ryzen 7 7600X at $300 featuring 6 cores, 12 threads and a 5.3 GHz boost frequency.

This pricing structure is a little different, especially for the Ryzen 9 7950X which now comes in $100 cheaper than the 5950X at launch, so that’s good to see. However, the 5950X has fallen to just $550, so compared to current pricing the 7950X presents a $150 price premium. The launch MSRP for the 12-core model stays the same, though with current pricing we’re looking at ~$150 premium for the 7900X over 5900X as well. Meanwhile, the 8-core model is a bit cheaper than what the 5800X launched at, $400 vs $450, though the model we’re getting is the 7700X so we guess this could be compared to the Ryzen 7 5700X at $300 or as low as $250 at the moment, another $150 price gap for the new generation. It is however $50 cheaper than the 5800X3D which is an interesting placement.

Less exciting news is the 6-core model coming in at $300, which is expensive in the current market. While this is the same launch price as the Ryzen 5 5600X back in 2020, that part has fallen to $200 in the face of strong competition from Intel, plus we now have the Ryzen 5 5600 occupying the $170 price point. AMD is leaving the door open for a cheaper Zen 4 offering with something like a R5 7600 non-X in the future, but for now, as we saw with Zen 3, the new Ryzen 7000 line is mostly premium priced.

IPC and Performance Claims

On a more positive note, AMD revealed the final IPC claims for Zen 4 and the total expected single-core performance gain. For IPC we’re looking at a 13 percent increase on average this generation compared to Zen 3, with both CPUs configured as an 8-core, 16-thread model at 4.0 GHz.

AMD provided this chart which shows where this number comes from, ranging anywhere from slim gains around 5% in workloads like Adobe Lightroom, to 9% in Cinebench R23, through to around that 13% mark in some productivity apps and gaming, and up to gains above 20% in specific tasks. The IPC claim is a mix of productivity and gaming and comes from various architectural improvements as you’d expect.

The total 29% performance boost claim is achieved through a significant frequency bump this generation, thanks to the shift to TSMC’s 5nm technology and a streamlined core design. A 5.7 GHz maximum rated clock is 16 percent higher than AMD’s best Zen 3 parts. This puts AMD ahead of Intel’s best Alder Lake frequency of 5.5 GHz with the Core i9-12900KS – although we expect Raptor Lake to increase frequencies to some degree in just a few months. This figure is much higher than what AMD was quoting back at Computex earlier this year, which is good news for AMD’s Zen 4 processor, but does make AMD’s strategy back then look a bit odd. By announcing a mere “greater than 15%” single thread performance uplift, the hype bubble surrounding Zen 4 was deflated somewhat.

Regardless of AMD’s move there, this sort of single-thread uplift is the highest seen in the Zen series so far. It’s the biggest frequency jump, beating the 9 percent gain we saw with Zen 2 over Zen+. The IPC increase isn’t as large as it was with Zen 2 or Zen 3, but in combination with such a high frequency we end up with a very significant performance increase. It’s interesting to see some of the performance claims AMD is making this generation. It’s interesting to see some of the performance claims AMD is making this generation. For example, AMD is claiming they have the fastest core in gaming, using Geekbench single-thread performance showing the Ryzen 9 7950X outperforming the Intel Core i9-12900K by 12 percent. That’s a good number, but I do find it curious AMD have switched from using their usual benchmark of Cinebench R23 single-thread to Geekbench, which does suggest that Cinebench numbers are less impressive – not that it matters too much as it’s the overall performance picture that matters most.

Generally speaking, what we’ve found is that the Core i9-12900K and 12900KS are between 20 and 30 percent faster than the Ryzen 9 5950X for single-thread workloads. A 29% percent performance uplift for AMD would allow the 7950X to match and slightly exceed Alder Lake in many instances, at the very least they should be trading blows with this sort of performance.

However, the question remains as to how well Zen 4 will compete against Raptor Lake, given it’s the 13th-gen parts that will become Ryzen 7000’s main competitor. If Intel is able to eke out a 5 to 10 percent single-thread performance uplift, they’ll likely be able to reclaim that single-thread crown. This is setting up a tasty battle between AMD and Intel for these sorts of workloads and I’m quite interested to see exactly where everything falls. It could be a very competitive generation.

Gaming Performance

We got a few benchmarks from AMD, one showing gains of up to 35%, although in a limited selection of titles. Unfortunately, AMD only is comparing their new Zen 4 parts to either the 5950X or 12900K, rather than their own gaming champion in the 5800X3D. As we know, the 5800X3D with its massive V-Cache is able to provide a ~15 percent performance uplift over other Zen 3 processors in CPU-limited gaming. These claims from AMD, which should always be taken with a grain of salt, do look decent but it’s hard to say whether it will end up outperforming the 5800X3D by a significant amount.

With that said, AMD is preparing a Zen 4 V-Cache release for later down the track, so if Zen 4 is able to roughly match the level of the 5800X3D without V-Cache, then adding that technology to the picture should give AMD a mighty processor for gaming. Also of interest are AMD’s claims for the Ryzen 5 7600X in gaming. They say it should be 5% faster than the Core i9-12900K on average when both use DDR5-6000 CL30 memory. We’ll have to see how that plays out across more titles, but that’s a very good position for the 7600X given the 12900K is nearly $600, compared to the 7600X at $300.

