Thursday, August 25th 2022
![AMD](https://tpucdn.com/images/news/amd-v1723816949774.png)
AMD Ryzen 7 7700X and Ryzen 5 7600X Cinebench R23 Single-Thread Numbers See it Trade Blows with the Competition
Alleged Cinebench R23 single-threaded benchmark numbers of the upcoming Ryzen 7 7700X and Ryzen 5 7600X "Zen 4" processors, leaked to the web by Greymon55, and tabulated by VideoCardz, show the two chips to be matching Intel's 12th and 13th Gen Core processors. The 7700X 8-core/16-thread processor is shown scoring anywhere between 2000 to 2099 points (denoted as 20xx), while the 7600X does anywhere between 1900 to 1999 points (19xx). This would see the two easily match/beat the 12th Gen Core "Alder Lake" P-cores, with the i9-12900K scoring 2000 points, and the i5-12600K getting 1920 points.
Numbers for the unreleased 13th Gen Core "Raptor Lake" put Intel at an advantage, with the i9-13900K allegedly scoring 2290 points, and the i5-13600K allegedly 1967 points, but what's important is that the single-thread performance, and application performance of less-parallelized workloads, such as games, could be highly competitive for "Zen 4" against Intel.
Sources:
Greymon55 (Twitter), VideoCardz, harukaze5719 (Twitter)
Numbers for the unreleased 13th Gen Core "Raptor Lake" put Intel at an advantage, with the i9-13900K allegedly scoring 2290 points, and the i5-13600K allegedly 1967 points, but what's important is that the single-thread performance, and application performance of less-parallelized workloads, such as games, could be highly competitive for "Zen 4" against Intel.
48 Comments on AMD Ryzen 7 7700X and Ryzen 5 7600X Cinebench R23 Single-Thread Numbers See it Trade Blows with the Competition
Yeah, I can see that :rolleyes:
AMD can/might take the lead in multi-threaded loads but with 8+16 combo it will be tough competition.
Ah!!!! this month of September is getting more and more interesting with leaks and number :D
And there will be no B650 boards for several months, meaning you'll have to go for X670 ... or Intel, hehe For higher refresh rate 4k it already matters, but a lot of so called 4k gamers are still on 60Hz screens and there it really doesn't all that much (at least if you have 8700k or better). Rampage! If this is anywhere near true, then zen4 won't have a single victory, be it single or multi thread! I seem to remember AVX512 being completely dismissed when AyyyMD didn't have it... :D Let's see how that goes when the price of the platform will be doubled! ;) :roll:
Let's wait to judge it until we see performance, availability, pricing, etc.
I'm more worried about whether they manage to get their firmware in order this time.
My prediction regarding 720p TPU scores is the following:
7950X 112.45%
7900X 111.9%
7700X 108.65%
7600X 106%
12900K 100%
But if AMD keeps the same Zen3 pricing level it won't be competitive, for example 13900K will at least match 7950X and 13700K will be faster than 7700X (still 13600K at 5.1GHz should have a little bit slower 720p performance vs 7600X) so taking account the multithreading performance and also the need for a good AM5 start regarding competitiveness/buzz/sales I think AMD must deliver also on the pricing level!
We don't have anything regarding gaming benchmark for Zen4.
The logic that I followed is the below:
I took 7600X as a starting point that has the biggest % PPT increase affecting Zen cores (if you deduct the IOD power consumption- along with 7900X).
One of the easily stand-out improvements of Zen4 architecture is that AMD doubled the L2 cache in relation with Zen 3.
As we have seen in the past cache related improvements depending the architecture have positive effect in gaming apps, and Ryzen core has proved again and again that it's such an architecture.
AMD claims around 8% average IPC increase, so taking account the successful gaming focus strategy of Ryzen that AMD followed the last 2 years and the L2 doubling, I can't imagine why in gaming the IPC increase won't be around the average advertised (8%).
So if you take account the frequency difference and the IPC, I came with around +24% in relation with 5600X.
(for the rest lineup I took as a base 7600X)
If the leaked numbers are correct in CB23 we have same or a little better increase, so the potential at least is there to consider such a scenario.
As you can understand this is a very positive scenario!
But if you read what I posted, you will understand also that what I claimed is that despite using a very positive scenario for AMD, this performance won't be enough to allow AMD to keep the same pricing strategy/level as Zen3 and still be considered competitive enough imo.
If the rumor regarding 7700X is correct ($299 according to 3 leaks) this means that at least in 2 SKUs (7600X also) AMD seems will be competitive enough with Intel.
But the odds based on last year's AMD's pricing strategy are not in favour of such pricing.Fingers crossed?
The IPC estimates is quite possibly fairly accurate, but translating this into gaming performance is not wise. Gaming performance is only indirectly affected by the CPU. Don't get me wrong, the CPU is important, but only up to a point. If you had a CPU with 10x the IPC, you wouldn't see a huge different in today's games, as most games are close to or not being bottlenecked by current fast CPUs. As CPUs get faster, you will first see the average framerate flatten out, then the minimum framerate, and beyond that faster CPUs will not benefit those particular games. This is the reason so many were disappointed by Alder Lake only showing marginal gaming gains (overall) despite a ~40% IPC change over the Skylake family (19% over Sunny Cove).
So I wouldn't expect a lot of gains in gaming, except for edge cases of course. But that doesn't mean the new CPUs will not be great for future games coming down the line, or practically anything non-gaming which benefits greatly from gains in single-treaded performance.
I just want to point out also the below in order to keep our minds open to possibilities:
5900X vs 3900X tested resulted in +19% claimed IPC increase from AMD.
5900X single core Turbo 4.8GHz
3900X single core Turbo 4.6GHz
This gives 1.19 X 4.8/4.6 = 1.2417X improvement possibility in ST.
According to TPU the 720p difference between 5900X/3900X is 103.9%/80.2%=1.2955X so higher than ipc x frequency difference! (1.2417 X)
At the same time ST CB23 improved only 1622/1357= 1.1952X way-way less than what was the gaming increase and the ipc x frequency increase.
Even if you replace frequencies with light threaded clocks instead of single core turbo, 5900X in 720p gaming still achieves at least IPC X light threaded clock difference result!
Now there are specific reasons for this result and doesn't mean that it will apply in Zen4 (it won't logically) but like i said i took a very positive scenario in order to make my case of what AMD pricing strategy must be even in this case.
As you see, while in 5900X case ST CB23 score difference vs 3900X was way less than the theoretical difference (claimed IPC increase X frequency difference) here we have exactly the theoretical difference (if the leaks are correct) that's a very good result and the reason i examined this scenario (my original 720p prediction for Zen4 was -1%, -1.5% slower than this positive scenario and still is my preferred scenario) (coincidentally this positive scenario would give around +15% 1080p 7700X/5800X gaming difference in current TPU testbed)
Also don't forget that there is the possibility regarding TPU testbed, that in Q4 maybe W1zzard will change RTX 3080 into something from Ada/RDNA3 series, so another reason to consider also (at launch of course the results will be based on RTX 3080, so slower vs what Ada/RDNA3 will give logically)
If you re-benchmark a new selection of games 2-3 years from now, you will probably see a slight shift and see larger benefits from Alder Lake and upcoming Zen 4.
Lastly, I hope everyone realize that 720p benchmarks are for technical curiosity, as no one should buy a high-end GPU to run 720p. The CPU bottlenecking on 1440p or 4K will be quite a different story. Apples and oranges.
A few things to clear up, firstly IPC is a measure of average instructions per clock, not performance. We use performance to estimate IPC, but that doesn't mean you can take clock speed and multiply by IPC gain to calculate gains in individual applications, as applications will stress different parts of a CPU. If one application relies heavily on SIMD and the new CPU mainly gets their IPC gain from frontend tweaks, caches and smaller improvements on the execution side, we can expect lower gains in that application. If another new CPU upgrade is a heavy SIMD upgrade, then we can suddenly see 30% gains in applications even if the CPU just offers 10% higher IPC.
Additionally, it's not unusual that some benchmarks will throttle the CPU too, especially on Intel, the actual clock speed is usually a lot lower than people think.
This is why I always say, performance matters, but IPC is what we use to explain some of it. :)
720p used mostly as an indication regarding how each CPU affecting gaming performance when we make it a limiting factor by lowering the rendered resolution. in general i agree.
I don't know why you're arguing since i already said that this a very positive/optimistic uptake regarding Zen4 possible gaming outcomes didn't I?
Is this positive scenario the most likely to happen, no, i already explained it, but it can happen anyway and i don't see in your arguments any specific Zen4 technical argument, only general well known (that i agree) facts regarding how each game/engine can potentially be affected or not by SIMD or IPC etc that can be applied to a vast number of CPUs anyway, do you have any argument based on specific Zen4 architecture that makes you exclude the possibility of such outcome?
Let's hope this results in price cuts for consumers, and not overpriced top end parts that use 200W more for 1% higher performance
But what we do know is which types of improvements which tends to help gaming performance. Let's look at Intel's two recent generational improvements; Sunny Cove (Rocket Lake) and Alder Lake, which offered +~18% and +~19% IPC gains respectively (and fairly similar clocks), or about ~40% in total over Skylake since Rocket Lake was only a few months on the market before Alder Lake, so it barely counts.
