Poll: How much do you think the AMD 5800X3D will cost at launch?

[ChocAndGray]

I think it's zero chance now AMD have spoken about the cache's benefits. If you read between the lines of the marketing spiel and interviews, productivity workloads don't benefit anywhere near as much from the cache as gaming does.

So, AMD chose to focus on "best use case" for the cache, and that was to create the ultimate AM4 gaming CPU in the 5800X3D. If you won't see much improvement in productivity workloads then there's little point in "wasting" precious 3D chiplets for 12 and/or 16 cores SKUs. Similarly, with Milan-X taking up the lion's share of the 3D chiplets, that precludes doing the 2nd ultimate AM4 CPU with a 5600X3D.

That's a shame, I'd have "upgraded" to a 5950x with 3d cache.

I guess the sensible choice for me would be to just wait for amd's next platform at this stage.

 

[LePhuronn]

That's a shame, I'd have "upgraded" to a 5950x with 3d cache.

Indeed. From a tech enthusiast standpoint, it's a shame the cache didn't boost productivity as well as gaming. Purely from a competition standpoint, a 5950X3D that definitively takes the productivity performance crown from the 12900K would've been great, and keeps AND's mindshare strong.

"Oh look, Intel have made great strides in returning to competitiveness with Alder Lake. But all we need to do is slap some cache on our 2 year old arch and it beats Intel again. So just think what our new one will do..."

 

[CAT-THE-FIFTH]

Indeed. From a tech enthusiast standpoint, it's a shame the cache didn't boost productivity as well as gaming. Purely from a competition standpoint, a 5950X3D that definitively takes the productivity performance crown from the 12900K would've been great, and keeps AND's mindshare strong.

"Oh look, Intel have made great strides in returning to competitiveness with Alder Lake. But all we need to do is slap some cache on our 2 year old arch and it beats Intel again. So just think what our new one will do..."

I wouldn't say it will have zero implications for productivity. The Zen2/Zen3 CPUs used to do relatively well in certain applications such as Adobe Lightroom with batch exports,and I suspect the 32MB of L3 cache was the reason. So I can see some situations where having 96MB of L3 cache will help.

 

[Perfect_Chaos]

That's a shame, I'd have "upgraded" to a 5950x with 3d cache.

I guess the sensible choice for me would be to just wait for amd's next platform at this stage.

If you go Intel you likely only have 1 upgrade left being Raptor Lake. Then you need to wait for Meteor Lake and the boards supporting it just to know you should get at least another chip to upgrade to.

So for upgraders the best choice is probably either wait for AM5 or wait for Meteor Lake. It depends who has the biggest advantage in overall speed per core for me, not the amount of cores.

 

[LePhuronn]

I wouldn't say it will have zero implications for productivity. The Zen2/Zen3 CPUs used to do relatively well in certain applications such as Adobe Lightroom with batch exports, and I suspect the 32MB of L3 cache was the reason. So I can see some situations where having 96MB of L3 cache will help.

True true, but I think the deal is the cache didn't offer enough of a boost in a large variety of workloads for AMD to bother. It's about marketing and optics after all.

If demand for Milan-X wasn't so high, I could see AMD doing a 5950X3D and then using very specific benchmarks to push the productivity angle, but there just ain't enough chiplets to go round.

No doubt Zen 4 has the cache as standard so we'll see plenty of productivity benches over Zen 3 when the times comes.

 

[CAT-THE-FIFTH]

True true, but I think the deal is the cache didn't offer enough of a boost in a large variety of workloads for AMD to bother. It's about marketing and optics after all.

If demand for Milan-X wasn't so high, I could see AMD doing a 5950X3D and then using very specific benchmarks to push the productivity angle, but there just ain't enough chiplets to go round.

No doubt Zen 4 has the cache as standard so we'll see plenty of productivity benches over Zen 3 when the times comes.

I think because its limited to an 8 core SKU,the 16 core models would beat it overall because of sheer core count. I do wonder whether we were meant to get a full range of models,but due to the pandemic it pushed this back,and its too close to the Zen4 for AMD to commit too many resources to it? IIRC,the initial demo was with a Ryzen 9 5900X.

 

[LePhuronn]

I think it simply comes down to demand for Milan-X taking all the chiplets.

AMD will have known from the outset that big chunks o cache will benefit some workloads better than others, it's not like they're slapping things on blindly. Since a chiplet is a chiplet is a chiplet, I do believe that AMD's intention was a full 5000 series refresh with 3D cache which would close the gap on Alder Lake for 6-9 months before Zen 4 comes along.

