AMD Confirms Ryzen 9 7950X3D and 7900X3D Feature 3DV Cache on Only One of the Two Chiplets

[Guwapo77]

Well just look at the boosts, the 7800X 3D is 5 GHz max. While the 7900 and 7950 3D chips are over 5.5 GHz max boost.

Maybe AMD did "solve" the voltage "issues" for OC'ing with V-cache.

From what I read a few minutes ago, OC'ing hasn't been solved. They did however up the voltage from 1.1v on 5800x3D to 1.4v on 7950x3D. Apparently, there are suppose to be some optimizations through Windows 11.

I can't wait for TPU and HU to go through their detailed breakdown of this CPU. 7950x3D is what I've been waiting for since the 5800x3D announcement.

I can't really vote on the poll. I kind of like this because cache is used mostly for games, which don't use more than 8 cores anyway, and run better on a single CCD as well. But that makes the existence of dual-CCD versions questionable. Who are they for?

Its for the work/play setups. Do some work...and run some games on breaks. The job won't even know.

 

[AnotherReader]

Because half the games tested are gpu bound even on a 4090. You can check the games here

RTX 4090 & 53 Games: Core i9-13900K vs Ryzen 7 5800X3D Review

Intel's new 13900K offers amazing gaming performance, thanks to improvements to caches, IPC and higher operating frequencies. But is that enough to beat the 3DV Cache-powered AMD Ryzen 7 5800X3D? What about platform cost?

www.techpowerup.com

I know for a fact that my 13900k at stock with 7600c34 ram is 15% faster than a maxed out 12900k running at 5.4ghz all core. I also know that said maxed out 12900k is faster than the 5800x 3d. Therefore it's obvious that the difference between the 13900k and the 5800x 3d cannot be 6% when fully cpu bound.

Unfortunately, TPU didn't test at resolutions lower than 1080p, but their conclusions disagree with yours. In their review, the 12900k is 7% slower than the 13900k when gaming at 720p. They also used DDR5 6800 overclocked to DDR5 7400 CL34 for the 13900k and DDR5 6000 CL36 for the 12900k. Now, there are games which show a greater gap between the 12900k and 13900k. Age of Empires IV shows a gap of 20%. It would probably be best to show 1% lows as they should be more sensitive to the CPU.

 

[Quicks]

Funny how when sales are not doing so well they can release a better products within 6 Months and drop prices on already released products without batting an eyelid!

 

[AusWolf]

Funny how when sales are not doing so well they can release a better products within 6 Months and drop prices on already released products without batting an eyelid!

Isn't that what you would do? With good sales, there's no rush.

 

[Makaveli]

my guess is the next x3d chips in a year or two will haver both ccd, its probably just a backup plan for if Intel comes out swinging again soon, they can just add another x3d cache to the other ccd and swing back to take the crown again.

lol. dumb, they should have just swung all the way and slam dunked.

Not sure about that.

The reason the cache is on 1 ccd is also to avoid the latency penalty of 2 ccd's.

Adding that on the other ccd will just cause both to have lower clocks instead of just 1 and will reintroduce that latency penalty.

With the Windows 11 tweaks games will choose the ccd with cache on it over the other.

Well just look at the boosts, the 7800X 3D is 5 GHz max. While the 7900 and 7950 3D chips are over 5.5 GHz max boost.

Maybe AMD did "solve" the voltage "issues" for OC'ing with V-cache.

The reason for this is the 7800X3D is a single CCD design with V-cache so clocks are lower.

The 7900 and 7950X3D chips only have v-cache on 1 ccd. The reason you are not seeing reduce clocks is because that is from the ccd without cache.

 

[JustBenching]

Unfortunately, TPU didn't test at resolutions lower than 1080p, but their conclusions disagree with yours. In their review, the 12900k is 7% slower than the 13900k when gaming at 720p. They also used DDR5 6800 overclocked to DDR5 7400 CL34 for the 13900k and DDR5 6000 CL36 for the 12900k. Now, there are games which show a greater gap between the 12900k and 13900k. Age of Empires IV shows a gap of 20%. It would probably be best to show 1% lows as they should be more sensitive to the CPU.

Some of the games they used are completely gpu bound in that test as well, like WD legions, rdr2, metro exodus, forza horizon 5, elden ring and doom eternal. When im testing I make sure that the bottleneck is the CPU and only the CPU, and I ended up at 15% difference between a stock 13900k and an overclocked 12900k. Whether you believe me or not, that's a different issue

 

[AusWolf]

Not sure about that.

The reason the cache is on 1 ccd is also to avoid the latency penalty of 2 ccd's.

