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

[Vattila]

Independent test results for Intel's 8-core i9-9900K are just around the corner, and the general consensus and expectations are that it will put Intel further in the lead with the "world's best gaming processor".

However, like most technology analysts and avid followers of x86 CPU architecture, I expect 7nm Zen 2 to bring substantial improvements next year, as it will have architectural refinements and — for the first time ever — be manufactured on a process node on par or even ahead of Intel.

In fact, many industry analysts now consider Intel's manufacturing lead permanently lost, and that TSMC will compete at the forefront, if not be the absolute leader, of semiconductor manufacture from here onwards.

Meanwhile, it is expected that the next-generation consoles will continue to use AMD semi-custom chips, and that those chips will make use of the Zen CPU architecture. So game developers will increasingly optimise for this architecture, likely an 8-core 2-CCX configuration, similar to what AMD offers in the mainstream.

Already, we are seeing parity in performance between Ryzen 2700X and i7-8700K in the most recent titles based on modern game engines optimised for the Zen architecture.

So based on all these developments, I expect Intel to lose its lead in gaming next year, and it may not regain the lead for the foreseeable future, as Zen 3 and successors, combined with AMD's console dominance and TSMC's manufacturing lead, all keep the pressure up on Intel.

What do you all think will happen, say, over the next three years?

 

[LTC8K6]

If I could predict the future...I'd be rich.

On the surface, current parity with an Intel 6 core chip wouldn't seem to say much about AMD 8 core chips in the future.
It also seems like a bit of an indication that maybe we don't need all these cores yet after all.

I suspect that Intel isn't as bad off as people like to think.
They've certainly had their troubles and AMD has been making nice strides.

A big clue will be when we get some more data on those Intel 10nm chips that are starting to slowly appear here and there.
We can get an idea of possible architecture improvements that Intel may be implementing.

 

[William Gaatjes]

Last thing i read about cpus manufactured by Intel using the 10nm process, is that those cpus, are lower clocked, more power consuming than anything Intel has made on 14nm (first version, + or ++ or +++).

 

[Vattila]

Last thing i read about cpus manufactured by Intel using the 10nm process, is that those cpus, are lower clocked, more power consuming than anything Intel has made on 14nm (first version, + or ++ or +++).

Yeah. That's my understanding as well. That's why I put "until Intel's 10nm matures in 2020" in the poll formulation. In other words, I think that Intel's first 10nm product in high-volume manufacture will not be a high-performance gaming CPU. The latter may come when yields mature, but may even have to wait for a further refinement of the process (10nm+).

The dark horse is Ice Lake and its schedule for introduction. It is unclear what architectural improvements it will bring. Whether it can beat Zen 2 is at this point unknown, as far as I know.

 

[lixlax]

The difference in gaming isn't huge in gaming between Zen+ and Coffee lake. To gain parity, let alone beat Intel in gaming will be a huge task and they will get their development going sooner or later. Lets put it this way- AMD is already close to ridge they're climbing, but the last part to reach it is extremely steep and requires lots of energy.

AMD needs another modest clock pump, more IPC and lower latencies which play important role in gaming perfomance.

 

[DigDog]

I may be *completely* wrong but my opinion is that Intel "struggles to keep up with the development stride of AMD" not because of incompetence but because - like EVERY large company - they have bureucracy problems; they are too big and too dumb to do things right.

Until AMD starts whoppin' they' ass, at which point they will pull another C2D miracle out of their ass. By which i mean "out of the R&D department".

TLDR AMD can put out a better cpu than Intel; they cant "beat" Intel.

 

[Hitman928]

I may be *completely* wrong but my opinion is that Intel "struggles to keep up with the development stride of AMD" not because of incompetence but because - like EVERY large company - they have bureucracy problems; they are too big and too dumb to do things right.

Until AMD starts whoppin' they' ass, at which point they will pull another C2D miracle out of their ass. By which i mean "out of the R&D department".

TLDR AMD can put out a better cpu than Intel; they cant "beat" Intel.

So. much. this.

Not saying their doomed or anything, just that intel's problems don't exist because of lack of talent or funding.

 

[Abwx]

If we are to believe the few hints we were given it s 10-15% better IPC and likely better latencies since Lisa Su stated that Zen 2 would be much better in gaming, so how much is "much better" when stated by an enginer who has forcibly the sense of measure..?.

Are 10-15% better gaming perf what we could call "much better"..?.

 

[maddie]

The difference in gaming isn't huge in gaming between Zen+ and Coffee lake. To gain parity, let alone beat Intel in gaming will be a huge task and they will get their development going sooner or later. Lets put it this way- AMD is already close to ridge they're climbing, but the last part to reach it is extremely steep and requires lots of energy.

