AMD Ryzen 9 7900X3D Drops to $409, to Clash with Core i7-14700K

[R0H1T]

Not true. Intel processors between Nehalem and Kaby Lake stayed at launch prices until inventory ran out years after launch.

It was longer ~ Intel only cut prices on any their desktop chips for well over a decade in 2018(?) IIRC.

This was one of the reasons why Intel was kinda hated post 2011-12 or so and NVIDIA in recent years!

 

[Panther_Seraphin]

It was longer ~ Intel only cut prices on any their desktop chips for well over a decade in 2018(?) IIRC.

This was one of the reasons why Intel was kinda hated post 2011-12 or so and NVIDIA in recent years!

Yeah 9th gen was the last time Intel held their value till late in the lifecycle.

Are you sure about that. I have been playing at 4K since this year began, are those numbers from Day one reviews? There have been updates to the micro code since then and MS have also done some refinement for Multi core CPUs. Using a 4090 means what? You do not get to use any of AMD software features that improve performance using an AM5 CPU with a Gen 5 NVME. You will probably tell me that Gen 5 NVME makes no difference too. I know that I feel a difference and at the end of the day that is all that matters. At $409 no one who buys this will be disappointed with it. It almost makes me want to send mine to TPU to show what it can do with a 7900 GPU but you can continue with your assessment as I am having too much fun to even contemplate that.

No thats how it works.....When your GPU is under heavier load the CPU difference means less. Even under a 4090 there is 1-2% at 4K but 5% at 1080. So with your 7900XT there is probably 0 difference at 4K and maybe 1-2% at 1080.

The only "fixes" M$ have added to the OS/game bar etc is better game detection so that the behaviour of locking the relevant CCDs is correct more often. When it was day 1 there was quite a few times when you would be "gaming" but the OS thought you were in "productivity" so would shift all processing to the non X3D CCD.

It was the one faux pas AMD did with X3D that Intel did better with P/E cores. AMD relied purely on a software based solution where Intel actually did hardware based scheduling in its silicon. AMD would either have to equip X3D chips with both X3D equipped CCDs or create a new I/O die revision with a new hardware scheduler in it to mitigate this completely.

 

[Imouto]

As someone who owns this chip I will tell you why this is better than the 7800X3D and 7950X3D. Heat output. You see the 7900X3D does not go above 80 degrees for me. That allows it to maintain a nice boost clock.

My 7800X3D doesn't go over 80ºC while doing a Blender render on a €40 TR Phantom Spirit. It doesn't go over 65ºC while gaming.

So maybe you wanna get your facts straight before claiming BS.

 

[Vayra86]

As someone who owns this chip I will tell you why this is better than the 7800X3D and 7950X3D. Heat output. You see the 7900X3D does not go above 80 degrees for me. That allows it to maintain a nice boost clock. I know the argument will come about it being a 6 core but that does not matter I am going to link 2 HW info screenshots.

View attachment 334695
This is after a nice session of Gaming, including TWWH3, The Gunk, AMS2 and Aliens Fire Team





View attachment 334697

As you can see all cores were loaded at 100%. Then you can see that all cores have average use at the lowest 1.4% to 11.8%

You're barely using the CPU what did you expect? Of course it won't have peak temps.

 

[kapone32]

Yeah 9th gen was the last time Intel held their value till late in the lifecycle.

No thats how it works.....When your GPU is under heavier load the CPU difference means less. Even under a 4090 there is 1-2% at 4K but 5% at 1080. So with your 7900XT there is probably 0 difference at 4K and maybe 1-2% at 1080.

The only "fixes" M$ have added to the OS/game bar etc is better game detection so that the behaviour of locking the relevant CCDs is correct more often. When it was day 1 there was quite a few times when you would be "gaming" but the OS thought you were in "productivity" so would shift all processing to the non X3D CCD.

It was the one faux pas AMD did with X3D that Intel did better with P/E cores. AMD relied purely on a software based solution where Intel actually did hardware based scheduling in its silicon. AMD would either have to equip X3D chips with both X3D equipped CCDs or create a new I/O die revision with a new hardware scheduler in it to mitigate this completely.

Indeed. Game Bar. That is not a Windows update. At the end of the day I know that I get better numbers at 4K using a 7900X3D vs any other CPU I own. So now E cores work better than the 2nd CCD in Windows? I can't believe a staff member agreed with that.
The issue though is that no one who is arguing has the actual chip. Just because you have a 7600 does not mean that is as far as it goes. Let's try an experiment since you have a different CPU. Let's get a Game at 4K that is CPU bound like TWWH3. Then let's see at 4K where the FPS in a Custom Battle using the same settings.

That will put to bed the 0% difference that you claim. As an example the 7900X3D feeds the GPU 3-5 GB/s vs the 5800X3D. I will be buying a 7900 soon and will put that to the test too. Don't forget Pcie 5.0 as well.

 

[dgianstefani]

Indeed. Game Bar. That is not a Windows update. At the end of the day I know that I get better numbers at 4K using a 7900X3D vs any other CPU I own. So now E cores work better than the 2nd CCD in Windows? I can't believe a staff member agreed with that.

Yes. Dual CCD chips have a notable disadvantage in gaming regardless of any software workarounds. It's an inherent issue. Last gen Intel is faster than your 7900X3D at 4K for this reason.

This also shows why 4K game testing for CPUs is sub optimal. I assume the Intel chips have more stable framerates, because in situations where the system is CPU limited not GPU limited, the 7800X3D is faster.

 

[kapone32]

You're barely using the CPU what did you expect? Of course it won't have peak temps.

I ran several 3D mark benchmarks during that too. I know that, we were talking about Gaming period. Not all Game engines are the same and some Games do push the CPU/GPU at 4K. I don't mean those E sports titles either. Look at the console ports that so many people cried about in 2023 that I am still trying to understand as I had none of those issues.

Yes.
This also shows why 4K game testing for CPUs is misleading.

View attachment 334732
View attachment 334733

LMAO a CPU that pulls over 150+ Watts when Gaming for less than 2% and once again there is no 7900X3D in those numbers. Did they use the Games I play in that?

 

[Tek-Check]

I'm entirely basing my statement on reality, which is that the 7950X3D (better than the 7900X3D in every way) tests 5-10% slower than a 7800X3D in gaming. The 7900X3D doesn't magically leapfrog both of those chips with a core deficiency to boot.

In fact, it's slightly slower than the 7950X3D since it has 2*6 instead of 2*8 and clocks lower. This would make it the slowest of the three in gaming. Find me a test that shows otherwise if you want to make claims.

7900X3D is a few percentages slower in gaming due to 6 cores, which is what is expected. Here is the biggest data point from 3DCenter's meta-review.
3D cache benefits 16 core CPU with ~16% uplift, 12 core CPU with ~10% uplift and 8 core CPU with ~17%

 

[Noyand]

What is that when it says all test suite? I am not talking about 1080P but 4K. Any CPU can do well at 1080P at 4K is where you see the separation. It is not 2015. Windows knows how to handle DUAL CCDs quite well. I am basing this on my use of the 5900x before I got a 5800x3D. BTW thanks for proving that a $409 is faster than a $600 Intel chip. Translation, I am quite pleased with my chip and would recommend it for high end GPUs like the 7900 and 4080/4090 GPUs. I am not talking about software but raw performance to give me over 350 FPS in AMS2.

The thing is, at launch, the 7950X3D had issues with scheduling, games would not always run on the CCD with 3D cache, wich is why the 7800X3D could be faster on average. There was no reason to think that the 7900X3d wouldn't suffer from the same issues since it's also using a hybrid CCD setup. And once the 7800X3D launched, the 7900X3D was just a jack of all trade master of none, especially for the price. You mainly game ? The 7800X3D is cheaper and faster. You mainly work ? The 7900x is cheaper and faster. Gamer Nexus reviewed all the X3d chips, and the 7900x3d was the one that they didn't like for its price performance. You are welcome to try and argue with Steve.

This review looks at AMD's new Ryzen 7 7800X3D CPU, integrating their "V-Cache" (or vertical cache) on top of a single-CCD configuration adjacent the I/O die. Unlike the AMD R9 7950X3D and AMD R9 7900X3D, this only contains one 8-core chiplet, and so should actually have some advantages in gaming scenarios when frequency isn't the main concern. Those advantages primarily show vs. the 7900X3D, which ends up a 6-core CPU (with SMT) when using the core parking feature to reduce cross-CCD latency on cache hits. Our review today primarily benchmarks the R 7 7800X3D vs. the i7-13700K, R9 7900X3D, R7 5800X3D, and 7950X3D.