Adding that on the other ccd will just cause both to have lower clocks instead of just 1 and will reintroduce that latency penalty.

With the Windows 11 tweaks games will choose the ccd with cache on it over the other.

Yep. Not to mention the extra heat.

 

[Hakker]

There are a few things going here which isn't completely explained. for the 7900X3D and 7950X3D. This will also clarify why the 7800X3D seems slower than the other two.
While the 7900X3D and 7950X3D have only 1 CCD with 3D Cache the other CCD will run at the faster speed. You can expect the 3D Cache CCD to also be limited to 4.0-4.2GHz instead of the higher base clock given. This is also why the 5800X3D was slower and why the 7800X3D has a lower base and boost clock than the other 2. The 3D Cache will likely have problems going faster than that so that CCD will run slower.
However we will see benchmarks in 6 weeks or so. However I don't see the need at this time for more than 8 cores for gaming. even seeing 6 cores used is still rare. As far as I know only Cyberpunk 2077 profits from having 8 cores. The rest is max 6 before the FPS stays the same.

 

[AnotherReader]

Some of the games they used are completely gpu bound in that test as well, like WD legions, rdr2, metro exodus, forza horizon 5, elden ring and doom eternal. When im testing I make sure that the bottleneck is the CPU and only the CPU, and I ended up at 15% difference between a stock 13900k and an overclocked 12900k. Whether you believe me or not, that's a different issue

It sounds reasonable. That's why I gave you an example of a game which showed a 20% difference. I don't know how many games you tested. However, different games stress CPUs differently, and on average, I expect the 7000X3D parts to be faster than Raptor Lake for gaming. There will be games that will favour Raptor Lake, e.g. Age of Empires IV, and others that will favour AMD. You're right that settings should be carefully chosen to minimize the GPU as a bottleneck.

 

[HD64G]

My only concern would be how they are going to keep the right workload on the 3DV cache.

What do you mean? The 3D cache is seen as a portion of the total L3 cache in the CPU and is used without any OS intervention. Only thing OS is doing is defining the order in which the CPU cores are used. And for Zen CPUs that is done in cooperation with the CPU driver which determines the best cores that are chosen to be sued firstly. If I remember correctly. If not, I am sorry for my mistake.

 

[Makaveli]

There are a few things going here which isn't completely explained. for the 7900X3D and 7950X3D. This will also clarify why the 7800X3D seems slower than the other two.
While the 7900X3D and 7950X3D have only 1 CCD with 3D Cache the other CCD will run at the faster speed. You can expect the 3D Cache CCD to also be limited to 4.0-4.2GHz instead of the higher base clock given. This is also why the 5800X3D was slower and why the 7800X3D has a lower base and boost clock than the other 2. The 3D Cache will likely have problems going faster than that so that CCD will run slower.
However we will see benchmarks in 6 weeks or so. However I don't see the need at this time for more than 8 cores for gaming. even seeing 6 cores used is still rare. As far as I know only Cyberpunk 2077 profits from having 8 cores. The rest is max 6 before the FPS stays the same.

There is more games that use 8 cores than cyperpunk.

Spiderman Remastered/ Miles Morales with RT on is very cpu heavy.

And there will be more coming this year.

 

[DemonicRyzen666]

epicfail.

Yup! It's definitely a epic fail even though there's no benchmarks yet, such a fail. SMH

Multi-socket, multi-core, multi-thread, big.LITTLE, asymmetric cache. What a scheduling nightmare.

Another half baked product in a row, keep 'em rollin'.

Is this all just speculation or do we actually know this for a fact? It sounds like speculation to me given that there is no source. I'm kind of taking issue with how this is being written as if it were fact when it's really just a guess.

Funny how when sales are not doing so well they can release a better products within 6 Months and drop prices on already released products without batting an eyelid!

Why is it the marjority of these people I quoted only post to trash something on this forum?
The trolling on this forum is getting out of hand anymore.

 

[Hakker]

DSOGaming tested it and it's also a hexacore title. Also RT does very little CPU wise. DSOGaming is a good source since it tests many games on how many cores they like.

 

[Lei]

What I don't understand is the logic behind the naming (as usual with AMD). We had the 5800X and 5800X3D, but they wanted to avoid the same confusion by having a 7700X and 7800X3D, which is fine. But then, we have a 7900X and X3D and a 7950X and X3D, which is weird. Also, what's with the rumored 10-core 7800X? Is it not a thing anymore?

Edit: IMO, it would be much simpler if we had xx00 numbers for normal CPUs, and xx50 for chips with the 3D cache. x600 could be 6-core, x700 8-core, x800 10 or 12-core and x900 16-core, and the xx50 would be the X3D.

