Speculation: i9-9900K is Intel's last hurrah in gaming

Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Will Intel lose it's gaming CPU lead in 2019?


  • Total voters
    184
  • Poll closed .

ondma

Platinum Member
Mar 18, 2018
2,718
1,278
136
Having an 8700k and reading what i’m reading about the 9900k. Getting any satisfaction from this upgrade (sidegrade?) would involve delidding and lapping IMHO.

So apparently it seems that even though they went back to solder which is what everybody wanted they managed to increase the thermal resistance of this setup?


The silicon die is thicker and the solder somehow prevents efficient transfer of heat?

So this company spending billions of dollars and struggling to deliver upgrade options for its gaming market is delivering this kind of construction in the CPU and IHS?

How about maybe taking some pocket change out of the billions they spend on R&D and maybe design a package that optimizes the removal of heat from the CPU?


Are they saving that ace up the sleeve for their next screwup?

We have suffered for years with this horrible design. Would it really cost them too much to ensure the die is prelapped and fully flat? To include a liquid metal TIM and copper IHS that is also flat?

On an industrial scale maybe $20 extra? I know that’s a lot of money. But we are talking about top end chips for gamers. Maybe $100 premium? It would certainly cost more to have it done by any independent service.

Is there something about offering that which is not feasible?

Right now they could make a 10000k which is a 9900k just properly lapped and sell it for $599. If people are paying $499 why wouldn’t they pay $599?

The coolers such purchasers buy are $150-450. Some people say they are running 3x360 loops. The issue is the CPU to Block heat resistance.

Given the huge engineering challenges they have which need $billions to overcome why would such a low hanging fruit be left just lying there?
According to Silicon Lottery, deliding has a minimal effect on temperatures of the 9900k. The heat issues come from the fact that there are 2 more cores to deal with and the cpu is capable of very high clockspeeds.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
It’s not surprising anywhere. It’s human nature to support the underdog.

The underdog in this case is the end user looking for a solid deal, not any of the megacorps themselves. It's why I can't rightly understand people blindly supporting Intel/AMD/Nvidia/Apple/Samsung/etc.

For those of who have been around since before the XT/AT days, we've seen all number of hype, success, and disappointment. Patterns start to emerge for sure. It's always exciting when there's a Radeon 9700, a GTX8800, a Conroe, or an Opteron/AMD64 moment in the industry. But for every one of those, there are at least as many underwhelming products, and between the GOATs and the duds, the vast majority of it is just an assortment of products that move the needle here or there, but whose primary purpose is to add to the coffers of the megacorps at the prices that the market will bear.

Given all of that, looking behind the scenes, my realistic expectations are for Zen 3000 to be another solid Zen family stack. Nothing shockingly good, and not embarrassing either. CPU performance has been pretty stagnant for so damn long that AMD improving to be a mix of better, on-par and 'close enough' with Intel is a good development. So I think we will see improved IPC, some pop to the clocks speeds, but I also think we aren't going to see quite the ludicrous offerings 'leaked' by the YouTube man either. Primarily it just doesn't make sense from a market standpoint. If you have something literally better in every way with double the cores and lower power consumption than the opposing flagship, why price it identical? Makes no sense, at best.

We've seen AMD with flagship offerings more than once, and when they command the top levels, those things are $1k. And, as much as we wouldn't like for the bargain hunters among us, you know the market would happily support something literally twice as good as a 9900k @ $1000. There's a reason why the 2080ti sells so well. If they were able to offer both superior IPC and hit 5+GHz, then at best they might slot in $449 8C/16T Zen2 there. Buyers would eat it up and have no real reason whatsoever to choose the 9900k in that case. And so on, undercutting Intel across the board with superior offerings.

The leaks OTOH seem like complete fanfiction, and they've been spread so far by that one flake that a ton of other outlets keep re-sourcing from each other in a weird online game of telephone. Mind you, this is a guy who told someone that disagreed with his opinion to kill themselves, who mass deletes his comments on Reddit, etc. Dude is a bit of a nutter butter, look it up.

I'll he hoping, but beneath a protective layer of cynicism :) I have a couple of waiting AM4 builds for some happy upgrades depending on how it falls.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
Having used both an 8700k, a 9700k, a delidded 8086k, and a 9900k, I have to say, they're all pretty interchangeable unless :

You have a 2080ti and are REALLY pushing for 165/244hz at higher ranges (in the few games that can utilize the 8C/16T to some extent)

You do a crapton of multitasking, or streaming while gaming. Or you want an extra push at large file encoding.

A blind switch between any of those 4 CPUs in 99% of games @ 1440p+ with anything less than a 2080ti, I don't think anyone would be able to tell the difference.

At best, I think it would be advisable to at least see the real offerings on Zen2 SKUs before investing in a 9900k. :)
 
  • Like
Reactions: Tlh97

happy medium

Lifer
Jun 8, 2003
14,387
480
126
Having an 8700k and reading what i’m reading about the 9900k. Getting any satisfaction from this upgrade (sidegrade?) would involve delidding and lapping IMHO.

So apparently it seems that even though they went back to solder which is what everybody wanted they managed to increase the thermal resistance of this setup?


The silicon die is thicker and the solder somehow prevents efficient transfer of heat?

So this company spending billions of dollars and struggling to deliver upgrade options for its gaming market is delivering this kind of construction in the CPU and IHS?

How about maybe taking some pocket change out of the billions they spend on R&D and maybe design a package that optimizes the removal of heat from the CPU?


Are they saving that ace up the sleeve for their next screwup?

We have suffered for years with this horrible design. Would it really cost them too much to ensure the die is prelapped and fully flat? To include a liquid metal TIM and copper IHS that is also flat?

On an industrial scale maybe $20 extra? I know that’s a lot of money. But we are talking about top end chips for gamers. Maybe $100 premium? It would certainly cost more to have it done by any independent service.

Is there something about offering that which is not feasible?

Right now they could make a 10000k which is a 9900k just properly lapped and sell it for $599. If people are paying $499 why wouldn’t they pay $599?

The coolers such purchasers buy are $150-450. Some people say they are running 3x360 loops. The issue is the CPU to Block heat resistance.

Given the huge engineering challenges they have which need $billions to overcome why would such a low hanging fruit be left just lying there?
You have a 8700k, the second fastest CPU on the planet, I don t see a reason why you need to upgrade anyhow.
Everybody with any sense knows that a 6 core Intel CPU at high clocks is fast enough for 99% of people out there.
No need for 10,12,or 16 cores ,it's a total waste of money/performance for 98% of us.
High frequency/IPC is where its at. Might need 8 cores in 2021.
 
  • Like
Reactions: CHADBOGA

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,784
136
High frequency/IPC is where its at. Might need 8 cores in 2021.

Console users are getting 8c chips next year. Might even have SMT. Those are gonna be some beefy cores, too. Not like the little Jaguar cores they had before, two of which had to be tied up by OS/system tasks.
 

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
Console users are getting 8c chips next year. Might even have SMT. Those are gonna be some beefy cores, too. Not like the little Jaguar cores they had before, two of which had to be tied up by OS/system tasks.

While true even at 1440p resolution the bottleneck already is stongly on the GPU side like the recent i7 2600k article here on AT showed. At 4k, the CPU hardly matters the 2600k just as good as a 9900k at least in single-player.
 

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
The underdog in this case is the end user looking for a solid deal, not any of the megacorps themselves. It's why I can't rightly understand people blindly supporting Intel/AMD/Nvidia/Apple/Samsung/etc.

For those of who have been around since before the XT/AT days, we've seen all number of hype, success, and disappointment. Patterns start to emerge for sure. It's always exciting when there's a Radeon 9700, a GTX8800, a Conroe, or an Opteron/AMD64 moment in the industry. But for every one of those, there are at least as many underwhelming products, and between the GOATs and the duds, the vast majority of it is just an assortment of products that move the needle here or there, but whose primary purpose is to add to the coffers of the megacorps at the prices that the market will bear.

