Monday, August 29th 2022
AMD Announces Ryzen 7000 Series "Zen 4" Desktop Processors
AMD today announced the Ryzen 7000 series "Zen 4" desktop processors. These debut the company's new "Zen 4" architecture to the market, increasing IPC, performance, with new-generation I/O such as DDR5 and PCI-Express Gen 5. AMD hasn't increased core-counts over the previous-generation, the Ryzen 5 series is still 6-core/12-thread, the Ryzen 7 8-core/16-thread, and Ryzen 9 either 12-core/24-thread, or 16-core/32-thread; but these are all P-cores. AMD is claiming a 13% IPC uplift generation over generation, which coupled with faster DDR5 memory, and CPU clock speeds of up to 5.70 GHz, give the Ryzen 7000-series processor an up to 29% single-core performance gain over the Ryzen 5000 "Zen 3."
At their press event, AMD showed us an up to 35% increase in gaming performance over the previous-generation, and an up to 45% increase in creator performance (which is where it gets the confidence to stick to its core-counts from). The "Zen 4" CPU core dies (CCDs) are built on the TSMC 5 nm EUV (N5) node. Even the I/O die sees a transition to 6 nm (N6), from 12 nm. The switch to 5 nm gives "Zen 4" 62 percent lower power for the same performance, or 49% more performance for the same power. versus the Ryzen 5000 series on 7 nm. The "Zen 4" core along with its dedicated L2 cache is 50% smaller, and 47% more energy efficient than the "Golden Cove" P-core of "Alder Lake."
The "Zen 4" CPU core gets a bulk of its 13% IPC gain from the core's front-end, followed by load-store, branch-prediction, and execution engine. The company also doubled the size of the per-core L2 cache to 1 MB. The core introduces support for AVX-512 instruction set. Eight cores share a 32 MB L3 cache on a CCD. The 6-core and 8-core SKUs in the Ryzen 5 and Ryzen 7 series, come with a single CCD, whereas the 12-core and 16-core Ryzen 9 parts come with two.
AMD introduces a brand new socket with Ryzen 7000, Socket AM5. This is a resilient 1718-pin LGA, with the ability to delivery up to 230 W of power, and comes with next-generation I/O that includes DDR5 and PCIe Gen 5. Physically, the coolers are compatible with Socket AM4 thermal solutions, so you can carry over your old coolers. AMD is promising to launch future generations of Ryzen processors that are AM5-compatible going up to 2025 at least.
There will be four chipset choices with Ryzen 7000, these include the X670E and X670 in the high-end; and the B650 and B650E in the mid-range. Motherboards with X670/E debut in September, and the B650/E in October. AM5 is the first platform with CPU-attached NVMe Gen 5, and the company predicts the first Gen 5 SSDs should arrive by November. We confirmed with AMD that they are not artificially limiting the performance of processors running on the B-Series chipsets vs the X-Series chipsets. The difference between B650 and B650E is that B650E offers support for PCIe Gen 5 for graphics cards and SSDs, while B650 non-E supports PCIe 5.0 SSDs, and PCIe 4 GPUs. AMD is introducing a new memory profile technology called EXPO that eases memory overclocking. It is a royalty-free technology, and includes memory settings specific to the AMD architecture. You are of course able to use Intel XMP-compatible DDR5 memory modules, these might just not have the most perfect settings out of the box. As many as 15 memory kits are being launched at speeds of up to DDR5-6400, from various manufacturers.
The AMD Ryzen 5 7600X is a 6-core/12-thread processor with 4.70 GHz nominal clocks. up to 5.30 GHz boost, 105 W TDP, and is being launched at $299. The Ryzen 7 7700X is 8-core/16-thread, clocked at 4.50 GHz, with up to 5.40 GHz boost, 105 W TDP, and is being launched at $399. The Ryzen 9 7900X is 12-core/24-thread, clocked at 4.70 GHz, with up to 5.60 GHz, 170 W TDP, and is being launched at $549. The top 7950X is 16-core/32-thread, clocked at 4.50 GHz, with up to 5.70 GHz boost, 170 W TDP, launching at $699. All SKUs available to purchase on September 27, 2022. This is an on-shelf date, not a preorder date (we have that confirmed personally).
