Info 64MB V-Cache on 5XXX Zen3 Average +15% in Games

[Kedas]

Well we know now how they will bridge the long wait to Zen4 on AM5 Q4 2022.
Production start for V-cache is end this year so too early for Zen4 so this is certainly coming to AM4.
+15% Lisa said is "like an entire architectural generation"

 

[Timorous]

It might have to be that much to be worth it compared to selling more Milan-X to Cloud. Either way it's not going to be cheap. I do get the impression that they are hesitating a bit and I assume it's because of the price.

I think they are just building stock with reject Milan-X dies but the process is mature enough that it is taking a while.

Pricing can be decided last minute. I see this as a great upgrade for those who went with a 2600(X), 2700(X), 3600(X) or 3700(X) based system. It gives them a pretty huge performance upgrade for what should be a lower cost than a 12700K + Z690 Mobo so even if price/perf of the CPU alone does not work out in AMDs favour it can still totally work out for someone with a compatible AM4 motherboard.

As a B350 Mobo owner I am hoping someone finds a way to hack that B350 Tomahawk bios that had Vermeer support and apply it to my Mortar because then it is a no brainer for me to go with a 5800X3D from my 2200G. That would see me through a GPU upgrade, maybe 2 and then I can look at the market in 5 or so years time.

 

[Kedas]

I was just checking when this thread started it's 7 months ago that they annouced it with the +15% in games, not much has changed since then.
I would think since he emphasis "it's hard" in the interview that they had some unexpected issues that pushed it back, they didn't gave a date so it could still be late spring.

 

[epsilon84]

I think they are just building stock with reject Milan-X dies but the process is mature enough that it is taking a while.

Pricing can be decided last minute. I see this as a great upgrade for those who went with a 2600(X), 2700(X), 3600(X) or 3700(X) based system. It gives them a pretty huge performance upgrade for what should be a lower cost than a 12700K + Z690 Mobo so even if price/perf of the CPU alone does not work out in AMDs favour it can still totally work out for someone with a compatible AM4 motherboard.

As a B350 Mobo owner I am hoping someone finds a way to hack that B350 Tomahawk bios that had Vermeer support and apply it to my Mortar because then it is a no brainer for me to go with a 5800X3D from my 2200G. That would see me through a GPU upgrade, maybe 2 and then I can look at the market in 5 or so years time.

Not sure I entirely agree about it being a 'great' upgrade for Zen+/Zen 2 owners, unless the rumoured price points are way off the mark and AMD prices the 5800X3D very aggresively. Generally speaking, people who actually upgrade their CPUs (whilst keeping the existing mobo) tend to stick to the same price brackets when it comes time to upgrade. You rarely see people upgrade from entry level / mainstream to flagship CPUs, for example.

Those who bought the 2600/2700/3600/3700 would have had value in mind when they initially bought those CPUs, at much lower MSRPs than the rumoured 5800X3D price. I personally fall into that bracket, my 3600/B450 combo cost me less than $300 USD back in 2019. I sure as hell ain't gonna spend $500+ now to upgrade what is a soon to be legacy platform. Now if AMD had offered a 5600X3D for $200-250, then I would be interested... that would prolong the useful lifespan of my B450 platform for another couple of years, for a reasonable cost, since I can still sell the 3600 to recoup some of that cost.

Good luck convincing someone who paid $150-$250 on their original Zen+/Zen 2 CPU to now pay 2-3 times that amount, just to have a 'halo' gaming CPU. Chances are their GPUs aren't even fast enough to truly exploit the IPC increase of the 5800X3D, unless they are running a 3080 / 6900 class GPU, which would be a very niche market indeed. How many people do you know that currently pair a 3080/3090/6900 with a Zen+/Zen 2 CPU? Many of them would already be running 5900X/5950X setups, in which case you are asking them to swap cores for IPC... probably an even harder sell!

 

[ryanjagtap]

Chances are their GPUs aren't even fast enough to truly exploit the IPC increase of the 5800X3D, unless they are running a 3080 / 6900 class GPU, which would be a very niche market indeed. How many people do you know that currently pair a 3080/3090/6900 with a Zen+/Zen 2 CPU? Many of them would already be running 5900X/5950X setups, in which case you are asking them to swap cores for IPC... probably an even harder sell!

Not disagreeing with you, but I thought people in this bracket would be on 1080p monitors and mostly would be CPU bound in games and there would be some uplift even in a low end GPU?

 

[epsilon84]

Not disagreeing with you, but I thought people in this bracket would be on 1080p monitors and mostly would be CPU bound in games and there would be some uplift even in a low end GPU?

People running 1080P monitors and low end GPUs don't spend $500+ on CPUs haha

For example, my 3600 is paired with a GTX 1070 (and I do indeed game at 1080P for this rig), which I guess you can call 'low end' these days. The system is generally not CPU bound unless we are talking older predominantly single-threaded games, in which case the performance is generally more than enough anyway, even if the CPU is the bottleneck. Without actually scientifically testing this, I would say that I am 90% GPU bound with this setup.

I cannot fathom a scenario where it makes sense for me to drop a $500-$600 CPU into this system for gaming, unless I upgrade to a high end GPU at the same time of course. Again, pretty niche scenario. If someone happens to fall into that category, then the 5800X3D is probably a good bet.

The 5800X3D kinda reminds me of the 10900K when it launched. It was the fastest gaming CPU, priced at $500+, but lacked the cores/multi-threading to challenge the higher core count Ryzen 3900X/3950X parts. It was indeed the best for gaming, but you do pay a hefty premium for the 'best gaming CPU', one that most people probably aren't willing to pay.

 

[ryanjagtap]

People running 1080P monitors and low end GPUs don't spend $500+ on CPUs haha

For example, my 3600 is paired with a GTX 1070 (and I do indeed game at 1080P for this rig), which I guess you can call 'low end' these days. The system is generally not CPU bound unless we are talking older predominantly single-threaded games, in which case the performance is generally more than enough anyway, even if the CPU is the bottleneck. Without actually scientifically testing this, I would say that I am 90% GPU bound with this setup.

I cannot fathom a scenario where it makes sense for me to drop a $500-$600 CPU into this system for gaming, unless I upgrade the GPU at the same time of course. Again, pretty niche scenario. If someone happens to fall into that category, then the 5800X3D is probably a good bet.

The 5800X3D kinda reminds me of the 10900K when it launched. It was the fastest gaming CPU, priced at $500+, but lacked the cores/multi-threading to challenge the higher core count Ryzen 3900X/3950X parts. It was indeed the best for gaming, but you do pay a hefty premium for the 'best gaming CPU', one that most people probably aren't willing to pay.

You're right, even if AMD cut prices and slot it in the 5800X place it's still not good for price conscious buyers. This maybe just a bragging rights CPU in limited quantity for gaming enthusiasts who don't go for 5900x and 5950x.
edit: They should launch 5800X3D at $499, if they're not gonna cut prices

 

[epsilon84]

You're right, even if AMD cut prices and slot it in the 5800X place it's still not good for price conscious buyers. This maybe just a bragging rights CPU in limited quantity for gaming enthusiasts who don't go for 5900x and 5950x.

Honestly, thats what I expected (or hoped) was going to happen - that the Zen3D parts will just slot into existing Zen 3 price points, and existing Zen 3 CPUs would get a price cut to better compete with the mainstream Alder Lake CPUs.

Now, with a single SKU, AMD finds itself in a tricky spot pricing wise with the 5800X3D, because its not only competing with Intel, but also with their own 5900X/5950X chips.

 

[HurleyBird]

I guess AMD still doesn't fully comprehend the value of mindshare like Nvidia does. If Jensen were in charge of AMD, we would see some exorbitantly binned (and priced) V-cache part whose sole reason for being wouldn't be to move volume itself, but rather serve as a halo product that by its nature, makes the entire Zen 3 product stack look better relative to the Alder Lake product stack.

