AMD Ryzen 7 9700X

[tfdsaf]

Solid power efficiency, solid thermals, but price is too high for its performance. Needs to be at most $30 dollars more expensive than the 7700x as the performance difference seems to be around 5-7% percent on average in applications and only 3% in games.

Really a missed opportunity for AMD here, wonder what the issue is with the lack of improvements over a span of almost 2 years!?

For apps Intel's 13600k seems really interesting and appealing, if only Intel recalled their bad CPU's and actually sold good CPU's that won't break in a year or two. As for games it seems like the 7800x3d remains the king.

I think AMD needs to lower their pricing in order to be competitive, at these price points the new CPU's are dead on arrival.

 

[Neo_Morpheus]

I have a feeling that AMD tried to avoid the criticism of the Zen 4 design (high temp/high power) and instead cut the power delivery way too low.

personally, i would’ve liked that they simply stated that and if you wanted top speed, use PBO.

As it is, they simply placed themselves right at the hands of the haters.

I am surprised that its not that much faster at stock, considering that even Jim The G.O.A.T. Keller mentioned how how good Zen 5 was going to be.

Thanks for the review W1zzard!

 

[A Computer Guy]

I would bet on AMD just moving from non-X/X naming to X/XT, so instead of 9700, we will see a 105W 9700XT later on.

Assuming you might be right I would be kinda pissed if this was the case and I ended up getting a 9950X when I would have intended to get a 9950XT. I guess I'm going to wait a bit longer and see if that trend emerges while also waiting for the new motherboards too. Even though 5950x is aging gracefully to the bottom of the charts it doesn't seem anywhere near unusable yet and I haven't found an AM5 Asrock board that will do x8, x8, x4, x1, x1 PCIe slots like my X570.

 

[Carillon]

I would bet on AMD just moving from non-X/X naming to X/XT, so instead of 9700, we will see a 105W 9700XT later on.

If true, that would be a really bad sign, as they would be taking away the stock cooler a non-X model usually comes with. Intel cheaped out on coolers too, the next step is toothpaste under the IHS, then going out of their way to have TSMC come up with a series of N4+++

 

[chrcoluk]

Looks a solid improvement, big power usage gains, and a 30C drop in CPU load temps compared to 7700X wow.

 

[sLowEnd]

I think AMD needs to lower their pricing in order to be competitive, at these price points the new CPU's are dead on arrival.

Launch prices are always like this. These will get sales in time after old stock has had some more time to clear out.

 

[Vayra86]

Looks a solid improvement, big power usage gains, and a 30C drop in CPU load temps compared to 7700X wow.

Yeah I didn't quite catch on to the 'underwhelming' part of this release. Sure, it doesnt break speed records, but you can't have that happening all the time, never did with Intel either.

Another thing that never happened with Intel is seeing a gen-to-gen power usage drops like that, so I think this is pretty good, there is lots of progress here. Can't always have raw performance ++. The other metrics enable more improvements in the future. Could AMD time and position their releases better... yep. But this is AMD.

 

[oZ65]

Made the chart for you:

http://imgur.com/HLLGSXh

No plans for dynamic charts at this time, even though I get why you're liking them and how random combinations could be useful.


AMD says "Memory: Some processors may be able to achieve EXPO 6400 1:1 with manual settings. By default AGESA will set any memory profile above 6000 MT/s to 1:2 mode, but an end user may override this to 1:1. Stability of this configuration will vary based on the specific processor. A latency optimized 1:1 EXPO memory profile will provide the best performance in a wide range of applications. There is no need to make any other adjustments. If you desire to make further tweaks AMD recommends trying to tighten the timings as much as possible as AUTO:1:1 DDR5-6000 MHz remains as the “sweet spot” for price and performance"

6000 works really well, 6400 requires a bit of luck and some tweaking of voltages, i.e. making it non-stock. DDR5-8000 MHz is possible, but due to the 1:2 mode it won't be that much faster. I have a G.SKILL kit coming, so will have data on this soon (not until after the 2nd round of reviews)

Great! I am eager to see 6400 1:1.

For some reason these new AMD CPU's really don't like to do AES or Powerpoint compared to their predecessors. Some office exec's will see this and be like sorry AMD we still need to buy Intel for now because Powerpoint. Maybe this is where AM5 EPYC fills the void since it's still 7000 series chips? I'm jesting a bit of course. Normally users probably won't be upgrading from 7700x to 9700x but from lower SKU's. I love this format of chart and it would be awesome to be able to pick any two CPU's from the test run to compare like this. I would want to see 5700x vs 9700x.