The complete slide-deck follows.
At their press event, AMD showed us an up to 35% increase in gaming performance over the previous-generation, and an up to 45% increase in creator performance (which is where it gets the confidence to stick to its core-counts from). The "Zen 4" CPU core dies (CCDs) are built on the TSMC 5 nm EUV (N5) node. Even the I/O die sees a transition to 6 nm (N6), from 12 nm. The switch to 5 nm gives "Zen 4" 62 percent lower power for the same performance, or 49% more performance for the same power. versus the Ryzen 5000 series on 7 nm. The "Zen 4" core along with its dedicated L2 cache is 50% smaller, and 47% more energy efficient than the "Golden Cove" P-core of "Alder Lake."
The complete slide-deck follows.
195 Comments on AMD Announces Ryzen 7000 Series "Zen 4" Desktop Processors
@Valantar See it is posters like the above that make me sound like a broken record. Yes, usually it's Intel that gets flamed for their cpus being "inefficient" cause of the huge power limits, but now it is amds turn and here I am defending amd. Even they themselves tested efficiency at same wattage. It seems very likely all of zen 4 will be less efficient than zen 3 out of the box, thanks to the increase in TDP, and so posters like the above won't understand that actually the CPUs are more efficient - but they are just pushed higher.They are very relevant for people that plan on keeping their CPU? If you don't test at 720p, how will you know which CPU is faster and therefore will last you longer?
If you need the high cores for whichever type of work you will be doing, then go for that.
If you need the low cores for cheap price but high(ish) fps, then go for that.
That's the whole point, there's a whole lineup of new CPUs being announced from AMD today (and coming from Intel very soon) that you go and pick whichever suits you the best.
If for some reason your circumstances change in the future, you can sell your existing hardware and upgrade/downgrade depending on where you start from and go to.
That caught up with the CPU side of the equation when Turing showed up, where higher end Turing models showed some fairly big gaps with even current model CPUs.
With Ampere and AMDs RX 6000 line - well, a lot of perfectly good CPUs from 3 years ago will now go CPU limited with anything over a 3060 Ti or 6700 XT. A lot of people didn't seem to notice this change, probably because nobody could afford $700 3060 Ti's in 2020 and 2021.
That is going to get a whole lot more pronounced with this next generation of GPUs. That IMO is going to drive enthusiast PC sales over the next couple of years in the CPU space, and I suspect this next gen of GPUs will be the one that shows up how GPUs have caught and passed the ability of many CPUs to keep up.
The days of seeing something like this, where a top SKU i7 and a bottom line i3 effectively perform the same, are pretty much over.
CPU scaling 11 years ago:
CPU scaling 2 years ago (RTX 3080):
It took about 3 months for Zen 2 and ~5 months for Zen 3, so let's not hope they continue that trend.
I hope reviews will uncover whether these boost clocks are obtainable at stock in most workloads, or if the typical boost is much lower like with Intel since Coffee Lake.3D V-Cache didn't cause a >30% performance uplift for Zen 3 in gaming (overall), and it's not likely to do for Zen 4 either, if anything the relative gain will probably be less.I've seen the slide saying 8%, but I haven't confirmed whether it's official or a fan-made fake. (if it originates from that YouTube guy then we know the answer)
But it's worth mentioning that IPC doesn't change though, not unless they disable features etc., and it certainly doesn't improve. AMD have known the IPC since before tapeout.
But if their "IPC" estimate is based on those benchmarks in there, then that's not IPC at all. A bunch of those are games. You can never approximate IPC through a non-CPU workload. :facepalm:
FWIW I think a good rule of thumb on Zen 3 / Alder Lake / Gen 10 is that unless you have more than a 3070, it doesn't matter as long as you have a 10600K / 5600X or better.
However, the 3060 Ti was roughly equivalent to a 2080 Super.
If we get that again, the 4060 Ti will be roughly equivalent to a 3080 Ti.
That means that for upper middle tier gamers for example, the 4000 Nvidia cards will make this chart relevant.