We're reviewing the AMD Ryzen 9 7900X3D CPU and benchmarking it vs. the R9 7950X3D, i9-13900K, i7-13700K, i5-13600K, and more. The R9 7900X3D is a peculiar part -- it needs to be roughly equal to the 7950X3D in as many scenarios as possible, otherwise it prices itself out of the market when considering the (now much cheaper) R9 7900X non-3D price. The 7900X has been in the $430 range lately, so considering we paid $630 for the 7900X3D, there has to be a large uplift to justify that large cost increase. In the very least, the power consumption improved. This benchmark looks at gaming performance for the best CPUs out right now (in early 2023), workstation and production performance for Adobe Premiere, Photoshop, coding, and more, and power efficiency.

 

[Onasi]

This also shows why 4K game testing for CPUs is sub optimal. I assume the Intel chips have more stable framerates, because in situations where the system is CPU limited not GPU limited, the 7800X3D is faster.

True, technically 720p and, preferably, with all settings dropped to their minimum would show the most differential. However, that would be very far from real life usage. As is 4K with a 4090, to be honest. The vast majority of people will be GPU bound in AAA titles with native 4K considering which GPUs are most popular.

 

[dgianstefani]

7900X3D is a few percentages slower in gaming due to 6 cores, which is what is expected. Here is the biggest data point from 3DCenter's meta-review.
3D cache benefits 16 core CPU with ~16% uplift, 12 core CPU with ~10% uplift and 8 core CPU with ~17%
View attachment 334734

Further details from the meta review -

True, technically 720p and, preferably, with all settings dropped to their minimum would show the most differential. However, that would be very far from real life usage. As is 4K with a 4090, to be honest. The vast majority of people will be GPU bound in AAA titles with native 4K considering which GPUs are most popular.

I don't know if I'd call it far from real life, just testing the actual differences in speed of an isolated component. Doing that 720p CPU testing at launch is useful because it simulates the results you'd get pairing the CPU with unreleased next gen GPUs at commonly used resolutions for instance.

People upgrade their GPU more frequently than their platform in my experience, so CPU isolation testing is always useful.

https://www.reddit.com/r/hardware/comments/12hth3b
^ meta review in full.

 

[kapone32]

The thing is, at launch, the 7950X3D had issues with scheduling, games would not always run on the CCD with 3D cache, wich is why the 7800X3D could be faster on average. There was no reason to think that the 7900X3d wouldn't suffer from the same issues since it's also using a hybrid CCD setup. And once the 7800X3D launched, the 7900X3D was just a jack of all trade master of none, especially for the price. You mainly game ? The 7800X3D is cheaper and faster. You mainly work ? The 7900x is cheaper and faster. Gamer Nexus reviewed all the X3d chips, and the 7900x3d was the one that they didn't like for its price performance. You are welcome to try and argue with Steve.

The same person that uses an AM3 system at home? He can say whatever he wants and you have just established something. The 5.6 GHz on the 2nd CCD is good for Games that do not support Vcache. There is no 7800x3D that can run at 5.4 GHz on a single core. I have a huge Game library and some Games do not even know that CPUs come with more than 4 cores.

I also do not take anything that he talks about very seriously while he is talking about a day one review.

 

[rv8000]

I'm entirely basing my statement on reality, which is that the 7950X3D (better than the 7900X3D in every way) tests 5-10% slower than a 7800X3D in gaming. The 7900X3D doesn't magically leapfrog both of those chips with a core deficiency to boot.

In fact, it's slightly slower than the 7950X3D since it has 2*6 instead of 2*8 and clocks lower. This would make it the slowest of the three in gaming. Find me a test that shows otherwise if you want to make claims.

When loads are properly assigned to vcache cores on the 7900X3D and 7950X3D it is actually IMPOSSIBLE for the 7800X3D to be faster as its boost clock is lower than both the 7900X3D/7950X3D vcache CCDs.

Thread assignment shenanigans make hybrid CPUs (yes intel too) very annoying to obtain consistent/accurate results.