Considering how long Zen 3 has been out, not to mention that the 5950X already boosted 200MHz higher on launch day, a clock speed regression is laughable. It's almost like AMD is sabotaging this launch on purpose, because they feel obligated to fulfill their promise but want to make the desktop part so unappealing (and late) that Milan-X production won't feel a blip.

As a gamer, I'd rather get Alder Lake which will effectively tie the 5800X3D in gaming but win at most everything else. For anyone tied to the AM4 platform, unless CPU-limited gaming performance is of extreme importance, the 5900X and 5950X both seem more attractive. Even the regular 5800X is likely to win in a lot of non-gaming benchmarks. And of course, by pushing the launch back to Q2 the looming shadow of Zen 4 starts to come into view.

Could AMD have made this product less attractive to buyers if they tried? Without getting to the point of downright absurdity, It's hard to imagine. As conspiratorial as it sounds, I am seriously wondering if this launch is being intentionally torpedoed.

It feels like for some reason, someone at AMD thought it to ask "what would the MBAs who did their best to sink Intel do?" when they should be asking "what would Nvidia do?" Nvidia is worth $700B for a reason, and a good part of that is because they don't make dumb moves like this. Don't want to impact Milan-X production? Instead of making a head-scratching disappointment, make a halo product and price it to align with very low volume. That's what Nvidia would do.

 

[eek2121]

Same process right? 400 watts, here we come!

I know you joke, but I wanted to mention power is one of the few things we do know about Raptor Lake. It will consume slightly less than ADL-S, but no big change.

 

[Hans Gruber]

People running 1080P monitors and low end GPUs don't spend $500+ on CPUs haha

For example, my 3600 is paired with a GTX 1070 (and I do indeed game at 1080P for this rig), which I guess you can call 'low end' these days. The system is generally not CPU bound unless we are talking older predominantly single-threaded games, in which case the performance is generally more than enough anyway, even if the CPU is the bottleneck. Without actually scientifically testing this, I would say that I am 90% GPU bound with this setup.

I cannot fathom a scenario where it makes sense for me to drop a $500-$600 CPU into this system for gaming, unless I upgrade to a high end GPU at the same time of course. Again, pretty niche scenario. If someone happens to fall into that category, then the 5800X3D is probably a good bet.

The 5800X3D kinda reminds me of the 10900K when it launched. It was the fastest gaming CPU, priced at $500+, but lacked the cores/multi-threading to challenge the higher core count Ryzen 3900X/3950X parts. It was indeed the best for gaming, but you do pay a hefty premium for the 'best gaming CPU', one that most people probably aren't willing to pay.

The problem with the Zen 2 (3600) is not the raw power of the chip. It's the latency issue that reduces gaming performance vs. Zen3 not to mention the 20% better IPC vs 3600. The memory latency issue was solved with Zen3 making gaming performance much improved. The raw power of the CPU was always there.

My issue is the niche v-cache concept. This is the very end of AM4. You lose mhz with v-cache vs. the standard 5800x. Then they talk about charging a premium over the 5800x even thought the CPU is well over a year old now. Add to that the Intel 12400 CPU and all the Alder Lake stuff outperforms Zen 3 now with the exception of the 5950x.

The question to current Zen3 CPU's users. Do you want to set a pile of money on fire or do you want to waste money on an end of the line Zen3 (AM4) platform CPU.

For me, I will be getting either a 5800x or 5900x. It all depends on CPU prices. I was among the earliest adopters of the Ryzen platform. With my test system a B350 motherboard R3 1200 with water cooling and B-die (Team Dark Pro) memory before anybody knew what it was. All of these components have been recycled, retired and unretired at times.

My investment was $90 in an outdated but still relevant Asus Rog Strix B-450 board. A Micron E-die 32GB dual rank 2x16GB kit for $104 on Amazon over a year ago. So I had to do something. I made a substantial investment in Windows 10 Pro keys years ago. I bought a new AIO RGB water cooler for $55 awhile back.

So basically I have two system now with one awaiting the Ryzen 3600 CPU and this current PC awaiting a Zen 3 CPU.

Honestly, thats what I expected (or hoped) was going to happen - that the Zen3D parts will just slot into existing Zen 3 price points, and existing Zen 3 CPUs would get a price cut to better compete with the mainstream Alder Lake CPUs.

Now, with a single SKU, AMD finds itself in a tricky spot pricing wise with the 5800X3D, because its not only competing with Intel, but also with their own 5900X/5950X chips.

Here is a question that needs clarification. Is the 15% in gaming performance across all resolutions 1080p/1440p/4k or are these golden numbers only @ 1080P? When you increase resolution the performance difference between CPU's is significantly less at higher resolutions. Nobody plays @1080p that is a serious gamer. 1440p is the sweet spot right now.

People seem to forget that Alder Lake has taken the performance crown with CPU's that are aggressively priced. The reason Zen 3 commanded a price premium was because it was the best CPU platform across the board.

If AMD reverts back to their previous pricing strategy on the consumer Ryzen line. Then the good times would be rolling again and all bets would be off. I am talking a 5950x for $500. That sort of thing.

 

[HurleyBird]

Another thing that wouldn't surprise me is if every 5800X3D is just a failed EPYC. The low clocks have less to do with power or heat, and are more about capturing as many of these poorly performing parts as possible. Helps explain the Q2 launch as it may take some time to build up enough inventory of low quality parts.

Of course, this theory only makes sense if for whatever reason binning needs to happen after V-cache integration and not before. Otherwise, the low quality compute dies never get selected for integration in the first place.

 

[epsilon84]

@Hans Gruber yeah I'm aware of the shortcomings of the 3600 gaming wise, though paired with a 1070 its somewhat moot. I got it for a great price in 2019 and saw a good upgrade path to Zen 3. Unfortunately I didnt anticipate a price hike for Zen 3, otherwise I would have upgraded to the 5600 or 5800 by now. AMD lost me when they wanted to charge an extra $100 for the 5600 over the 3600... and hence why I am still rocking the 3600 today. Thought Zen3D would see AMD have a top to bottom stack of SKUs... priced competitively to attack Alder Lake. Instead we are left with a single overpriced* SKU. (*supposedly)

WR to the 15% improvement in gaming, surely that is at 1080P. Would probably drop to 5-10% at 1440P and virtually no gains at 4K.

 

[Hans Gruber]

@Hans Gruber yeah I'm aware of the shortcomings of the 3600 gaming wise, though paired with a 1070 its somewhat moot. I got it for a great price in 2019 and saw a good upgrade path to Zen 3. Unfortunately I didnt anticipate a price hike for Zen 3, otherwise I would have upgraded to the 5600 or 5800 by now. AMD lost me when they wanted to charge an extra $100 for the 5600 over the 3600... and hence why I am still rocking the 3600 today. Thought Zen3D would see AMD have a top to bottom stack of SKUs... priced competitively to attack Alder Lake. Instead we are left with a single overpriced* SKU. (*supposedly)

WR to the 15% improvement in gaming, surely that is at 1080P. Would probably drop to 5-10% at 1440P and virtually no gains at 4K.

I hear you and completely understand your way of thinking. I do think AMD has lost some customers. If people were honest. Anybody who had an AMD system prior to Ryzen and after Core 2 Duo was released by intel. They would be laughed out of any forum and people would think whoever sold them a system with an AMD system rolled them. I think I paid $172 for my Ryzen 3600 through newegg via Ebay.

I am not calling the B550 and X570 motherboards a scam. But the VRM and PCB layers (luxury motherboards) are overpriced and do not return any real performance gains over standard B350/X370 and B450/X470 motherboards. I realize they added PCI-4 to the current generation of motherboards. PCI-5 is here so that was an evolutionary thing that only shows up on benchmarks and not real world performance.

The big Ryzen selling point was the performance, low energy consumption and affordable/inexpensive motherboards. I am not saying backplates on motherboards are not a nice to have feature. But $200+ motherboards with enhanced VRM performance and more PCB layers is not worth the increase in price based on performance numbers.

I have been playing the waiting game with GPU and CPU upgrades.

I pointed out in a previous post that AMD's entire line of Zen 3 CPU's are in stock at both newegg and Amazon. AMD continues to talk of supply issues but the marketplace has no shortage of AMD CPU's.

What is worse. When RDNA2 was released AMD had an executive taking bets on AMD's ability to meet the demand of AMD GPU's. People do not believe that scarcity is a manufactured phenomena. Nvidia said there was a supply issue and Nvidia has GPU's at a higher availability rate compared to AMD.

I think the 5800x3d would be 5% improvement gaming @ 1440p and 2-3% @ 4K gaming. The 15% is obviously 1080p where both AMD and Intel love to play the numbers game at 1080P. It looks good on paper.

 

[Insert_Nickname]

I sure as hell ain't gonna spend $500+ now to upgrade what is a soon to be legacy platform.

On the other hand "legacy" also means well tested and stable. AM5 is bound to bring it's own share of early adaptor bugs/issues, just like the first gen AM4.

Let's not forget the X570 USB issues either. Though they seem to be sorted.

I made a substantial investment in Windows 10 Pro keys years ago.

So long as you have a retail key, you can transfer it to a new system, no problem. Absolute worst case, you'll have to call MS to activate, but that's about it. You just can't keep running the old system with the same key, as that's against the licence.

At least in my country, MS support is very good. Can't really go into specifics, but they've been very large to me on a couple of occasions.

 

[jpiniero]

What is worse. When RDNA2 was released AMD had an executive taking bets on AMD's ability to meet the demand of AMD GPU's. People do not believe that scarcity is a manufactured phenomena. Nvidia said there was a supply issue and Nvidia has GPU's at a higher availability rate compared to AMD.

I see a couple 6700 XT and 6900 XT in stock. You won't believe the prices though.

 

[Abwx]

Another thing that wouldn't surprise me is if every 5800X3D is just a failed EPYC. The low clocks have less to do with power or heat, and are more about capturing as many of these poorly performing parts as possible. Helps explain the Q2 launch as it may take some time to build up enough inventory of low quality parts.

Of course, this theory only makes sense if for whatever reason binning needs to happen after V-cache integration and not before. Otherwise, the low quality compute dies never get selected for integration in the first place.

You re completely off, that s not a question of die quality but of thermals.
The Vcache increase substancially the thermal resistance from the CPU die to the heatsink, so they had to reduce W/mm2.

Besides die quality is quite homogenous, there isnt even 5% dispersion for the critical characteristics, otherwise there would be more than 10% frequency capability difference from a die to another.

 

[biostud]

I really don't believe AMD thought of the 5800X3D as anything but an enthusiast chip, a niche product for us enthusiast to play around with and try to keep the gaming crown. They have never planned for this to be a full update to zen3 line, but more of an experiment to see how it would be received in the market. In that regard it is more like intel's *KS chips, but without "furnace mode" and not just higher clocked, but actually a different CPU.
So those in the market for the fastest gaming CPU and those who want to see what else the extra cache can do will buy it, but otherwise it is a niche product.

 

[Shivansps]

Yeah im also on that point with a 3600, 32GB of DDR4 and a good B450 board... heck, even the 5700G is an upgrade...
So to me, i go to a 5800X or i wait for 5800X3D.

 

[Timorous]

People running 1080P monitors and low end GPUs don't spend $500+ on CPUs haha

For example, my 3600 is paired with a GTX 1070 (and I do indeed game at 1080P for this rig), which I guess you can call 'low end' these days. The system is generally not CPU bound unless we are talking older predominantly single-threaded games, in which case the performance is generally more than enough anyway, even if the CPU is the bottleneck. Without actually scientifically testing this, I would say that I am 90% GPU bound with this setup.

I cannot fathom a scenario where it makes sense for me to drop a $500-$600 CPU into this system for gaming, unless I upgrade to a high end GPU at the same time of course. Again, pretty niche scenario. If someone happens to fall into that category, then the 5800X3D is probably a good bet.

The 5800X3D kinda reminds me of the 10900K when it launched. It was the fastest gaming CPU, priced at $500+, but lacked the cores/multi-threading to challenge the higher core count Ryzen 3900X/3950X parts. It was indeed the best for gaming, but you do pay a hefty premium for the 'best gaming CPU', one that most people probably aren't willing to pay.

But they spend $500 + on CPU + RAM + Mobo combined when building a new platform? Budget gamers re-use parts and a $500 or less 5800X3D allows best in class performance for less than a platform switch to similar performance costs. Z690 DDR4 + 12700K will cost more than the 5800X3D on its own so upgraders with A320 / B450 + motherboards have an option.

As a gamer, I'd rather get Alder Lake which will effectively tie the 5800X3D in gaming but win at most everything else. For anyone tied to the AM4 platform, unless CPU-limited gaming performance is of extreme importance, the 5900X and 5950X both seem more attractive. Even the regular 5800X is likely to win in a lot of non-gaming benchmarks. And of course, by pushing the launch back to Q2 the looming shadow of Zen 4 starts to come into view.

Which ADL part? 12900K / KS? That + Z690 will cost more than a 5800X3D so why pay more for less? Zen 4 will be expensive with DDR5 as a requirement and if Genoa adoption is anything like Milan / Milan-X don't expect a huge stack with budget offerings.

Do people not buy parts that they think will last a long time? For a budget gamer who went B450 + Zen+/Zen2 they have a great chance to drop in a CPU that should give their platform another 5 years of life with just GPU upgrades and perhaps another 16GB of ram if they only have 16GB at the moment.

 

[biostud]

On the other hand "legacy" also means well tested and stable. AM5 is bound to bring it's own share of early adaptor bugs/issues, just like the first gen AM4.

Let's not forget the X570 USB issues either. Though they seem to be sorted.

That is true, but I can always buy into the AM4 platform (or RP-L) in '23 if everything turns sour

Video card pricing is still the deciding factor for my next upgrade...

 

[LightningZ71]

I don't really see the 5800x3d as being a good upgrade choice for anyone running any of the 5000x series Ryzen processors. It's still not going to be a big upgrade over the 5600x with respect to gaming. If you've got a B550 or X570 board and you're running any of the 5000g series processors, it brings both significantly better performance and platform upgrades through PCIe 4.0 on the board. If you've got any 3000 or 4000 series chip on a B450 or better board, it makes sense from a performance perspective, though, it falls short when it comes to platform by a decent amount as compared to going with a 12400-12600K processor on an H670 board with a usable amount of DDR5 or well specced DDR4. For those few out there with a 2000 series processor on an X570 board, it does make sense as it will bring a LOT of improvement for the least pain. It does nothing for all but a very few on 300 series boards.

Essentially, if you're getting the 5800x3d, you're either looking to extend your investment in an already expensive X570/B550 board by a few years (we're talking over $400 boards here), or, you have a decent 400 series board and you REALLY don't feel like ripping up your computer to get a big performance boost. My son is considering either the 5800x3d, 5900x or 5950x because he has a 2700x and had to replace his B450 board a year ago when it died and got a refurb X570 board that has been working great. The 12th gen platform really doesn't bring a lot to the table for him above what going to a 5000x series processor would do.

(EDIT) To prove my point, I found the following:
At shopblt or Neobits, there's a 2 X 8 GB DDR54800 kit for ab out $140.
ASUS has anounced their initial H670 boards and the Prime H670 Plus D4 is $160
The MSRP on the 12600KF is $264
That's a grand total of $564
That $564 gives you essentially everything from a similarly priced x570 board.