I would bet that when we see this type of test on a 4080, the delta will become dramatic. Think about the fact that there's not even a Zen 2 on this chart - probably a reason for that no? Zen 2 generally performed about like a 10400, which would mean there is a 20% delta in FPS at 1440P with a 3080 vs a top line Alder Lake (and probably more vs Zen 4).
Napkin math - with 4000 series GPUs, there will be a 20%+ delta between 3 year old CPUs using a 4060 Ti.
Take note, this is also at 1440p, it's worse at 1080p.
Now can't wait to see independent benchmarks (TPU YES PLS)!! Please add real-time gaming power consumption+temp findings to the review too. Synthetic or non-gaming workload specific benchmarks are boring for us gamers.
Looks like a way better cpu than Zen 3 and v-cache models are coming earlier than expected in Q1 2023 and will boost perforamnce even more than it did with 5800X3D and there will be no (or only a small) clock penalty (100MHz). The v-cache in Zen 4 is much more advanced.
Personally can't wait for 7900X3D.
Doesn't matter, if there is a 20% relative difference, as long as it can keep your target. For "future-proofing", sure it matters.
If your target is 1440p@60
Borderlands 3, being the lowest performing game in the 5700X review, still have over 60fps (1440p).
If your target is higher than 60, then fewer and fewer processors are "good enough"
from 5700X review. www.techpowerup.com/review/amd-ryzen-7-5700x/16.html
It is different when you have a 170w CPU you under-volt it (I did that with my 6900xt for instance) you don't lose performance but due to your good chip bin, you get same performance or 1% lower for a decent decrease in power consumption lets say 150w from 170w, lowering heat output at the same time. That is a different case than getting a 105w CPU and limit it to 65w. Power consumption will go down noticeably and performance as well.
What you are saying really really makes no sense to me. Cause the same way, limiting the 7600x to 105w also means you lose performance compared to someone limiting it at 200w. So what?
Power limiting high-end, high-TDP CPUs makes sense if for example you want a silent system without using a water cooler, or you care about processor efficiency.
I love that AMD has built-in Eco Mode settings for all their CPUs - 65W for anything 105W TDP; 45W for anything 65W TDP, etc. But these are very rarely enabled - again, enthusiasts would most likely prefer hand tuning (if interested in reducing power at all), and other users don't know about them. But at least it saves them from Intel's massive SKU overload. It'll be interesting to see what Eco Mode settings these 170W chips will have though.
AMD has presented a chart, where 7950x is 67% more efficient than a 5950x when both capped at 65w. That is great and in a laptop environment it will do wonders though but it will still be slower than a desktop 7950x with advertised 170w. Like I said, any CPU limited or capped with power will be more efficient but that is not what you paid for. Your limiting factor is huge and you will lose a lot of performance to achieve your power goal. It is not worth it and simply silly approach to prove some a point or something.
A 7950x at 105w is way more efficient than a 7950x at 170w,so by limiting it i lose performance but get efficiency. So if i care about efficiency, i get more for what i paid for, since it's more efficient at 105w!!
Dude, you can't focus on efficiency only or on power consumption only. It is like comparing CPUs' performance by frequency only.
What do you mean you get more? You get less performance for less power usage. How is that more? Your problem is you focus on TDP or power consumption not performance. You pay for a 7950x for instance which will crunch data (blender) that fast. You limit it and you didnt achieve more performance but less but obviously it will use less power.
For me this is simple but you just don't understand, that CPUs are not for efficiency but performance and efficiency is just accompanying the performance it delivers.
To help understanding the idea, here is a graph I made with my i7-12700K at default frequencies, but optimized voltages (my motherboard gives plenty of voltage with default settings). With the settings it has at the moment, the CPU won't require more than 150W, but under overclocked conditions it could take up to 250W or more if cooling allows. Two things are clear from the graph: increasing power yields progressively diminishing performance returns and increases the energy (Joules) required to perform the same work (here, a render test similar to that used in TPU reviews).
All CPUs both from Intel and AMD will exhibit a behavior along these lines.
Get a 7950X, shut down 4 cores, limit to 105-95w and enjoy most of the pref at half the watteg and $ cost.
:ohwell: