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

Page 12 - 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 .

sxr7171

Diamond Member
Jun 21, 2002
5,079
40
91
I wonder what the next gen console cpu will look like.

What kind of specialized memory cache might they use to mitigate that memory latency? Consoles tend to have some specialized memory in them. Can that translate to a desktop system? Can they end up like “super console” chips? Games would be optimized around it since it is common on both consoles plus on PC - user willing.
 
Last edited:

killster1

Banned
Mar 15, 2007
6,208
475
126
Maybe there should be a Las Vegas betting pool on this. I'd love amd to win but when was that last happening operon 165? 9900k beats zen 95+% and Intel is hardly even trying.
 

Thunder 57

Platinum Member
Aug 19, 2007
2,647
3,706
136
Maybe there should be a Las Vegas betting pool on this. I'd love amd to win but when was that last happening operon 165? 9900k beats zen 95+% and Intel is hardly even trying.

So doubling their core count and disregarding power use and heat to hit 5GHz since Zen has launched is "hardly even trying"?

As to the original question, there are reasons to think the 9900k will be the best Intel has to offer for some time. 10nm on the high end is nowhere in sight. Will they make Sunny Cove on 14nm? Not according to the latest leaked roadmap, which also shows 14nm all they way through 2020. When we do see 10nm, will it clock as high as a highly mature 14nm? Not at first, I don't think.

Intel also will have to give up its ring bus for the mesh at some point as core counts rise. Look at the power usage of the ring vs the mesh:

Uncore%208700Ka.png


Uncore%207980XE.png

You're looking at about 8W vs 39W at 12 threads when you move to the mesh. That power will have to come out of the core power budget, meaning again likely lower clock speeds. AMD is already dealing with high IF power, but can get away with it because the Zen cores aren't as power hungry.

Finally, though this one is hard to say since data is scare, Intel may lose some performance in games when they increase the L2 cache size and change the L3 to a victim cache. The increased L2 offsets this in some games but others see performance plummet. Since there is no equal core count equal clock speed chips with both designs, this harder to say. Play around a bit in Bench comparing a 9900k and Skylake-x. Skylake-x is clocked lower, but not enough to explain some of the discrepancies.

Put it all together and I think the 9900k will be the best performing in games for some time. Can AMD match or surpass it with Zen 2? That I can't say. I think Zen 2 wins more of the non gaming benches and the 9900k wins more games than it loses.
 
Last edited:
  • Like
Reactions: Tlh97

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
Skylake-x is clocked lower, but not enough to explain some of the discrepancies.

As far as I remember overclockign the very low clocked uncore in skylake-x resolved the slower gaming performance. So the mesh mostly has a power issue but not a performance issue.
 
  • Like
Reactions: ZipSpeed

Thunder 57

Platinum Member
Aug 19, 2007
2,647
3,706
136
As far as I remember overclockign the very low clocked uncore in skylake-x resolved the slower gaming performance. So the mesh mostly has a power issue but not a performance issue.

I don't think the mesh is a performance issue. If anything as the number of cores grows, the mesh should be a net gain over the ring bus. I never really thought about the low uncore clock but suspect honestly. I know the mesh always got a bad rap, but I always figured the performance characteristics were more because of the complete revamping of the L2/L3.
 

TheELF

Diamond Member
Dec 22, 2012
3,967
720
126
So doubling their core count and disregarding power use and heat to hit 5GHz since Zen has launched is "hardly even trying"?
For PURE GAMING the 2700x competes (as in looses) against the 7700k so yeah the 9900k isn't even trying.
Intel just smelled a cash grab and went with it,hey if they can sell their aging 14nm for a premium why shouldn't they.
https://www.techspot.com/review/1613-amd-ryzen-2700x-2600x/page3.html
https://www.gamersnexus.net/guides/3423-intel-i7-7700k-revisit-benchmark-vs-9700k-2700-9900k
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
as the thread is saying 9900K last hurrah in gaming, 9900K is a pretty hefty performing last hurrah
everything else will be trying to catch up in next year

IMO SKL-X (mesh, cascade lake etc) overclocking we have seen helps a lot but it doesn't even challenge the ring bus
if intel discontinues the ring bus they need to have something prepared- they cannot compete with just core count or AVX 512 non existent on desktop (ATM) advantage, not talking about price
 

killster1

Banned
Mar 15, 2007
6,208
475
126
Exactly they are using old tech and still winning that's why gb I say they are not trying. They had no reason to try (new fab) since they are number one.. it would be like honda dumping money into tech for motors when they already have better gas mileage and power then the competition right? It was trying yes but barely as it took 0$ for them compared to building a new factory
So doubling their core count and disregarding power use and heat to hit 5GHz since Zen has launched is "hardly even trying"?