Given all of that, looking behind the scenes, my realistic expectations are for Zen 3000 to be another solid Zen family stack. Nothing shockingly good, and not embarrassing either. CPU performance has been pretty stagnant for so damn long that AMD improving to be a mix of better, on-par and 'close enough' with Intel is a good development. So I think we will see improved IPC, some pop to the clocks speeds, but I also think we aren't going to see quite the ludicrous offerings 'leaked' by the YouTube man either. Primarily it just doesn't make sense from a market standpoint. If you have something literally better in every way with double the cores and lower power consumption than the opposing flagship, why price it identical? Makes no sense, at best.

We've seen AMD with flagship offerings more than once, and when they command the top levels, those things are $1k. And, as much as we wouldn't like for the bargain hunters among us, you know the market would happily support something literally twice as good as a 9900k @ $1000. There's a reason why the 2080ti sells so well. If they were able to offer both superior IPC and hit 5+GHz, then at best they might slot in $449 8C/16T Zen2 there. Buyers would eat it up and have no real reason whatsoever to choose the 9900k in that case. And so on, undercutting Intel across the board with superior offerings.

The leaks OTOH seem like complete fanfiction, and they've been spread so far by that one flake that a ton of other outlets keep re-sourcing from each other in a weird online game of telephone. Mind you, this is a guy who told someone that disagreed with his opinion to kill themselves, who mass deletes his comments on Reddit, etc. Dude is a bit of a nutter butter, look it up.

I'll he hoping, but beneath a protective layer of cynicism :) I have a couple of waiting AM4 builds for some happy upgrades depending on how it falls.

My thoughts (posted in the Ryzen 3000 speculation thread) are that AMD does have something that matches or beats Intel current offerings, because Intel got caught on 14nm and Zen2 was targeted to compete with Intel 10nm. AMD knew what Zen2 needed probably before Zen had an alpha sample made.

Zen+ only "loses" right now because of memory latency (20ns slower?) and clock speed.

I am mystified that people do not realize these architectures are planned years in advance and simulated to the nth degree.

All the compromises made to get Zen out the door are going to be fixed in Zen2.

So, they will have something that matches the i9 9900K for cheaper and something faster that,will cost the same or more (~$500), and they will fly off the shelves.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
My thoughts (posted in the Ryzen 3000 speculation thread) are that AMD does have something that matches or beats Intel current offerings, because Intel got caught on 14nm and Zen2 was targeted to compete with Intel 10nm. AMD knew what Zen2 needed probably before Zen had an alpha sample made.

Zen+ only "loses" right now because of memory latency (20ns slower?) and clock speed.

I am mystified that people do not realize these architectures are planned years in advance and simulated to the nth degree.

All the compromises made to get Zen out the door are going to be fixed in Zen2.

So, they will have something that matches the i9 9900K for cheaper and something faster that,will cost the same or more (~$500), and they will fly off the shelves.

I'm not sure you completely understand me here :)

I would not be surprised in any way if AMD is able to notch ahead here. The catch is :

Total IPC win. Ok, possible.
Total Clock Speed equal or win. Again, possible.
16C/32T variant on AMD. Again, possible.

Where it falls apart is IF all three of those come together, choosing to sell a 16C/32T capable of ~5Ghz @ $499. Just on pure common sense, it doesn't mesh with reality. Because that would mean that a MUCH smaller 8C/16T would be already capable of being far superior to the 9990k. Thus, if these 'leaks' were correct in all manner, then the pricing would perhaps undercut Intel at every level, but not to that degree. Extreme buyers (the 2080ti/Titan types) would happily eat up 16C/32T 5Ghz @ $999, or maybe cut them at a break and go $899.