But then everything got in the way so AMD changed their plans.

I recall AMD saying that demand for Milan-X was higher than anticipated, so straight away that means Ryzen X3D is getting cut back. Then I'm sure something else Covid-y got in the way which pushed Ryzen X3D release back to Q2. So now AMD are in the situation where they don't have as many chiplets as they planned for Ryzen and the delayed release puts it within 6 months of Zen 4.

So, spin the marketing narrative that 3D cache is part of AMD exploring diversification into custom SKUs, and the best desktop use for 3D cache is for gaming, so they release the definitive gaming CPU: 5800X3D.

I think because its limited to an 8 core SKU,the 16 core models would beat it overall because of sheer core count.

But would it? How much does a 5950X beat a 5800X in gaming? That's the marketing narrative for the 5800X3D: it's a dedicated gaming CPU. So would a 5950X3D beat a 5800X3D in gaming?

The "limited to 8 core SKU" is simply because they don't have the chiplets to make any other SKUs. Had Milan-X not been as popular then AMD will have more 3D chiplets to go around, including the 6 core chiplets for 6 and 12 core CPU options.

Even if AMD were genuinely surprised in final tests that the cache boost only benefited gaming on desktop, the fact we don't also have a 5600X3D to use those 6 core chiplets pretty much confirms Milan-X is gobbling everything up.

 

[CAT-THE-FIFTH]

But would it? How much does a 5950X beat a 5800X in gaming? That's the marketing narrative for the 5800X3D: it's a dedicated gaming CPU. So would a 5950X3D beat a 5800X3D in gaming?

The "limited to 8 core SKU" is simply because they don't have the chiplets to make any other SKUs. Had Milan-X not been as popular then AMD will have more 3D chiplets to go around, including the 6 core chiplets for 6 and 12 core CPU options.

Even if AMD were genuinely surprised in final tests that the cache boost only benefited gaming on desktop, the fact we don't also have a 5600X3D to use those 6 core chiplets pretty much confirms Milan-X is gobbling everything up.

I was talking more about non-gaming and why AMD didn't talk so much about non-gaming performance of a Ryzen 7 5800X3D. A Ryzen 9 5900X/5950X probably is faster in many of those scenarios(apart from the few I mentioned) as a lot of applications can leverage the extra threads.

It's a shame as AMD demoed the technology on a Ryzen 9 5900X. But at the same time,from a marketing perspective,it would be far better to compare Zen4 to the bog standard Ryzen 9 CPUs. I suspect a Ryzen 9 5950X3D would be a productivity monster. It's why so many server/supercomputer CPUs tended to have large amounts of cache like Milan X.

 

[LePhuronn]

I know, and that's why Milan-X and AMD's marketing spiel is important.

Yes, by virtue of having double the cores, a 5950X3D would outperform the 5800X3D in non-gaming tasks, in exactly the same that the 5950X outperforms the 5800X. But the crucial point is does the additional cache benefit enough of those non-gaming tasks in a significant way to be worth taking away from Milan-X stock?

The answer is "more than likely not", otherwise we'd have a 5950X3D and the marketing spiel of "we focussed on best use cases for our technology, and the best desktop use is gaming. So here's the definitive gaming CPU".

 

[CAT-THE-FIFTH]

Personally I think they should have done a Ryzen 7 5800X3D and a Ryzen 9 5950X3D,that way you would have the best gaming CPU and the fastest desktop socket CPU ever made. But the rumours of Zen4 being pushed forward,again makes me think Zen3D was more likely meant to have launched by now,and the pandemic screwed up timelines. So AMD has decided to launch one SKU in limited quantities just to poke Intel in the eye.

In some ways Intel launching with DDR5 first probably has helped AMD a bit,because Intel will have to take all the flack for DDR5 teething problems and costs.

 

[LePhuronn]

...makes me think Zen3D was more likely meant to have launched by now,and the pandemic screwed up timelines. So AMD has decided to launch one SKU in limited quantities just to poke Intel in the eye.

It does look like that. Intel have been banging on about gaming performance every since Coffee Lake, and now there's a very real chance Alder Lake retaking the crown will be short lived.

 

[CAT-THE-FIFTH]

It does look like that. Intel have been banging on about gaming performance every since Coffee Lake, and now there's a very real chance Alder Lake retaking the crown will be short lived.