Not all tests show the 7950X3D slower by default, so incorrect.

Having owned and thoroughly tested a 7900X3D, disabling the non vcache CCD, strictly for testing purposes, it is indeed faster than a 7800X3D. It’s a shame thread hoping across CCDs can hamper performance.

*edit for some context while testing CP2077 awhile ago

Disabling the non vcache CCD resulted in +15% avg FPS and +20% min FPS

Scaling 7900X3D

CP2077 Stock Bios Settings - No EXPO/XMP,Percentile Differences RUN 1,RUN 2,RUN 3,RUN 4,AVG,Stock vs EXPO/XMP,6.90% AVG,235.56,230.3,231.8,231.67,232.3325,Stock vs 6200 Manual,9.31% MIN,90.25,101.59,55.03,55.74,75.6525,Stock vs Best Config,11.91% MAX,362.69,378.34,365.04,361.95,367.005,Stock vs...

docs.google.com

 

[Tek-Check]

Yes. Dual CCD chips have a notable disadvantage in gaming regardless of any software workarounds. It's an inherent issue. A 13600K is faster than your 7900X3D in 4K.

This also shows why 4K game testing for CPUs is sub optimal. I assume the Intel chips have more stable framerates, because in situations where the system is CPU limited not GPU limited, the 7800X3D is faster.

In 4K, the first ten CPUs or so have negligible difference of ~5fps. Nothing to split hair in half. Largely irrelevant for everyday experience.
More relevant is what happens in 720/1080p, as this is CPU-bound scenario.
In 720p, 7900X3D would land ~5% slower than 7800X3D, so somewhere near 7950X3D/14900K

 

[Tek-Check]

Gamer Nexus reviewed all the X3d chips, and the 7900x3d was the one that they didn't like for its price performance. You are welcome to try and argue with Steve.

He would now change his mind as the price has dropped almost $200. This changes the perspective radically.

 

[Colddecked]

It's not actually binning, it's simply that the cache enforced voltage limitation forces the chips to run much closer to their efficiency sweet spots, compared to the standard Zen 4 parts which try to max out frequency.

This is also shown in how late production 5800X3D chips are worse and run hotter than early production ones, since the good bins go to server chips, and just about any die will hit the target frequency anyway since it's so much lower than the standard parts non X3D.

Anecdotal and IDK if this applies to 7000 series, but my 5800x3d is undervolted by quite a bit more than what other 5000 series can do. And from what I have seen on here this seems to be quite average for x3ds. I think there is more efficient silicon being used for x3ds. Probably not the most efficient, but a bin below.

 

[dgianstefani]

When loads are properly assigned to vcache cores on the 7900X3D and 7950X3D it is actually IMPOSSIBLE for the 7800X3D to be faster as its boost clock is lower than both the 7900X3D/7950X3D vcache CCDs.

Thread assignment shenanigans make hybrid CPUs (yes intel too) very annoying to obtain consistent/accurate results.

Not all tests show the 7950X3D slower by default, so incorrect.

Having owned and thoroughly tested a 7900X3D, disabling the non vcache CCD, strictly for testing purposes, it is indeed faster than a 7800X3D. It’s a shame thread hoping across CCDs can hamper performance.

A hypothetical scenario where software scheduling has no issues doesn't change the testing done in reality, where the 8 core 7800X3D is always faster than the 6+6 core 7900X3D. Frequency isn't the only difference.

When scheduling works, and there's no issues with latency, you're right, the 7950X3D is faster than the 7800X3D, the problem is it doesn't always work in reality, as testing shows.

Anecdotal and IDK if this applies to 7000 series, but my 5800x3d is undervolted by quite a bit more than what other 5000 series can do. And from what I have seen on here this seems to be quite average for x3ds. I think there is more efficient silicon being used for x3ds. Probably not the most efficient, but a bin below.

You can undervolt non X3D chips to get the same efficiency gains.

The X3D chips can undervolt further because they don't have to hit the same frequency targets. You can't expect a 5 GHz chip to undervolt to the same voltage as a 4.5 GHz chip, while maintaining performance.