Similarly, Intel has a ton of 2C and 4C items that sell widely to OEMs. Ryzen 2 '3' level products being 6C/12T to start seems like a waste, when a 2C/4T and 4C/8T offering @ 7nm with these types of gains would be massively effective at cutting Intel out of every price point yet maximizing sales. Now the argument against this may simply be timing, if 7nm capacity can't be sourced to run that many types of Zen 7nm yet, so they're focusing on larger chiplets to start, with smaller monodies to follow? Nearly impossible to say until we get more confirmation on the lineup.

Finally, how often have we seen gigantic leaps with a revision? Intel went from 32nm to 14nm on Core (Sandy to Coffee), and we didn't see that much movement in raw clock speed potential. Now that can be due to balancing pipelines for IPC and all other types of considerations, but the range between ~4 to 5+ seems particularly sticky. One could observe that the iGPU portion of these Intel gens kept growing in space, thus using up the potential improvements in core count/etc. Which I totally agree with in so many ways. The iGPU on Intel CPUs is pretty darned big in die space. But it doesn't seem to count much in relation to the cores that actually ARE on the SKUs. On the flip side, we've seen Vega 64 to Vega 7, 14nm to 7nm, nearly identical cores, very modest clock boost, yet it actually increased in power consumption. The performance gains seemed more closely tuned to the improved HBM and memory performance overall. And, how was such a thing priced? Right around its competition, more or less.

So you see, if the list had leaked with no mention of Ryzen 3 6C/12T, and no prices listed, I'd be much more trusting towards it. Applying business-sense logic to it with those things in consideration however, makes me look at it as probably fanfic.

As for planning things and simulating things, sure. But between the drawing board and final silicon, a lot happens. Sometimes entirely unforeseen situations arise, or things don't pan out as expected. Intel's own late 10nm is one example among so many. I doubt ATI was planning on the 2900XT being kinda bad. I doubt Nvidia planned on Geforce FX being a laughable meme. I doubt Intel thought Larrabee and Itanium were roads to nowhere. I know AMD didn't expect Phenom I to be basically uncompetitive AND late. I know Intel once expected P4/Netburst to continue to scale with clockspeed and die shrinks, but ran into disaster with Prescott and had to radically alter their roadmaps. Etc.

Much more common than these disasters are moderate improvements, more so now that process tech is slowing down in regards to silicon. "7nm" isn't even true "7nm" in the old ways of measurement either. It's mostly PR speak, even from Intel side of things 14nm is not uniformly 14nm, and there are a load of other factors that determine how well an IC is going to fare on a node.

Idk. I hope its great, and another legendary product. But I won't be surprised if it's just 'pretty good'.
 
  • Like
Reactions: Tlh97 and Pilum

TheELF

Diamond Member
Dec 22, 2012
3,967
720
126
Zen+ only "loses" right now because of memory latency (20ns slower?) and clock speed.
And I'm not a millionaire only because I don't have a million dollars.
It's extremely difficult to get a high performing CPU to clock that high and it's just as difficult to interconnect a lot of cores with that low a latency.
I agree with arkaign,if and that's a huge if, they manage to surpass intel they are going to ask for ridiculous prices,just as they did with all their halo products.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,784
136
All the compromises made to get Zen out the door are going to be fixed in Zen2.

Don't be so sure about that. Intercore latency is going to be a possible issue thanks to the chiplet design. At least on 16c chips anyway.

if and that's a huge if, they manage to surpass intel they are going to ask for ridiculous prices,just as they did with all their halo products.

Hector Ruiz isn't running the company anymore. Lisa Su has either directly or indirectly provided up to two major advantages over AMD's position in 2004-2005: she has increased CPU supply, and she has reduced desktop prices for AMD's product. AMD may be on the cusp of regaining the performance lead in nearly every possible segment - something they haven't accomplished since the x2 days. The 1800x came close (it was competitive with both the 7700k and 6800k), but it will take Zen2 to get the job done. And Intel will have few answers until 2021. So AMD can pull a Hector and try to increase margins, which leaves Intel an opening to counterattack at a comfortable margin position (something Intel's investors love) with 7nm products. Or they can repeat what they did in 2017, keep prices capped at the $500 and below range, and soak up market share as Intel flounders.