The bigger issue for AMD isn’t so much the 7600X vs 12900K though, but how this sort of six core processor will compare to Intel’s more wallet-friendly offerings from not just the 12th-gen, but 13th-gen as well. The 12900K is only about 5 to 10 percent faster in games than a CPU like the Core i5-12600K, which currently retails for $240. For AMD’s cheapest Zen 4 processor to offer competitive cost per frame compared to the 12600K, it would have to be over 20 percent faster which could be tough to achieve going on what these performance numbers suggest.

And we guess this is where the disappointment at the $300 price point lies. If it was sitting closer to $250 it would sound like an extremely impressive deal, but at the current MSRP it will require a full benchmark analysis.

Productivity Performance

The really beastly aspect to Zen 4 appears to be productivity performance. AMD is showing gains between 32 and 48 percent across a range of rendering workloads like V-Ray and Corona comparing the 16-core 7950X to the 16-core 5950X, which is a massive performance uplift in a multi-thread benchmark for two CPUs with the same core count. This is achieved through an enormous all-core frequency increase, with the 7950X able to hit at least 5 GHz all core, giving a 25% jump there alone, then we see an additional 13 percent average IPC increase, and yeah, that’s a BIG difference in performance.

The really beastly aspect to Zen 4 appears to be productivity performance. In V-Ray, AMD showed the 7950X demolishing the Core i9-12900K by up to 57%, which will give AMD heaps of performance buffer up against 13th-gen parts coming soon if these performance claims are accurate across more workloads. Raptor Lake is expected to increase the E-core count from 8 to 16 cores in the flagship model, but that’s unlikely to be sufficient to close a gap that large.

AMD is also claiming a 47% performance per watt advantage over the 12900K, suggesting that the 7950X will consume far less power than the 12900K. This is no surprise considering AMD’s big lead in efficiency with the previous generation, and we expect that to only improve with a new process node. So while the 170W TDP is higher than we’ve seen from current models like the 5950X, we’re getting a much larger rise in performance and power consumption still below that of Intel’s top end parts. At the same TDP, AMD is expecting 35% performance gains at 170W and a huge 74% at 65W, that itself bodes well for Zen 4-based mobile CPUs to come in the future.

A New Long-Term Platform

AMD answered the key question today with news that AM5 will be supported through 2025+, which should mean at least four years of support, matching AMD’s launch claims for AM4.

At a minimum this would support Zen 4 and the Zen 5 generation coming in 2024, with probable support for a third generation as well. This continues to blow Intel’s support out of the water, as Intel continues to only offer two years of support with each motherboard socket and series at best.

The motherboards will be split into four series: X670 and X670 Extreme coming in September, plus B650 and B650 Extreme in October. The Extreme appears to mean additional PCIe 5.0 lane support for graphics and storage versus the non-Extreme models, as well as just being the best boards with the most features. Entry-level motherboards are expected to start at $125, which is a bit more expensive than AM4.

Adding to the platform cost is support for DDR5 memory only. While AMD says they are easing the transition to DDR5 and talk about technology like EXPO for DDR5 overclocking, DDR5 is still more expensive than DDR4 at the moment. A 32GB DDR5-5600 kit from Corsair will set you back about $170 these days, compared to $100 for DDR4-3600. This is a better price difference than at the launch of Alder Lake for sure, but still a cost to factor in when moving to the AM5 platform. Meanwhile, PCIe 5.0 SSDs are expected to be available in November.

Next-Gen Radeon Teaser?

AMD briefly mentioned RDNA3, with a graphics card that looks like this…

That, and a gameplay demo with absolutely no extra information. That pretty much covers it for AMD’s announcements today, with plenty more to come closer to launch day in terms of additional architecture details, and of course, performance benchmarks, which we’ll bring on with timely reviews.

Takeaways

The most important thing to discover today was that AM5 will have long platform support like AM4 had, which makes it much easier to justify jumping into a new platform with DDR5 technology and gives AMD a serious competitive advantage over Intel. The expectation on Intel’s side is that the LGA 1700 socket will be replaced after 13th-gen, meaning people buying into Intel’s new CPU family this year will have effectively no upgrade path later. It’s absolutely essential for Intel’s next socket and motherboard series to support more than two generations, at this point it’s almost a non-negotiable feature.

The performance side of the equation is looking solid. AMD reversed their hype-deflating announcement from earlier this year with a big performance jump in line with expectations and enough to make the battle between Zen 4 and Intel’s CPUs quite interesting. Multi-thread performance looks mighty good thanks to big frequency increases, while gaming is still a bit of a mystery as to how it stacks up compared to the 5800X3D. In terms of value, a part like the Ryzen 5 7600X feels expensive at $300 despite AMD claiming it’s better at gaming than a Core i9-12900K. We just don’t think that price will be compelling against Intel, which in the mid range is typically not that much slower than Core i9 models yet are priced closer to $200 or below. There’s really no counter for something like the Core i5-12400 at $190 in the Zen 4 lineup yet, which will limit upgrades, and the value proposition over something like a $170 Ryzen 5 5600 is also questionable.

This should be tempered by the fact that the AM5 platform will be expensive to upgrade to, so perhaps opting for just higher-end models to begin with makes sense. New AMD AM5 boards will start around $125, compared to ~$100 for budget B550 boards, while DDR5 memory is at least 50% more expensive than DDR4 per GB. This will add $50 to $100 in platform costs over a new AM4 build or Intel DDR4 12th-gen build. Plus, each CPU is $100 to $150 more expensive right now than their Zen 3 predecessors, despite matching or coming in cheaper than launch pricing.