AMD needs another modest clock pump, more IPC and lower latencies which play important role in gaming perfomance.

I may be *completely* wrong but my opinion is that Intel "struggles to keep up with the development stride of AMD" not because of incompetence but because - like EVERY large company - they have bureucracy problems; they are too big and too dumb to do things right.

Until AMD starts whoppin' they' ass, at which point they will pull another C2D miracle out of their ass. By which i mean "out of the R&D department".

TLDR AMD can put out a better cpu than Intel; they cant "beat" Intel.

Two opposing views that are remarkably popular.

1) CPU performance/clock is close to the peak.
2) Intel can pull a Conroe if the will is there.

For the first we appear to have Apple refuting this, and for the second, miracles on demand should be left to Deities.

 

[epsilon84]

I'm not so sure, and I'll try to explain why. Just to put things into perspective - it takes an overclocked 2700X @ 4.2GHz to draw level with an i5 8400 @ 3.8GHz:

Let's pause for a sec and consider that point. AMD's best CPU today, the 2700X, needs a 10% clockspeed advantage plus a massive thread advantage just to merely *match* a 6C/6T CFL chip at 3.8GHz in games.

The 9900K is literally 2 tiers above the 8400. 8C/16T vs 6C/6T. 4.7GHz vs 3.8GHz, plus the scope to overclock up to ~5.2GHz. 16MB L3 vs 9MB L3. Are we really expecting AMD to suddenly leapfrog that with a single uarch refresh?

AMD will surely close the gap with Zen 2, but to actually overtake Intel in gaming? I'm not so sure. Adding more cores won't do anything for gaming, so we are talking about AMD not only matching, but exceeding Intel IPC and Intel clockspeeds in order to overtake a 9900K.

I'm more confident about Zen 3 being able to beat a 9900K at gaming, but if I was a betting man, I'd put my money on the 9900K keeping the gaming crown against Zen 2. I'll be more than happy to be proven wrong on this though! Competition is always great and if AMD can actually beat Intel at its own strengths that will surely only drive Intel to improve and innovate more.

Looking beyond the 9900K though, a lot also depends on how high Intel's own 10nm or 10nm++ refresh will clock, as well as the IPC levels of Intel's next uarch - we've been at Skylake levels since 2015, don't forget. So Intel has had 4 years to develop something better, architecture wise. How much better exactly? Like LTCK86 said, I'd be rich if I could predict the future

 

[frozentundra123456]

I'm not so sure, and I'll try to explain why. Just to put things into perspective - it takes an overclocked 2700X @ 4.2GHz to draw level with an i5 8400 @ 3.8GHz:

Let's pause for a sec and consider that point. AMD's best CPU today, the 2700X, needs a 10% clockspeed advantage plus a massive thread advantage just to merely *match* a 6C/6T CFL chip at 3.8GHz in games.

The 9900K is literally 2 tiers above the 8400. 8C/16T vs 6C/6T. 4.7GHz vs 3.8GHz, plus the scope to overclock up to ~5.2GHz. 16MB L3 vs 9MB L3.

AMD will surely close the gap with Zen 2, but to actually overtake Intel in gaming? I'm not so sure. It would require AMD to bridge the IPC gap, plus the clockspeed gap. I'm more confident about Zen 3 being able to beat a 9900K at gaming, but if I was a betting man, I'd put my money on the 9900K keeping the gaming crown against Zen 2. I'll be more than happy to be proven wrong on this though! Competition is always great and if AMD can actually beat Intel at its own strengths that will surely only drive Intel to improve and innovate more.

Looking beyond the 9900K though, a lot also depends on how high Intel's own 10nm or 10nm++ refresh will clock, as well as the IPC levels of Intel's next uarch - we've been at Skylake levels since 2015, don't forget. So Intel has had 4 years to develop something better, architecture wise. How much better exactly? Like LTCK86 said, I'd be rich if I could predict the future

Yea, I dont buy the assumption that a lot on these forums are making that Zen 2 will automatically beat intel in gaming. They have both a considerable clockspeed deficit to make up and a higher latency with the CCX vs ring bus. So lets wait and see. Not saying it *cant* happen, but I am not making a lot of optimistic (for AMD) assumptions like some posters either. As far as "being optimized for Zen", we heard the same thing when the last round of gpus came out. Oh, buy AMD, they have the console contracts, their gpus are more future proof, same argument for DX12. Well we only have to look at how long AMD has been without a competitive top end card to see how that worked out.