Hector Ruiz honestly believed that Intel would be stuck in Netburst hell forever (they wouldn't). I think Su is smarter than that. Intel will be back. AMD can't afford to charge $1k for a top-end Zen3 sku when competing with next-gen Intel CPUs.
 

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
I'm not sure you completely understand me here :)

I would not be surprised in any way if AMD is able to notch ahead here. The catch is :

Total IPC win. Ok, possible.
Total Clock Speed equal or win. Again, possible.
16C/32T variant on AMD. Again, possible.

Where it falls apart is IF all three of those come together, choosing to sell a 16C/32T capable of ~5Ghz @ $499. Just on pure common sense, it doesn't mesh with reality. Because that would mean that a MUCH smaller 8C/16T would be already capable of being far superior to the 9990k. Thus, if these 'leaks' were correct in all manner, then the pricing would perhaps undercut Intel at every level, but not to that degree. Extreme buyers (the 2080ti/Titan types) would happily eat up 16C/32T 5Ghz @ $999, or maybe cut them at a break and go $899.

Similarly, Intel has a ton of 2C and 4C items that sell widely to OEMs. Ryzen 2 '3' level products being 6C/12T to start seems like a waste, when a 2C/4T and 4C/8T offering @ 7nm with these types of gains would be massively effective at cutting Intel out of every price point yet maximizing sales. Now the argument against this may simply be timing, if 7nm capacity can't be sourced to run that many types of Zen 7nm yet, so they're focusing on larger chiplets to start, with smaller monodies to follow? Nearly impossible to say until we get more confirmation on the lineup.

Finally, how often have we seen gigantic leaps with a revision? Intel went from 32nm to 14nm on Core (Sandy to Coffee), and we didn't see that much movement in raw clock speed potential. Now that can be due to balancing pipelines for IPC and all other types of considerations, but the range between ~4 to 5+ seems particularly sticky. One could observe that the iGPU portion of these Intel gens kept growing in space, thus using up the potential improvements in core count/etc. Which I totally agree with in so many ways. The iGPU on Intel CPUs is pretty darned big in die space. But it doesn't seem to count much in relation to the cores that actually ARE on the SKUs. On the flip side, we've seen Vega 64 to Vega 7, 14nm to 7nm, nearly identical cores, very modest clock boost, yet it actually increased in power consumption. The performance gains seemed more closely tuned to the improved HBM and memory performance overall. And, how was such a thing priced? Right around its competition, more or less.

So you see, if the list had leaked with no mention of Ryzen 3 6C/12T, and no prices listed, I'd be much more trusting towards it. Applying business-sense logic to it with those things in consideration however, makes me look at it as probably fanfic.

As for planning things and simulating things, sure. But between the drawing board and final silicon, a lot happens. Sometimes entirely unforeseen situations arise, or things don't pan out as expected. Intel's own late 10nm is one example among so many. I doubt ATI was planning on the 2900XT being kinda bad. I doubt Nvidia planned on Geforce FX being a laughable meme. I doubt Intel thought Larrabee and Itanium were roads to nowhere. I know AMD didn't expect Phenom I to be basically uncompetitive AND late. I know Intel once expected P4/Netburst to continue to scale with clockspeed and die shrinks, but ran into disaster with Prescott and had to radically alter their roadmaps. Etc.

Much more common than these disasters are moderate improvements, more so now that process tech is slowing down in regards to silicon. "7nm" isn't even true "7nm" in the old ways of measurement either. It's mostly PR speak, even from Intel side of things 14nm is not uniformly 14nm, and there are a load of other factors that determine how well an IC is going to fare on a node.

Idk. I hope its great, and another legendary product. But I won't be surprised if it's just 'pretty good'.

The i9 9900K "beater" would be a 12c / 24t CPU ~$500 - $600, anything beyond that I think would be a different tier of product (Threadripper).
 

