AMD confirms Ryzen 7 5800X3D launches this spring, Zen4 Raphael in 2H 2022

[random_matt]

Makes you think what is the best method for cooling. Although, as it is all locked down then perhaps even a full loop won't help enough. Still may buy one out of curiosity.

 

[humbug]

As i said it will be more optimised for Zen 4. Kudos to AMD for the innovation, it can significantly boost gaming performance without architectural changes, i can't wait for Zen 4 to see it, ofcourse it all depends how good RPL is, if RPL fail then they won't need to release special 3d version.

Rumour is still 16 core CPU but 5Ghz+ all core and 20 to 30% IPC

At 4.8Ghz a 5950X scores 32,393, with clock speed + IPC (lets say 25% over that 5950X) the 7950X should score 40,500, i think it will be tough for RTL to beat that and AMD aren't even adding cores. they still have that option.

 

[Vulture]

Rumour is still 16 core CPU but 5Ghz+ all core and 20 to 30% IPC

At 4.8Ghz a 5950X scores 32,393, with clock speed + IPC (lets say 25% over that 5950X) the 7950X should score 40,500, i think it will be tough for RTL to beat that and AMD aren't even adding cores. they still have that option.

Chiplet design allows them to pack more cores, and all big unlike Intel. RPL with monolithic 8 big cores and 16 small will have huge die size, which is costly.

 

[humbug]

Chiplet design allows them to pack more cores, and all big unlike Intel. RPL with monolithic 8 big cores and 16 small will have huge die size, which is costly.

Right, just imagine that 40,000 score and another 50% cores on top, making it a 24 core, a score of 60,000, a 2.2X the performance the 12900K.

 

[Joxeon]

Rumour is still 16 core CPU but 5Ghz+ all core and 20 to 30% IPC

At 4.8Ghz a 5950X scores 32,393, with clock speed + IPC (lets say 25% over that 5950X) the 7950X should score 40,500, i think it will be tough for RTL to beat that and AMD aren't even adding cores. they still have that option.

I'd be supprised if it runs 5ghz all core in cinebench, AMDs demo said 5ghz but that was in a gaming workload so that would imply heavy workloads won't hit 5ghz.

 

[humbug]

I'd be supprised if it runs 5ghz all core in cinebench, AMDs demo said 5ghz but that was in a gaming workload so that would imply heavy workloads won't hit 5ghz.

Getting my 5800X running all core 4.8Ghz in Cinebench is not difficult and people have done it with the 5950X, i can do it without unlocking the socket limits tho for the 5950X you certainly do, 4.7 - 4.8Ghz on Zen 3 in heavy workloads is already completely doable, in fact the 5800X runs at 4.6 to 4.7Ghz all core out of the box in Cinebench.

The extra 300Mhz is not so much an architecture limitation, tho 100 or 200Mhz might be with Zen 3, with a few tweaks and most importantly much higher power levels with the LGA socket yes i can see Zen 4 doing 5Ghz all core in heavy workloads, maybe even slightly higher, Zen 3 is almost there.

 

[Grim5]

Rumour is still 16 core CPU but 5Ghz+ all core and 20 to 30% IPC

At 4.8Ghz a 5950X scores 32,393, with clock speed + IPC (lets say 25% over that 5950X) the 7950X should score 40,500, i think it will be tough for RTL to beat that and AMD aren't even adding cores. they still have that option.

lol nah, go look at what the clock speed and IPC gain is between zen 2 and 3 and then compare r23 scores from reviews between 3950x and 5950x. 5950x is only scoring 20% higher score despite running several hundred MHz faster AND having 20% IPC advantage - since I own both I can also tell you that my 3950x with PBO turned on scores higher than my 5950x with PBO off

 

[humbug]

lol nah, go look at what the clock speed and IPC gain is between zen 2 and 3 and then compare r23 scores between 3950x and 5950x. 5950x is only scoring 20% higher score despite running several hundred MHz faster AND having 20% IPC advantage - since I own both I can also tell you that my 3950x with PBO turned on scores higher than my 5950x with PBO off

Its 12%, and the Zen 3 core is an iteration of the Zen 2 core, Zen 4 is a different core.

 

[Vulture]

Zen 4 will be built on completely new platform so no limits of AM4 and also new node which is superior to Intel's, so i believe they will beat Intel in ST performance. Talking about AM4 it is impressive how much gains AMD achived on single platform and same node (Zen 2 to Zen 3).

 

[Joxeon]

Getting my 5800X running all core 4.8Ghz in Cinebench is not difficult and people have done it with the 5950X, i can do it without unlocking the socket limits tho for the 5950X you certainly do, 4.7 - 4.8Ghz on Zen 3 in heavy workloads is already completely doable, in fact the 5800X runs at 4.6 to 4.7Ghz all core out of the box in Cinebench.

The extra 300Mhz is not so much an architecture limitation, tho 100 or 200Mhz might be with Zen 3, with a few tweaks and most importantly much higher power levels with the LGA socket yes i can see Zen 4 doing 5Ghz all core in heavy workloads, maybe even slightly higher, Zen 3 is almost there.

4.8 is not 5ghz though and even that's a stretch for most chips which tap out at 4.7, if the all core boost in gaming is 5ghz then you can bet the all core boost in heavy workloads will be less, even Intels 5.3ghz boosting chips struggle to hit 5ghz all core depending on the workload. Then there's trying to cool those small 5nm chiplets which if the TDP is increasing will be pulling more power.

The 13900k should score around 39000 as 8 Ecores will add 8000 then 10% IPC on top.