I think this puts a reasonable ceiling on the price of the 5800x3d at around $550-$600. Anything more than that and it makes zero sense to upgrade as the 12600kf should be within a few percent of it in most ST scenarios and potentially lead it in many MT scenarios.

 

[Shivansps]

I don't really see the 5800x3d as being a good upgrade choice for anyone running any of the 5000x series Ryzen processors. It's still not going to be a big upgrade over the 5600x with respect to gaming. If you've got a B550 or X570 board and you're running any of the 5000g series processors, it brings both significantly better performance and platform upgrades through PCIe 4.0 on the board. If you've got any 3000 or 4000 series chip on a B450 or better board, it makes sense from a performance perspective, though, it falls short when it comes to platform by a decent amount as compared to going with a 12400-12600K processor on an H670 board with a usable amount of DDR5 or well specced DDR4. For those few out there with a 2000 series processor on an X570 board, it does make sense as it will bring a LOT of improvement for the least pain. It does nothing for all but a very few on 300 series boards.

Essentially, if you're getting the 5800x3d, you're either looking to extend your investment in an already expensive X570/B550 board by a few years (we're talking over $400 boards here), or, you have a decent 400 series board and you REALLY don't feel like ripping up your computer to get a big performance boost. My son is considering either the 5800x3d, 5900x or 5950x because he has a 2700x and had to replace his B450 board a year ago when it died and got a refurb X570 board that has been working great. The 12th gen platform really doesn't bring a lot to the table for him above what going to a 5000x series processor would do.

Memory is another problem, people with 16GB DDR4 should look forward to AM5 and DDR5. But the ones, like me, with 32GB DDR4 (or more), we are in a point were it is better to go to Intel 12th and keep using DDR4 rather than going AM5.
We are talking about +$500 for a really crappy DDR5-4800 kit just to to keep the same amount of ram as before. Of course, this SHOULD be a lot less by the time AM5 launches.

 

[HurleyBird]

You re completely off, that s not a question of die quality but of thermals.

Launch 5950X's already had a 4.9 GHz boost, and with six quarters of process maturity you should be able to eeke out a couple hundred extra MHz too. Let's be generous and call it 100MHz. AMD said that the V-cache has a small effect on thermals. Is 500 MHz (or more) a small effect? And if the bottleneck is thermals, then also dropping base clocks by the same amount versus the vanilla 5800X doesn't make sense.

 

[Ajay]

Memory is another problem, people with 16GB DDR4 should look forward to AM5 and DDR5. But the ones, like me, with 32GB DDR4 (or more), we are in a point were it is better to go to Intel 12th and keep using DDR4 rather than going AM5.
We are talking about +$500 for a really crappy DDR5-4800 kit just to to keep the same amount of ram as before. Of course, this SHOULD be a lot less by the time AM5 launches.

True.. PMICs and other VRM components for DDR5 won’t be able able to meet demand until sometime in 22H2.

 

[epsilon84]

But they spend $500 + on CPU + RAM + Mobo combined when building a new platform? Budget gamers re-use parts and a $500 or less 5800X3D allows best in class performance for less than a platform switch to similar performance costs. Z690 DDR4 + 12700K will cost more than the 5800X3D on its own so upgraders with A320 / B450 + motherboards have an option.

Do people not buy parts that they think will last a long time? For a budget gamer who went B450 + Zen+/Zen2 they have a great chance to drop in a CPU that should give their platform another 5 years of life with just GPU upgrades and perhaps another 16GB of ram if they only have 16GB at the moment.

Anyone seriously considering buying a 5800X3D probably doesn't care that much about value though, let's be real here. If you want value for gaming you'll probably just get an i5 12400 / B660 / 32GB DDR4 and still possibly have change left over compared to a 5800X3D on its own.

I just don't see many existing Zen+/Zen2 owners rushing out to get one, assuming the rumoured $500+ pricepoint is true. The earlier Zen chips were all about value and price/performance, and I just don't see the 5800X3D being that appealing to that crowd. Agree to disagree here.