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
Don't be so sure about that. Intercore latency is going to be a possible issue thanks to the chiplet design. At least on 16c chips anyway.



Hector Ruiz isn't running the company anymore. Lisa Su has either directly or indirectly provided up to two major advantages over AMD's position in 2004-2005: she has increased CPU supply, and she has reduced desktop prices for AMD's product. AMD may be on the cusp of regaining the performance lead in nearly every possible segment - something they haven't accomplished since the x2 days. The 1800x came close (it was competitive with both the 7700k and 6800k), but it will take Zen2 to get the job done. And Intel will have few answers until 2021. So AMD can pull a Hector and try to increase margins, which leaves Intel an opening to counterattack at a comfortable margin position (something Intel's investors love) with 7nm products. Or they can repeat what they did in 2017, keep prices capped at the $500 and below range, and soak up market share as Intel flounders.

Hector Ruiz honestly believed that Intel would be stuck in Netburst hell forever (they wouldn't). I think Su is smarter than that. Intel will be back. AMD can't afford to charge $1k for a top-end Zen3 sku when competing with next-gen Intel CPUs.

On latency: IF got upgraded to IF 2, so, I am guessing that fixes a lot of latency issues.

I agree with the rest.

Dr. Su started a core war, it's not ending anytime soon.
 
  • Like
Reactions: Tlh97

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
And I'm not a millionaire only because I don't have a million dollars.
It's extremely difficult to get a high performing CPU to clock that high and it's just as difficult to interconnect a lot of cores with that low a latency.
I agree with arkaign,if and that's a huge if, they manage to surpass intel they are going to ask for ridiculous prices,just as they did with all their halo products.

I disagree on the asking a lot of $$$ for something better.

My "surpassing" the Intel i9 9900K is 12c / 24t and probably ~10% better performance across all benchmarks against it for $500 - $600 or heck, maybe $450.

This whole architecture was designed to be cheap to make so it could be sold cheap.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
I disagree on the asking a lot of $$$ for something better.

My "surpassing" the Intel i9 9900K is 12c / 24t and probably ~10% better performance across all benchmarks against it for $500 - $600 or heck, maybe $450.

This whole architecture was designed to be cheap to make so it could be sold cheap.

The problem with this is the claims of :

5ghz+
Superior IPC
16C/32T $499

These factors would mean an 8C/16T Zen2 would easily be better than a 9900k, a 12C or 16C would be in another league entirely (50%+ and 100%+ faster respectively in scalable loads).
 
  • Like
Reactions: Tlh97

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
The problem with this is the claims of :

5ghz+
Superior IPC
16C/32T $499

These factors would mean an 8C/16T Zen2 would easily be better than a 9900k, a 12C or 16C would be in another league entirely (50%+ and 100%+ faster respectively in scalable loads).

I think we saw the part that matches i9 9900K in January at CES.

AND, if 9900K (stock) runs all core at 4.7, I think a Zen2 something will be running at that or a little slower, and match it in perf.

Zen2 needs what, 5 - 10 % better perf to match the 9900K?

Zen was released on a mobile first cpu tech, Zen2 is using a high perf option of 7nm.

I really don't see how Zen2 7nm matching and slightly beating Intels best 14nm part is so far fetched. Again, this has been planned and tested for for years, they didn't just whip up Zen2 after Zen was launched to the public.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
I think we saw the part that matches i9 9900K in January at CES.

AND, if 9900K (stock) runs all core at 4.7, I think a Zen2 something will be running at that or a little slower, and match it in perf.

Zen2 needs what, 5 - 10 % better perf to match the 9900K?

Zen was released on a mobile first cpu tech, Zen2 is using a high perf option of 7nm.

I really don't see how Zen2 7nm matching and slightly beating Intels best 14nm part is so far fetched. Again, this has been planned and tested for for years, they didn't just whip up Zen2 after Zen was launched to the public.

It's not the matching or perhaps beating 9900k that I really find doubtful. It's the list of SKUs.