Don't worry, they can unlaunch it and make it 399$
Perhaps because 7800x3d is 5GHz, they used naming to cover up its low clock speed.

Hardly a deal breaker...

I wish they could sell naked GPUs (only PCB)

 

[JAB Creations]

while they should have been 204 MB or 208 MB, respectively.

According to who? Does endlessly adding cache scale 1:1 for performance for eternity? I mean AMD can make this but they're not smart enough to put it together, test it and come to the conclusion that the value doesn't make sense for the potential customers?

 

[Lei]

According to who? Does endlessly adding cache scale 1:1 for performance for eternity? I mean AMD can make this but they're not smart enough to put it together, test it and come to the conclusion that the value doesn't make sense for the potential customers?

tell me why they have iGPU on a processor that is specifically meant for gaming. Who has 450$ for a cpu but not already owning a dedicated graphics card?
bcz for a non-gaming applications; 7700x is faster and cheaper, so why 7800x3d comes with embedded graphics?

 

[R-T-B]

The part I find more interesting is that the 7950X3D has a 120W TDP compared to 170W on the regular version.

Probably because OC-locked again.

No... Games have been multithreaded for a decade now. Back to 2008, most games from codemasters, activision etc... was already multithreaded. And from 2013/2014, almost every game is multithreaded.

I'm not sure if you live in triple-AAA land or what, but that's not most games. Most of the games I play still use two threads or less.

 

[Wirko]

tell me why they have iGPU on a processor that is specifically meant for gaming. Who has 450$ for a cpu but not already owning a dedicated graphics card?
bcz for a non-gaming applications; 7700x is faster and cheaper, so why 7800x3d comes with embedded graphics?

You want 7800FX3D?

 

[AusWolf]

tell me why they have iGPU on a processor that is specifically meant for gaming. Who has 450$ for a cpu but not already owning a dedicated graphics card?
bcz for a non-gaming applications; 7700x is faster and cheaper, so why 7800x3d comes with embedded graphics?

Probably because the iGPU is an integral part of the Zen 4 IO die.

I personally believe that all CPUs should come with an iGPU, at least for troubleshooting purposes.

 

[Godrilla]

Some of the games they used are completely gpu bound in that test as well, like WD legions, rdr2, metro exodus, forza horizon 5, elden ring and doom eternal. When im testing I make sure that the bottleneck is the CPU and only the CPU, and I ended up at 15% difference between a stock 13900k and an overclocked 12900k. Whether you believe me or not, that's a different issue

You can always upgrade to 10 ghz ddr5 to stay competitive.


Does anyone know if AMD used a 4090 or 7900 xtx for those first party charts?

It seems not only pricing but scheduling will make or break R9 Zen4 3d cpus.

Just for reference the 5800x3d went from $449 to $329 from January 2022 to November 2022 ( $400 in October). The potential customer base for 5800x3d was multiple generations of AM4 customers. Ryzen 4 3d chips might seemed very hyped initially but the demand should be significantly less than 5800x3d. Plus we now have 3 cpus to choose from. Hence why AMD didn't finalize prices yet.
Final thoughts wait for multiple 3d party benchmarks before purchasing decision these will likely have enough stock just based on the higher AM5 cost for new customers to swallow compared to 5800x3d drop in existing am4 mobo user base.

 

[The_Enigma]

Well, there goes my plans to buy an AM5 platform and a 7900X3D. Not happy that the vcache die is not going to boost well and the big asymmetric performance between those chiplets. Guess I'll delay till next gen again.

 

[Makaveli]

Well, there goes my plans to buy an AM5 platform and a 7900X3D. Not happy that the vcache die is not going to boost well and the big asymmetric performance between those chiplets. Guess I'll delay till next gen again.

lol you decided all of this and there hasn't even been a product review yet?

 

[The_Enigma]

lol you decided all of this and there hasn't even been a product review yet?

Yep, not shelling out $1000 for a new platform with a half assed product.

 

[umeng2002]

It's clear that there are technical limitations to making them boost with the increased voltage requirements. Performance is all that matters.

Even with the limited boost on the 7800X3D and 5800X3D, they have world class performance for gaming. Now, you literally can have your cake and eat it too.

Thread management will be an issue if Windows thinks that the highest boosting cores are the best.

 

[PenguinBelly]

my guess is the next x3d chips in a year or two will haver both ccd, its probably just a backup plan for if Intel comes out swinging again soon, they can just add another x3d cache to the other ccd and swing back to take the crown again.

lol. dumb, they should have just swung all the way and slam dunked.

You have no clue as to how these things work.