As to the original question, there are reasons to think the 9900k will be the best Intel has to offer for some time. 10nm on the high end is nowhere in sight. Will they make Sunny Cove on 14nm? Not according to the latest leaked roadmap, which also shows 14nm all they way through 2020. When we do see 10nm, will it clock as high as a highly mature 14nm? Not at first, I don't think.

Intel also will have to give up its ring bus for the mesh at some point as core counts rise. Look at the power usage of the ring vs the mesh:



You're looking at about 8W vs 39W at 12 threads when you move to the mesh. That power will have to come out of the core power budget, meaning again likely lower clock speeds. AMD is already dealing with high IF power, but can get away with it because the Zen cores aren't as power hungry.

Finally, though this one is hard to say since data is scare, Intel may lose some performance in games when they increase the L2 cache size and change the L3 to a victim cache. The increased L2 offsets this in some games but others see performance plummet. Since there is no equal core count equal clock speed chips with both designs, this harder to say. Play around a bit in Bench comparing a 9900k and Skylake-x. Skylake-x is clocked lower, but not enough to explain some of the discrepancies.

Put it all together and I think the 9900k will be the best performing in games for some time. Can AMD match or surpass it with Zen 2? That I can't say. I think Zen 2 wins more of the non gaming benches and the 9900k wins more games than it loses.


And yes of course Intel won't be releasing the 19990k as their first chip release they will milk it and release that after all the slower models (frustrating but that's how it aways is) so I guess it all depends on this mythical amd chip for how fast intel needs to be)
 
Last edited:

exquisitechar

Senior member
Apr 18, 2017
655
862
136
Surprising that more people than not think that AMD will have the best gaming CPU with Zen 2. IMO, it's almost certainly not happening.
Exactly they are using old tech and still winning that's why gb I say they are not trying. They had no reason to try (new fab) since they are number one.. it would be like honda dumping money into tech for motors when they already have better gas mileage and power then the competition right? It was trying yes but barely as it took 0$ for them compared to building a new factory



And yes of course Intel won't be releasing the 19990k as their first chip release they will milk it and release that after all the slower models (frustrating but that's how it aways is) so I guess it all depends on this mythical amd chip for how fast intel needs to be)
This is ridiculous, to think that Intel is willingly in their 14nm hell. Intel did build new fabs for 10nm and put a large amount of money in it, it's just been a disaster and things have been looking better only recently.
 

TheELF

Diamond Member
Dec 22, 2012
3,967
720
126
This is ridiculous, to think that Intel is willingly in their 14nm hell. Intel did build new fabs for 10nm and put a large amount of money in it, it's just been a disaster and things have been looking better only recently.
Sure they have trouble in making 10nm but it's not hurting them as much as people try to make it look,intel had growth in the desktop market in q1 2019,growth...with 14nm.

Surprising that more people than not think that AMD will have the best gaming CPU with Zen 2. IMO, it's almost certainly not happening.
At most what could happen is more parity in more games as is already happening in so many games because they all push graphics and only use the CPU to feed the GPU and today's GPUs strong as they are still have very real limits.
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
my personal opinion is that Intel has more engineering a leading teams and the 14nm production engineering team is much more capable than their 10nm
14nm results are surprisingly good for 14nm but 10nm team are bad enough not to be better than 14nm after 3 years....
 
  • Like
Reactions: Tlh97

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
my personal opinion is that Intel has more engineering a leading teams and the 14nm production engineering team is much more capable than their 10nm
14nm results are surprisingly good for 14nm but 10nm team are bad enough not to be better than 14nm after 3 years....
My money goes on upper management, not engineering teams. They had the biggest warning sign with 14nm delays, Broadwell barely made it though in time to be called a product. You can only hold so many keynote presentations claiming a 3 year advantage in process node tech before you start believing it yourself.

In fact no, I blame the 14nm team - they did not fail bad enough to wake up management. :p
 

mopardude87

Diamond Member
Oct 22, 2018
3,348
1,575
96
I would say that the 9700k should be mentioned as the chip to get, not the 9900k. I do suspect that AMD will take the lead though, either with Zen 2 or soon after.

If what i hear about the next gen Playstation having 16 threads is true, i got a bad feeling about anything with 8 threads. Trust me i debated between my i7 8700 non k and that 9700k for days but ultimately decided more threads would work best for me. The i5 8400 before it already felt a bit weak in BF1 given some maps can dang near max that thing out and i felt a hesitation as well. Game feels quite a bit smoother on the i7 8700 and way more of a proper upgrade over the i5 4460 i had before the coffee lake stuff.