View attachment 357927

Really, as a programmer, I would really like to understand what the heck happened with AES.

 

[W1zzard]

Really, as a programmer, I would really like to understand what the heck happened with AES.

https://gpuopen.com/gdc-presentations/2024/GDC2024_AMD_Ryzen_Processor_Software_Optimization.pdf

This should provide some basic insights into what's involved to optimize for a given arch

 

[EcoFreak]

Wizzard, great review as always. Something important to me is idle power. I wonder if you have any insights as to how much difference motherboards make. E.G. The MiniForum options:

 

[Athlonite]

Wizzard, great review as always. Something important to me is idle power. I wonder if you have any insights as to how much difference motherboards make. E.G. The MiniForum options:

Pretty much no difference it's the IOD that causse's high idle power draw not the cores which are really efficient even if you turn off the inbuilt iGPU you'll only lose a couple of watts draw at idle

 

[oZ65]

https://gpuopen.com/gdc-presentations/2024/GDC2024_AMD_Ryzen_Processor_Software_Optimization.pdf

This should provide some basic insights into what's involved to optimize for a given arch

Thank you for your reply, but a long time ago (6-7 years ago now) I worked for a couple of years in a GPGPU environment where we wrote OpenCL code for a large verity of architectures. So I understand the problem well.

Just that such a sharp regression in AES, is not common. It would almost make me wonder if AIDA64 didn't do some strange optimization.

 

[Zach_01]

Steve at Hardware unboxed was only getting all core clocks of 4.4GHz, vs 7700X getting 5.2GHz. No wonder it uses less power. Hardly an architectural improvement.

Please think again while considering performance too
9700X does all core 4.4-4.5GHz and performs the same as 7700X at 5.2-5.3GHz.
Same performance for 35+% less power?

What exactly is an improvement to you?

I have a feeling that AMD tried to avoid the criticism of the Zen 4 design (high temp/high power) and instead cut the power delivery way too low.

personally, i would’ve liked that they simply stated that and if you wanted top speed, use PBO.

As it is, they simply placed themselves right at the hands of the haters.

I am surprised that its not that much faster at stock, considering that even Jim The G.O.A.T. Keller mentioned how how good Zen 5 was going to be.

Thanks for the review W1zzard!

I dont think its about criticism for power/temp
To me it looks like more distinct segmentation between low-high SKUs and 3DVcache ones.

 

[THU31]

Please think again while considering performance too
9700X does all core 4.4-4.5GHz and performs the same as 7700X at 5.2-5.3GHz.
Same performance for 35+% less power?

What exactly is an improvement to you?

To me this is not a valid comparison. This is 9700X by name only. It should be compared to a CPU with the same power target, which is the 7700.

9700X and 7700 both use 80 W in Blender, and 9700X is 11% faster at the same power. Sure, it's something, but this is what Intel was doing a decade ago every generation and people laughed at it.

When you enable PBO on the 9700X, you get 13% more performance in Blender at 112% more power. More than double the power! This is the only reason why the 7700X seems much less efficient, because it has a higher stock power target.

 

[oZ65]

To me this is not a valid comparison. This is 9700X by name only. It should be compared to a CPU with the same power target, which is the 7700.

9700X and 7700 both use 80 W in Blender, and 9700X is 11% faster at the same power. Sure, it's something, but this is what Intel was doing a decade ago every generation and people laughed at it.

When you enable PBO on the 9700X, you get 13% more performance in Blender at 112% more power. More than double the power! This is the only reason why the 7700X seems much less efficient, because it has a higher stock power target.

Speaking of which, the 7900 is available for 375€. The 9700X is expected at 400€. Maybe lord W1zzard ( ) could add a comparison chart for productivity since they also consume the same.

 

[Nam-A-M-D]

The latest Ryzen 7 9700X really is a disappointment. I was really hoping this was going to be worth the upgrade for lower end systems. There really wasn't any point to these at all after looking at the stats.

 

[THU31]

Speaking of which, the 7900 is available for 375€. The 9700X is expected at 400€. Maybe lord W1zzard ( ) could add a comparison chart for productivity since they also consume the same.

Yeah, that would be interesting to see. 7900 is only 15% faster in Blender, so it definitely shows that a 65 W TDP limits 12 cores a bit. But it would be cool to see all the apps in one chart.