Both of these architectures offered marginal gains in gaming performance, so let's look at some of their improvements;
Sunny Cove:
- Frontend - larger instruction window, better branch pred. etc. (usually affects gaming greatly)
- More L2 cache (can affect games)
- More L1 size/bandwidth (can affect games)
- Larger uop cache and TLB (can affect games)
- Double load/store bandwidth (little effect on games)
- Significantly increased INT MUL/DIV performance (can have some effect on games, but not a lot)
Alder Lake:
- Massive frontend improvements, pretty much every aspect of it, especially decoding, instruction window (usually affects gaming greatly)
- More L2 cache (can affect games)
- Larger uop cache and TLB (can affect games)
- Added another ALU (can have some effect on games, but not a lot)
So both of these architectures should have the foundation to offer much greater gaming performance, but the reason they don't is that most current games are not demanding enough. But they can show greater advantage in the future if new games become more demanding.
Back to Zen 4, with Zen 4 offering ~8% IPC improvements, double L2 cache and higher clocks, we are probably not looking at something which will outperform Alder Lake overall. As for gaming specifically, we can't know for sure, but even if we assume it's a lot of frontend improvements or even if AMD have made a genius improvement, it will still run into the same issue as Alder Lake, as many games get to the point where they basically fully saturate the GPU, especially on 1080p and higher. You got me wrong, I'm not criticizing Wizzard's game selection. I'm claiming that future games will probably be gradually more demanding, therefore creating a new bottleneck to overcome. So 2-3 years from now, we should expect Alder Lake to look better vs. Comet Lake.
Looking back at games from the past 10+ years you'll see that games are slowly getting more demanding for CPUs, but that demands in GPU performance is growing much quicker. And considering the graphics APIs are focusing on batching operations and offloading this to the GPU, we can expect this trend to continue. This is why until recently Haswell and even Sandy Bridge CPUs have been "good enough" for gaming at realistic resolutions with a new mid-range GPU, some may argue they are still good enough.
Essentially you are taking historical data from other completely different architectures from other companies and apply the results here based on good assumptions, that i already told you that i agree, i wrote from the start that my hypothesis was a very optimistic Zen4 720p performance outcome, didn't i? So what is the reason for repeated arguements since as far as I'm aware we agree on this.
Unless you want to say that this optimistic scenario that i took isn't just not very likely to happen as as i wrote, but it has 0% chance happening.
in that case i think is unwise knowing so little for Zen4 to make so strong assumptions! I may be wrong, but my understanding regarding Cypress Cove is that is different from Sunny Cove regarding IPC.
My understanding is around 10% IPC improvemens on average for Cypress Cove vs Comet Lake and a little more if you measure SPECrate differences (+1.5% , +2% more) and Golden Cove around +19% IPC improvemens on average vs Cypress Cove (or -1.5%, -2% from that if you measure SPECrate differences.
So Golden Cove around +31% vs Comet Lake IPC (and regarding SPECrate around 28%-28.5%)
But anyway it doesn't matter what other architectures achieve in 720p gaming vs IPCxclock difference, because all your arguments (like that IPC improvements claims don't translate 100% to actual gaming performance increase which in general is valid) could be applied to Zen3 also and fail since 720p gaming performance is higher than IPCxclock difference.
Before you start to explain why Zen3 is different, don't bother, is known quantity, what is unknown is Zen4, another reason i told you for specific Zen4 technical arguments. From what i wrote you understood that i thought that you criticizing W1zzard?
My argument was that current gen games selection is plenty enough to generate distinctly different results in 720p if the testbed's VGA change from RTX 3080 to 4080, there is no need to look into the future 2-3 years from now to expect Zen4 to look better than Zen3 (or Alder Lake vs Comet Lake) in relation with what 720p difference we have now.
That's your assumption, as i replied my assumption is that current TPU games selection isn't at this stage that you assume regarding 720p and the results regarding Zen4/Zen3 difference will prove me right in few months from now when TPU upgrades from RTX3080 to something like RTX4080 or whatever from ada/RDNA3.
We will see in a few months, I won't reply again regarding this, there is no point... I told you i may be wrong, that was my impression i don't won't to research it again from the start, but can you help me point out an Intel's slide claiming Cypress Cove IPC is 18% faster than Comet Lake? (Sunny Cove comparison slides are irrelevant and also irrelevant for example are comments from reviewers using vague wording like Cypress Cove IPC seems to be similar to Sunny Cove etc, Intel slide please?)