 

[Charlie22911]

Last thing i read about cpus manufactured by Intel using the 10nm process, is that those cpus, are lower clocked, more power consuming than anything Intel has made on 14nm (first version, + or ++ or +++).

I hope that’s not the case for intels sake, because Broadwell (1st gen 14nm) was a turd in high performance incarnations.

My 6900k is at 4.2GHz all core and 1.275v (for AVX stability), under loads like handbrake it draws close to 250w at 75c-85c depending on ambient. 4.2 GHz all core is the highest I’ve even gotten any of the multiple Broadwell-E CPUs I’ve had to run stable. That’s with a full custom loop (EK Supremacy EVO, 6x120mm worth of radiator area).


This will be a great time for AMD to grab some market share and feed its R&D budget. But I also think it is important to keep expectations in check, I fully expect intel to maintain a small lead in performance across the board even when compared to 7nm zen 2. And that’s not a knock against AMD, on the contrary it’s impressive.

 

[DisarmedDespot]

I can't see Intel losing their lead to Zen 2. AMD will get damn close, but there's still a big enough gap I can't see it being closed in one generation.

I'm not sure what Intel can do right now. In some ways they're in a worse situation than the Core 2 Duo days. Their process lead might as well be gone, what with how many issues 10nm has been having, and the company doesn't seem to have any direction at all. Heck, their high margins might be a pair of golden handcuffs keeping them from competing on price with AMD as effectively as they could. It's in no way an existential threat (even if AMD pulled a rabbit out of their hat with Zen 2 and totally thwomped Intel across all segments, Intel wouldn't be in serious danger), but they're pretty clearly losing their groove and are guaranteed to lose more marketshare than they would otherwise unless they wake up.

 

[deathBOB]

I don’t think we can make conclusions about 7nm when we haven’t seen the high power process. You can’t assume a shrink will lead to higher clock speeds now.

 

[Borealis7]

both "yes" and "no" options make sense.
Yes - because it makes sense for Intel to want gamers to move on to the more expensive HEDT platform, but that could also make them lose that very same market segment to AMD with their cheaper offerings. Lets be frank here, as long as your CPU doesn't critically bottleneck the GPU, you could game even at 4K with a mainstream desktop chip.
and No - because Intel still want to sell mainstream desktop chips and now AMD has flipped a Zippo lighter under their behinds, so they have to keep churning out those 5% increments every year.

we'll have to wait and see.

 

[TheGiant]

IMO getting from 4, 4.2, 4,3 GHz etc and 4,6,8 cores to 5GHz while increasing IPC a task that not easy is (hello Yoda)
Everyone predicts AMD with 7nm will trounce Intel's offerings but the manufacturing node itself doesn't make the difference alone.
I will believe it when I see it

 

[ChiefBigFeather]

Zen is a workstation architecture that happens to be decent at gaming. While I suspect the gap will be very narrow, I don‘t think AMD will release a faster gaming chip. Zen just isn‘t build for doing that.

Frametime experience will probably be close to indistinguishable, especially with tweaked settings (good ram etc).

 

[TheELF]

Already, we are seeing parity in performance between Ryzen 2700X and i7-8700K in the most recent titles based on modern game engines optimised for the Zen architecture.

Some games just have internal timers or other limitations that make them just stop at some point,look at this all cores have some idle time GPU usage is low but FPS won't increase no matter what.
(AMD used this title to show off that Ryzen is "as fast as intel" before launch)
Just because some games run the same, because of bad coding or the GPU wall or whatever, doesn't mean that two CPUs have the same speed.

 

[cytg111]

Seems to me that AMD and Intel is converging at that same point of leveled off top-of-whats-possible for single threaded performance. So I guess my answer to that poll should be : Doesnt matter.

 

[NTMBK]

We don't know what Zen 2's architectural improvements are, we don't know what clock speeds 7nm will hit, and we don't know when Intel will finally fix their 10nm. So in short, nobody knows.

 

[beginner99]

no intel will not loose it but it also doesn't matter all that much. It only matters if you game at 1080p >=120hz. At higher resolutions you will be GPU limited and with a 60 hz display most CPUs are good enough.

Anything above 1080p you will need to spend >1000 on GPU for the CPU to matter. What will matter more is price. AMD has an opening now because of intels supply issues. However they may loose that opening with 7 nm. I suspect we will see a price increase with Zen 2.

 

[coercitiv]

We don't know what Zen 2's architectural improvements are, we don't know what clock speeds 7nm will hit, and we don't know when Intel will finally fix their 10nm. So in short, nobody knows.

When has that stopped us?!?!

 

[Vattila]

Some games just have internal timers or other limitations that make them just stop at some point,look at this all cores have some idle time GPU usage is low but FPS won't increase no matter what.