7900X is only 29% faster than the 9700X at 2.5x the power, which shows that high boost clocks are extremely inefficient, just like on Intel. All CPUs pretty much start crumbling above 5 GHz.

 

[Nhonho]

If Intel CPUs had high IPC and low power consumption, AMD wouldn't be as accommodated as it is.

AMD should have launched the Ryzen 9000 CPUs with the memory controller integrated on the same die of the x86 cores to minimize latency the maximum.

 

[Od1sseas]

What "Magic" are we talking about exactly? My 14600KF limited to 90W, BEATS the R7 9700X.
So not only Zen 5 isn't more efficient, it still has a trash memory controller, even worse than Intel 12th gen memory controller. Laughable how behind AMD is.

 

[AMDK11]

In terms of cpu design of what was done 15 years ago with Bulldozer this is a revoultionary ground up design. As there in minal loss in any ipc in most areas.

Was Bulldozer revolutionary? It all depends on your point of view. Instead of one wide Integer block, AMD took the easy way out by using 2 smaller Integer blocks. It is known that there was no chance for two smaller blocks to cooperate within one thread. AMD could very well use 8 traditional and narrower cores in the Bulldozer instead of 4 modules (4 cores with a double Integer cluster). The "module" (core) takes the easy route because a small block of integers is easier to design and replicate within a "module" (core) than a wide block of integers. Moreover, this solution makes it much easier to fully utilize the 4-Wide decoder when two Integer clusters are loaded. AMD had a problem with core expansion, so they found an easier way to expand it and optimally use it (they had no choice)

Following the words of Mike Clark, who admitted that only the approach with two decoders in Zen5 allowed for the expansion of the Front-end. Is history repeating itself?

At least Zen5 has no IPC regression compared to Bulldozer.

The Zen5 Core is no less "revolutionary" than the Bulldozer "Module".

 

[BoggledBeagle]

What "Magic" are we talking about exactly? My 14600KF limited to 90W, BEATS the R7 9700X.
So not only Zen 5 isn't more efficient, it still has a trash memory controller, even worse than Intel 12th gen memory controller. Laughable how behind AMD is.

13600 and 14600 CPUs have so many cores that AMD simply cannot compete with them in multithread workloads with an 8 core CPU. No comparisons make sense, even an efficiency comparison.

 

[oZ65]

Yeah, that would be interesting to see. 7900 is only 15% faster in Blender, so it definitely shows that a 65 W TDP limits 12 cores a bit. But it would be cool to see all the apps in one chart.

7900X is only 29% faster than the 9700X at 2.5x the power, which shows that high boost clocks are extremely inefficient, just like on Intel. All CPUs pretty much start crumbling above 5 GHz.

Really. I am assembling a new AM5 system mainly for chess (Stockfish) and I am tempted to get a 7900 and power tune it. In theory it should give better performance/watts than the 9700X.

 

[chrcoluk]

When you enable PBO on the 9700X, you get 13% more performance in Blender at 112% more power. More than double the power! This is the only reason why the 7700X seems much less efficient, because it has a higher stock power target.

This tells me they made the right choice on the power design.

I think a chip that ran 13% faster but consuming way more power would have been a worse product.

Maybe there is an argument to be made this gen shouldnt have been released until something more substantial was ready, but I agree with their decision to push efficiency. The PC industry has been on a bad path for a while and this is a step in the right direction in making efficiency a focus again.

 

[Od1sseas]

13600 and 14600 CPUs have so many cores that AMD simply cannot compete with them in multithread workloads with an 8 core CPU. No comparisons make sense, even an efficiency comparison.

The amount of cores doesn't matter, its the final performance that matters.
The i7 12700K at 90W also scores the same as the R7 9700X.
Congrats to AMD for matching the perf/watt of a 2021 architecture. Maybe they can match Intel's 12th gen memory controller with Zen 6 in 2027 (lol)

 

[oZ65]

The amount of cores doesn't matter, its the final performance that matters.
The i7 12700K at 90W also scores the same as the R7 9700X.
Congrats to AMD for matching the perf/watt of a 2021 architecture. Maybe they can match Intel's 12th gen memory controller with Zen 6 in 2027 (lol)

What a strange hill to die on. You cannot reduce the entire performance analysis to a single piece of software. There are plenty of scenarios where both Intel and AMD suck