One buddy told me i made a mistake with the 8700 given what the next generation consoles will be running. My money is on 12 threads with a very likely 2x the core clock over 16 very slow cores. Pretty sure a 9900k upgrade will be cheaper then a move to a Ryzen 3000 platform and way more practical too assuming 12 threads just isn't cut it. Then again i am making the move to 4k today so its gonna be a minute before a gpu even ends up being a bottleneck at such a resolution.
 
  • Like
Reactions: Tlh97 and krumme

rbk123

Senior member
Aug 22, 2006
743
345
136
All the speculation points to Zen2 eeking out a win over the 9900K at best, mostly due to a high probability of not hitting, or getting close enough to, 5Ghz. If this proves to be true, then Intel will be able to add refinements to their refinements and probably eek back on top. It would take a solid gap between Zen2 over Intel for it to be the "last hurrah" in gaming for Intel. While this would be nice, it's not likely.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
-> Intel is skipping over 10nm straight to 7nm for 2020 products.
-> 7nm is HVM at Intel and the first chip is the Next Core; 128 KB L1i and 64 KB L1d.

Comet Lake(10c) might be getting cut down to low-volume, for a new platform 14nm SKU & 7nm SKU on a new desktop socket. This should be noticed later this year (2H2019), but I can't really extrapolate timelines from Intel profiles well yet.
 
Last edited:

exquisitechar

Senior member
Apr 18, 2017
655
862
136
-> Intel is skipping over 10nm straight to 7nm for 2020 products.
-> 7nm is HVM at Intel and the first chip is the Next Core; 128 KB L1i and 64 KB L1d.

Comet Lake(10c) might be getting cut down to low-volume, for a new platform 14nm SKU & 7nm SKU on a new desktop socket. This should be noticed later this year (2H2019), but I can't really extrapolate timelines from Intel profiles well yet.
We've just learned 7nm is in 2021, though?
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
We've just learned 7nm is in 2021, though?
My non-exact more detailed extrapolation;
Announce in December 2019 for the 14nm SKU and December 2020 for the 7nm SKU.

7nm was pathfinding finished by 2017, 2017-2018-2019-2020 is four years needed to get to production.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
10nm Server in 1H 2020 seems like a pipe dream. None of their roadmaps show anything but small, limited releases of small-die 10nm parts in late 2019. I am not sure how they are going to launch relatively-large Xeon dice on that process. Still no word of when 10nm comes to the desktop . . . if ever.

7nm in 2021 seems plausible, so I won't count that out.
 

nenforcer

Golden Member
Aug 26, 2008
1,767
1
76
One buddy told me i made a mistake with the 8700 given what the next generation consoles will be running. My money is on 12 threads with a very likely 2x the core clock over 16 very slow cores. Pretty sure a 9900k upgrade will be cheaper then a move to a Ryzen 3000 platform and way more practical too assuming 12 threads just isn't cut it. Then again i am making the move to 4k today so its gonna be a minute before a gpu even ends up being a bottleneck at such a resolution.

No your perfectly fine with the i7 8700 at the moment since even the XBOX One devotes a core to the OS and a core to VR / Kinect limiting games to 6 cores. They removed the VR / Kinect core and let games use a 7th core if necessary I believe. The next consoles wont even be out intul late 2020 and by that time AMD will have Ryzen 4000 series on .7nm+ and Intel will have at least 10 core if not 12 or 14 core desktop processors. A used Core i7 9700 or Core i9 9900K will be substantially cheaper at that point, possibly less than $300, so you will have an easy upgrade path to a 16 thread processor. Obviously Intel is pretty confident their 10 core / 20 thread .14nm Comet Lake processor is going to be competitive with AMD Ryzen 3000 series 16 core / 32 thread processors for 2019. (at least for single threaded performance)
 
  • Like
Reactions: Tlh97

ondma

Platinum Member
Mar 18, 2018
2,718
1,278
136
Surprising that more people than not think that AMD will have the best gaming CPU with Zen 2. IMO, it's almost certainly not happening.

This is ridiculous, to think that Intel is willingly in their 14nm hell. Intel did build new fabs for 10nm and put a large amount of money in it, it's just been a disaster and things have been looking better only recently.
its not surprising at all in these forums.
 
  • Like
Reactions: Tlh97 and Pilum

sxr7171

Diamond Member
Jun 21, 2002
5,079
40
91
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?