 

[Tek-Check]

Further details from the meta review -

Exactly what I wrote above.
Further details from 3DCenter just confirm that price correction was the right move, as it compensates for smaller v-cache uplift on 7900X3D.
At ~$400, now suddenly this CPU is back to life and very competitive.

 

[GenericNinja]

Exactly what I wrote above.
Further details from 3DCenter just confirm that price correction was the right move, as it compensates for smaller v-cache uplift on 7900X3D.
At ~$400, now suddenly this CPU is back to life and very competitive.

As a gaming CPU how does being slower than the 7800X3D and more expensive make the 7900X3D "very competitive"? If you claim higher clocks better just disable two cores on the 7800X3D to achieve the same clocks on your 3D cores and don't worry about 2nd CCD issues (and Ryzen 7 still cheaper).

 

[dgianstefani]

Further details from 3DCenter just confirm that price correction was the right move, as it compensates for smaller v-cache uplift on 7900X3D.
At ~$400, now suddenly this CPU is back to life and very competitive.

Sure, but this still applies.

You mainly game ? The 7800X3D is cheaper and faster. You mainly work ? The 7900x is cheaper and faster.

 

[rv8000]

Exactly what I wrote above.
Further details from 3DCenter just confirm that price correction was the right move, as it compensates for smaller v-cache uplift on 7900X3D.
At ~$400, now suddenly this CPU is back to life and very competitive.

It’s not really that theres a smaller performance uplift from the vcache, its just the 7900 is more likely to hop CCDs and hamper performance. Funny thing is a 7600X3D that clocked like a 7900X3D would be faster than a 7800X3D in any game not utilizing 8 cores.

The 7900X3D probably should’ve never been a thing without better thread assignment, which makes it look a lot worse than it actually is when it comes to gaming.

 

[dgianstefani]

It’s not really that theres a smaller performance uplift from the vcache, its just the 7900 is more likely to hop CCDs and hamper performance. Funny thing is a 7600X3D that clocked like a 7900X3D would be faster than a 7800X3D in any game not utilizing 8 cores.

The 7900X3D probably should’ve never been a thing without better thread assignment, which makes it look a lot worse than it actually is when it comes to gaming.

It exists to offload defect 8 core dies, where not all the cores work. As you say, a single CCD solution would be faster, or an 8+4 setup. But the 6+6 is a manufacturing solution to minimize waste, not a product designed to make sense for consumers first and foremost.

Reminds me of when AMD sold dual CCD 3+3 5600 and 4+4 5800 without telling the consumer or changing model numbers.

Dual-CCD Ryzen 5 5600X and Ryzen 7 5800X In the Wild

Certain AMD Ryzen 5 5600X and Ryzen 7 5800X processors are physically based on a dual-CCD design, according to an investigative report by Igor's Lab and Yuri "1usmus" Bubliy. The 5600X and 5800X are normally meant to be single-CCD processors owing to their core-counts. Based on the "Vermeer"...

www.techpowerup.com

 

[kapone32]

It exists to offload defect 8 core dies, where not all the cores work. As you say, a single CCD solution would be faster, or an 8+4 setup. But the 6+6 is a manufacturing solution to minimize waste, not a product designed to make sense for consumers first and foremost.

Reminds me of when AMD sold dual CCD 3+3 5600 and 4+4 5800 without telling the consumer or changing model numbers.

Dual-CCD Ryzen 5 5600X and Ryzen 7 5800X In the Wild

Certain AMD Ryzen 5 5600X and Ryzen 7 5800X processors are physically based on a dual-CCD design, according to an investigative report by Igor's Lab and Yuri "1usmus" Bubliy. The 5600X and 5800X are normally meant to be single-CCD processors owing to their core-counts. Based on the "Vermeer"...

www.techpowerup.com

So I guess the 5900X was useless based on your assessment

 

[Onasi]

So I guess the 5900X was useless based on your assessment

Can confirm as an owner. The bloody thing constantly tells me to go f**k myself any time I try to run Blender or load up a game. Just a BSOD with a “Go eff yourself. Love, your 5900X”. Lazy bum.

 

[dgianstefani]

So I guess the 5900X was useless based on your assessment

Less useless than the 7900X3D which is gaming focused due to being X3D, but slower than cheaper alternatives for both gaming and productivity.

At least the 5900X is faster in it's intended role - productivity, than the 5800X.