Yeah. The Forza Horizon 4 results at 1080p, showing the exact same frame rates, are odd and uncharacteristic. You would normally expect the little gap at 1440p to grow somewhat at 1080p. So I may have overstated my point with that particular result. But my general point is that with modern games you will not see the 50% outliers (such as the very old CS:GO). Rather, the difference will continue to narrow with modern well-written code optimised for Zen.

I don’t think we can make conclusions about 7nm when we haven’t seen the high power process. You can’t assume a shrink will lead to higher clock speeds now.

Well — simply put — if TSMC's entry into the high-performance space with their 7nm HPC process is unable to match Intel's frequencies on 14nm, then they have fallen way short. I think that is a very pessimistic expectation. Make no mistake about it, TSMC is dead set on targeting the HPC market, to ensure further growth and to diversify their customer base away from being overly reliant on the mobile market, and on one single big customer, Apple, in particular.

Also, I suspect that AMD and TSMC have worked on this for a long time, more closely than most analysts realise, with TSMC being well aware of the needs and frequency targets AMD has for their roadmap.

In short, expect TSMC's HPC version of the 7nm process to clock much better than their mobile-oriented mainstream version of the process.

They have […] a higher latency with the CCX vs ring bus.

There is a myth that a ring bus is good for gaming. No, it is good for bad code. As far as I understand (as a programmer myself), ring bus is simply more amenable to badly written multithreaded code that has a lot of interdependencies (shared memory accesses) and thus requires a lot of synchronisation (locks). Such code does not scale (and high-performance parallel code, such as for server and super-computer, is hence not written like that).

The ring bus architecture may somewhat alleviate such bad code by providing a relatively uniform latency between threads, while the CCX architecture will strongly penalise interdependent threads across CCXs. However, if your code has threads with less interdependency, and keeps the most inderdependent threads grouped within CCXs, you should see better results than the ring bus architecture, as intra-CCX latency can be near optimal within the tightly integrated 4-core complex, with direct connections between cores.

In short, my expectation is that modern game code that is well written and optimised for the Zen architecture will reduce the Intel lead substantially, with the remaining gap mostly due to the frequency deficit.

Zen is a workstation architecture that happens to be decent at gaming. While I suspect the gap will be very narrow, I don‘t think AMD will release a faster gaming chip. Zen just isn‘t build for doing that.

Your reply is characteristic of many posts here expressing scepticism about what AMD can achieve architecturally. Apart from the Bulldozer years (which was a difficult time for AMD, due to PC market slowdown and market shifts, ATI integration, conversion to a fabless model, big changes in design methodology, settlement of longstanding litigation with Intel, and the death of one their corporate fellows and key architectural contributors, chief architect Chuck Moore), AMD has a long history of beating Intel with architectural innovation.

Personally, I have no doubt that the engineers at AMD are equally as capable when it comes to x86 architecture and system design compared to Intel. There is no reason to believe otherwise. And in this area it is not a question of resources. While their product and market focus may be narrower due to less resources, you can be sure AMD has the lead designers and ground-floor engineers they need to work on the Zen architecture.

So, to be blunt, if AMD is unable to beat 14nm Coffee Lake (a Skylake-derivate with insignificant architectural advances), then they have fallen well short of their target. It is obvious, and has been explicitly stated by AMD executives in presentations and interviews, that Zen 2 targeted 10nm Ice Lake. It makes no sense for AMD to target anything less. It would be defeatist.

By the way, if rumours are to believed, then it looks like Ryzen 3000 will be 8-core, which strongly suggests that AMD no longer thinks they need a core-count advantage to compete.

Can they execute — unlike in the Bulldozer years? That is the question. AMD's CEO Lisa Su seems confident. She has under-promised and over-delivered so far.

 

[dacostafilipe]

If they don't mess up the architecture, I see no reason for AMD not to outperform Intel's actual architecture.

If they don't mess up the architecture, I see no reason for Intel not to outperform AMD's next architecture.

If they don't mess up the architecture, ...

 

[Hans Gruber]

Zen is a workstation architecture that happens to be decent at gaming. While I suspect the gap will be very narrow, I don‘t think AMD will release a faster gaming chip. Zen just isn‘t build for doing that.

Frametime experience will probably be close to indistinguishable, especially with tweaked settings (good ram etc).

Neither Intel nor AMD is a gaming chip. Because Intel has been the market leader for a decade. Game developers have optimized their games for intel CPU's. Obviously AMD has probably helped out developers to better optimize Ryzen CPU's in the last year. Many of you are computer noobs. The CPU and GPU war dates back decades.