Per core superior IPC + 5Ghz~ capability would mean 8C/16T is > 9900k.

Yet the list shows Ryzen 5 3600x 65w, unlocked 8C/16T for $250. Literally half the price for a superior CPU.

In fact the entire lineup just looks so suspect as listed. The 3300x is 6C/12T, but only 3.5/4.3? Why? 3700x 12C/24T 4.2/5.0, but 3850x 16C/32T is 4.3/5.1? Why? Why would the clocks keep going UP for higher core count units? That would be the opposite of normal tiering. They could easily offer more tiers in each stack from 4 to 5Ghz to stretch yields and make more money, and offer more performance. Holding ~5Ghz stock purely for 12/16C products is .. bizarre.

The devil is in the details, and an unlocked 9900k killer being offered at $250 is .. well let's call it what it is : kind of silly sounding, especially when you consider the leaker's history of being totally off with AMD news. He said Navi was going to be unveiled in full at CES, made zero mention of Vega 7, and on and on. Yet his frankly strange claims have kept bouncing around constantly. Then the engineer guy on Twitter gives the 3.3/4.2 info on a Zen2 CPU and he basically freaks out.

Again, for the umpteenth time : plans do not always equal reality, all the more so when you don't even own your own fabs. You can estimate all you want in advance, but things can go way better or way worse than expected.
 
  • Like
Reactions: Tlh97

TheELF

Diamond Member
Dec 22, 2012
3,967
720
126
I think we saw the part that matches i9 9900K in January at CES.

AND, if 9900K (stock) runs all core at 4.7, I think a Zen2 something will be running at that or a little slower, and match it in perf.

Zen2 needs what, 5 - 10 % better perf to match the 9900K?

Zen was released on a mobile first cpu tech, Zen2 is using a high perf option of 7nm.

I really don't see how Zen2 7nm matching and slightly beating Intels best 14nm part is so far fetched. Again, this has been planned and tested for for years, they didn't just whip up Zen2 after Zen was launched to the public.
If cinebench alone were a metric for overall performance then FX would have already killed off intel...
Yes Zen2 needs about 5 - 10 % better perf to match the 9900K in cinebench...15...
version 20 came out a couple of months ago and changed things.
You can't take one single result and use it as a blanket performance number,you have to bench a lot of apps and find an average.
If you take one single result then why take the one with the smallest difference?Why not the one with the biggest?
Then it would be 85% ...
https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/5
102013.png
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
Hell, they JUST changed their roadmap again during the recent quarterly report and investor calls. As is done frequently in the industry. This isn't like making sedans or furniture, there are simply too many variables to nail it down perfectly, and you can't be sure of what you have until it gets through the spins and respins and samples and etc.

I mean we have boundless examples through the years of missed targets, despite all efforts to the contrary, from AMD, from Intel, from Nvidia, and so on.
 

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
It's not the matching or perhaps beating 9900k that I really find doubtful. It's the list of SKUs.

Per core superior IPC + 5Ghz~ capability would mean 8C/16T is > 9900k.

Yet the list shows Ryzen 5 3600x 65w, unlocked 8C/16T for $250. Literally half the price for a superior CPU.

In fact the entire lineup just looks so suspect as listed. The 3300x is 6C/12T, but only 3.5/4.3? Why? 3700x 12C/24T 4.2/5.0, but 3850x 16C/32T is 4.3/5.1? Why? Why would the clocks keep going UP for higher core count units? That would be the opposite of normal tiering. They could easily offer more tiers in each stack from 4 to 5Ghz to stretch yields and make more money, and offer more performance. Holding ~5Ghz stock purely for 12/16C products is .. bizarre.

The devil is in the details, and an unlocked 9900k killer being offered at $250 is .. well let's call it what it is : kind of silly sounding, especially when you consider the leaker's history of being totally off with AMD news. He said Navi was going to be unveiled in full at CES, made zero mention of Vega 7, and on and on. Yet his frankly strange claims have kept bouncing around constantly. Then the engineer guy on Twitter gives the 3.3/4.2 info on a Zen2 CPU and he basically freaks out.