Well we all know Intel being too comfortable doesn't end well.

 

[LePhuronn]

Well we all know Intel being too comfortable doesn't end well.

Look, I know we shouldn't wish ill on anybody because it can stifle competition, but if things play out the way they seem to be playing out, I am gonna chuckle most heartily at Intel getting spanked with zero response for another year.

Don't get me wrong, Alder Lake is a great return to form, but if all it takes to regain the gaming crown is slap some cache on a 2 year arch then Raptor Lake is gonna be in for a world of hurt for 12 months in the face of Zen 4; Raptor Lake is incremental and refinement, Zen 4 is brand new.

 

[CAT-THE-FIFTH]

Look, I know we shouldn't wish ill on anybody because it can stifle competition, but if things play out the way they seem to be playing out, I am gonna chuckle most heartily at Intel getting spanked with zero response for another year.

Don't get me wrong, Alder Lake is a great return to form, but if all it takes to regain the gaming crown is slap some cache on a 2 year arch then Raptor Lake is gonna be in for a world of hurt for 12 months in the face of Zen 4; Raptor Lake is incremental and refinement, Zen 4 is brand new.

Well,at least it will mean the motherboards will be forced downwards in price!

 

[LePhuronn]

Well,at least it will mean the motherboards will be forced downwards in price!

pfft, fat chance

 

[CAT-THE-FIFTH]

pfft, fat chance

It will be with falling sales when Intel will need to rebate their complaining board partners!

 

[Joxeon]

Look, I know we shouldn't wish ill on anybody because it can stifle competition, but if things play out the way they seem to be playing out, I am gonna chuckle most heartily at Intel getting spanked with zero response for another year.

Don't get me wrong, Alder Lake is a great return to form, but if all it takes to regain the gaming crown is slap some cache on a 2 year arch then Raptor Lake is gonna be in for a world of hurt for 12 months in the face of Zen 4; Raptor Lake is incremental and refinement, Zen 4 is brand new.

Don't forget that Zen 3 was on the same 7nm as zen 2 yet AMD managed big gains so there is no reason why Intel can't achieve decent gains especially now they are finally off 14nm.

 

[Vulture]

Don't forget that Zen 3 was on the same 7nm as zen 2 yet AMD managed big gains so there is no reason why Intel can't achieve decent gains especially now they are finally off 14nm.

You are right, that's why im so excited about Zen 4 cuz it will be based on completely new node, 5nm is good improvement when compared to 7nm, and also I/O die will be based on TSMC 6nm which is huge improvement to gloflo 12nm which was little bottleneck, so when you add all of this into equation 25%+ IPC gain is realistic.

 

[LePhuronn]

Don't forget that Zen 3 was on the same 7nm as zen 2 yet AMD managed big gains so there is no reason why Intel can't achieve decent gains especially now they are finally off 14nm.

Zen 3 achieved big gains by being a significant redesign of Zen 2. Intel's 10nm SuperFin looks already maxed out, Raptor Lake is only a refinement of Alder Lake and Meteor Lake is on a different node again. Intel being off 14nm seems to have done very little because their 10nm is a mess, and the same caveats will come into place with Meteor Lake when they're finally off that 10nm.

The potential is there of course, but just because company X can do something doesn't mean company Y is similarly capable.

 

[Joxeon]

Zen 3 achieved big gains by being a significant redesign of Zen 2. Intel's 10nm SuperFin looks already maxed out, Raptor Lake is only a refinement of Alder Lake and Meteor Lake is on a different node again. Intel being off 14nm seems to have done very little because their 10nm is a mess, and the same caveats will come into place with Meteor Lake when they're finally off that 10nm.

The potential is there of course, but just because company X can do something doesn't mean company Y is similarly capable.

I wouldn't really call ADL a mess when it beats Zen 3 in ST MT and gaming, RPL will be over doubling the cache which according to AMDs own Vcache gives a large boost in gaming and coupled with the increased core counts which will further increase Intels MT lead especially on the lower end SKUs where that lead is already 40% it's looking decent for RPL unless Zen 4 will see an increase to core counts throughout the stack.

Also when you factor in being able to pick up a cheap Z690 once 700 series boards are out and some cheap DDR4 ram or use existing ram vs X670 and the costly DDR5 which will still not be hitting the peak speeds for another couple of years yet then it becomes a very costly upgrade to Zen4 as no doubt AMD will also look to increase CPU prices again.