 

[Vulture]

4.8 is not 5ghz though and even that's a stretch for most chips which tap out at 4.7, if the all core boost in gaming is 5ghz then you can bet the all core boost in heavy workloads will be less, even Intels 5.3ghz boosting chips struggle to hit 5ghz all core depending on the workload. Then there's trying to cool those small 5nm chiplets which if the TDP is increasing will be pulling more power.

The 13900k should score around 39000 as 8 Ecores will add 8000 then 10% IPC on top.

I doubt, it is only polished ADL based on the same node, quote from one site: "During its Investors Meeting event for 2022, Intel demoed an engineering sample of next-generation Raptor Lake, while promising up to double-digit performance uplift compared to today’s Alder Lake silicon. Reading between the lines, IPC is likely to increase by only a few per cent."

So up to double digit performance uplift, hmmm.

 

[humbug]

4.8 is not 5ghz though and even that's a stretch for most chips which tap out at 4.7, if the all core boost in gaming is 5ghz then you can bet the all core boost in heavy workloads will be less, even Intels 5.3ghz boosting chips struggle to hit 5ghz all core depending on the workload. Then there's trying to cool those small 5nm chiplets which if the TDP is increasing will be pulling more power.

The 13900k should score around 39000 as 8 Ecores will add 8000 then 10% IPC on top.

Even assuming that's true, assuming 4.8Ghz is the all core clock, with 20% IPC it still comes to 39,000.

 

[Joxeon]

I doubt, it is only polished ADL based on the same node, quote from one site: "During its Investors Meeting event for 2022, Intel demoed an engineering sample of next-generation Raptor Lake, while promising up to double-digit performance uplift compared to today’s Alder Lake silicon. Reading between the lines, IPC is likely to increase by only a few per cent."

So up to double digit performance uplift, hmmm.

We know 8 Ecores adds around 8000 so the minimum score for RPL will be 35500 and that's just by adding Ecores, I'm sure there will be some IPC and clock improvements as like you say it's on the same node so will see refinements as the silicon quality improves.

 

[Joxeon]

IMO all Intel needs to do with RPL is be close enough to Zen 4 while having the substantially cheaper platform costs.

 

[Vulture]

Even if MT performance are similar it is still win for AMD, they dind't need to increase cores and change design which saves them money, while having tangible advantage in ST perfomance. As i said, Intel monolithic 8P + 16e cores will have huge die size, and then the cost of software optimizations for big/little. They will have to migrate to MCM design so they could add more P cores and completely eliminate those e cores which is painful from software perspective, and i'm quite sure they will kill big/little once they solve a challenge of creating MCM design.

 

[humbug]

Even if MT performance are similar it is still win for AMD, they dind't need to increase cores and change design which saves them money, while having tangible advantage in ST perfomance. As i said, Intel monolithic 8P + 16e cores will have huge die size, and then the cost of software optimizations for big/little. They will have to migrate to MCM design so they could add more P cores and completely eliminate those e cores which is painful from software perspective, and i'm quite sure they will kill big/little once they solve a challenge of creating MCM design.

Just keep piling little cores in to it, i mean you're already getting 8 + 16 with RPL, then what? 8+ 24? Now another 8 doesn't really make much difference so its 8 + 36 and on and on....

To put that into perspective Genoa (Zen 4 Epyc) will have 96 BIG cores and 192 threads, Bergamo (Zen 4+ Epyc) 128 BIG cores and 256 threads, in 2013 did we ever think that in 10 years we would see 100 core CPU's and beyond that.... ? 256 cores and 512 threads anyone?

That's the power of AMD's "Glue"

 

[Perfect_Chaos]

IMO all Intel needs to do with RPL is be close enough to Zen 4 while having the substantially cheaper platform costs.

I think there's a good chance Intel will offer better price/performance, not sure about their boards though. Not going to change platform to Raptor Lake when the boards won't allow any upgrades either, it's either Zen 4 or Meteor Lake for me so i can get at least 1 upgrade out of a new board.

 

[Calin Banc]

The new parts MSRPs really goes to show how much AMD was milking consumers with the original SKUs especially so given the subsequent TSMC price rises and the fact that AMD can still afford to cut 50% off the mainstream lineup and are making a profit.

They are a premium company, so customers were paying a premium price ).

Jokes aside, they're in a position now to charge more and still sell as they have gained some mindshare back. Same practice applies to GPU market as well.

Really? If so then you could say this about teh manufactureere of most products. How many price reductions do Playstations and Xbox have over there life cycle? Does this mean the Sony and Microsoft was milking the customer when laynching the product or where they covering R&D costs and production costs of a new product (which would decrease over time) most manufactureers reduce prices towards the end of a products life cycle. It is as simple as that

Lol, how much of a CPU, MB, GPU, fast SSD, PSU and so on you'd buy from the same price of a console? And wasn't that Microsoft and Sony were making profit on those little prices anyway? PS5 has a lower price than PS3 if I'm not mistaken.

 

[q974739]

IMO all Intel needs to do with RPL is be close enough to Zen 4 while having the substantially cheaper platform costs.

The platform cost has to be cheap enough to compensate for the AMD motherboard reuse advantage. In the same way that the current generation has all the upgraders looking at the 5800x3D, knowing that you can get a cheap chip and board and then upgrades later...

That's going to have to be a significantly cheaper platform, or decent performance advantage.

 

[cliffy]

Lol, how much of a CPU, MB, GPU, fast SSD, PSU and so on you'd buy from the same price of a console?

Lol, that was not the point