Again, for the umpteenth time : plans do not always equal reality, all the more so when you don't even own your own fabs. You can estimate all you want in advance, but things can go way better or way worse than expected.

I agree the leak is kinda ridiculous, but, I still think we are going to be surprised with the Zen2 offerings and will channel the Fry "Shut up and take my money" meme.
 

B-Riz

Golden Member
Feb 15, 2011
1,482
612
136
If cinebench alone were a metric for overall performance then FX would have already killed off intel...
Yes Zen2 needs about 5 - 10 % better perf to match the 9900K in cinebench...15...
version 20 came out a couple of months ago and changed things.
You can't take one single result and use it as a blanket performance number,you have to bench a lot of apps and find an average.
If you take one single result then why take the one with the smallest difference?Why not the one with the biggest?
Then it would be 85% ...
https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/5
102013.png

I'm guessing they have something that is 5% -10% faster across a benchmark suite than Intel's i9 9900K. And it's not going to be cheap.
 

moinmoin

Diamond Member
Jun 1, 2017
4,933
7,618
136
Literally half the price for a superior CPU.
Did you miss the original Ryzen launch? AMD was offering double the cores for the same price point of Intel CPUs across the whole product range. It's not unthinkable they'll do something similar again when the can.

They want to eat into Intel's market share big time while the CPU design is optimized for high yield (= low production cost at scale) from the start. They don't need high prices to make big short term profit, they need higher market shares first to achieve sustainable profit. The fact Intel is far from crashing under the price pressure so far (thanks to many sales channels being slow to change) shows exactly why sustainable profit is preferable.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
Zen 1xxx wasn't an IPC winner that clocked highly, or was all that friendly with memory to begin with. Depending on app, it could lose against i5s. It's real excellence was in heavy MT stuff.

It also premiered at loftier prices than many remember, only to come sharply down later. It became a situation where the price was hard to ignore as a serious value. Even if Intel was faster in certain areas, you could build a better Ryzen system most of the time with the same budget.

But a Zen2 where it wins literally everything (superior IPC + 5Ghz range + moaaaar cores), I think they would price that at a level the market would support instead of roughly half price across the board.
 
  • Like
Reactions: Pilum

scannall

Golden Member
Jan 1, 2012
1,944
1,638
136
The i9 9900K "beater" would be a 12c / 24t CPU ~$500 - $600, anything beyond that I think would be a different tier of product (Threadripper).
I'm going to disagree a bit here. We've already seen an 8 core engineering sample, running at an unknown clock edging out a 9900K. It will be 8 cores head to head for a 9900K beater. Adding a 2nd chiplet for 12 and 16 core variants adds heat, which will have them clock lower than a single 8 core chiplet. My guess would be the 8 core version will be the best bet for gamers. 12 and 16 core versions for people who use more heavily threaded types of applications.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,478
14,432
136
I'm going to disagree a bit here. We've already seen an 8 core engineering sample, running at an unknown clock edging out a 9900K. It will be 8 cores head to head for a 9900K beater. Adding a 2nd chiplet for 12 and 16 core variants adds heat, which will have them clock lower than a single 8 core chiplet. My guess would be the 8 core version will be the best bet for gamers. 12 and 16 core versions for people who use more heavily threaded types of applications.
I agree. The thing is, I would love 16 core AM4 chips, but that makes all my threadrippers not as good. I hope they expand the TR4 socket as well, less expensive 24 and 32 core or more, and at less price and heat.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
I agree. The thing is, I would love 16 core AM4 chips, but that makes all my threadrippers not as good. I hope they expand the TR4 socket as well, less expensive 24 and 32 core or more, and at less price and heat.

All of this makes sense to me. I even think an eventual 16C/32T AM4 part is possible, but at lower clocks, while the TR family is expanded higher. But I don't expect them out of the gate. Seems like AMD has pushed the roadmap around a bit during the last QR.