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

Page 26 - 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
  • This poll will close: .

Zucker2k

Senior member
Feb 15, 2006
809
106
136
Are you suggesting the 9900K should ONLY be run with DDR4-2666 because that is the 'official' spec? Who buys a $500 CPU to gimp it with slow memory, especially since DDR4-3200 isn't really that much more expensive these days? You're literally paying $10 - $20 more for a 3200 kit instead of 2666, for a ~10% increase in performance on a $500 CPU, thats a no brainer. So Im not really buying that argument. If anything, I would think that most 9900K owners (and 3900X as well) would be running high end RAM kits with 3600 or even 4000 speeds. If you're gonna go high end you go high end with ALL components, not super expensive CPU with dirt cheap RAM.
I want to see his reaction when those 5.2Ghz 9900ks pop up in benchmarks! Heh

I really don't understand the double standards. AMD can push their chips to the hilt (which is okay) but exploiting the free headroom baked into Intel chips is sacrilegious /s. I am reminded of that popular remark: "Ryzen loves fast ram." If there ever was a compelling reason to run benchmarks @3466 speeds on Ryzen systems, this is the one! LOL
 

Zucker2k

Senior member
Feb 15, 2006
809
106
136
How much more frequency, in absolute and percentage terms, does Zen 2 need on the active cores throughout the benchmarks, to tilt the balance?

Clocks alone won't do it, unless we assume that latency will scale with clocks in a linear fashion. The bigger cache on the 3950x will help but I don't see it as a game changer.
 

Zucker2k

Senior member
Feb 15, 2006
809
106
136
As for overclocking a 3900X, isnt that rather pointless? PBO gets you 99% there anyway, and you don't lose the higher single core boosts either.

In fact, it seems the best thing to do for a 3900X isn't to overclock it, it's to get a better cooler on it and let PBO do its thing.
Only way you can do better than PBO is to invest in exotic cooling (which doesn't count) So I think 100% may be more appropriate for PBO boosts in this iteration. Even seasoned overclockers are struggling to hit 4.5GHz all clocks without resorting to exotic cooling.
 

Abwx

Diamond Member
Apr 2, 2011
8,939
298
126
I want to see his reaction when those 5.2Ghz 9900ks pop up in benchmarks! Heh

I really don't understand the double standards. AMD can push their chips to the hilt (which is okay) but exploiting the free headroom baked into Intel chips is sacrilegious /s. I am reminded of that popular remark: "Ryzen loves fast ram." If there ever was a compelling reason to run benchmarks @3466 speeds on Ryzen systems, this is the one! LOL
If that s stock why not..?

The double standard is to technically overclock a chip and do as if it were stock while it s on user setings, AMD has validated their IMC at 3.2 so it s fair to do first comparisons at manuacturers specs.

Besides the 9900K wont gain that much past 3200, we ll see once we have some more exhaustive reviews on this matter.
 
Sep 4, 2016
82
49
61
I agree that clocks alone won't solve the problem... On those games that performance are really lagging behind would be good to have some code profiling done. Based on the way IPC is for most applications right now, I am sure some code tuning can greatly reduce the gap or even turn the game around...
 

Zucker2k

Senior member
Feb 15, 2006
809
106
136
If that s stock why not..?

The double standard is to technically overclock a chip and do as if it were stock while it s on user setings, AMD has validated their IMC at 3.2 so it s fair to do first comparisons at manuacturers specs.

Besides the 9900K wont gain that much past 3200, we ll see once we have some more exhaustive reviews on this matter.
How can you say the 3900x is "stock" when PBO is actively maximizing (overclocking) the chip in real time, much more effectively/efficiently than even a manual overclock? There were lots of noise made here when Intel announced the 9900k. People condemned Intel for factory-overclocking the chip and leaving little room for overclocks. Well, let's see the performance of your overclocked 3900x against the "stock" 3900x then.
 

Abwx

Diamond Member
Apr 2, 2011
8,939
298
126
How can you say the 3900x is "stock" when PBO is actively maximizing (overclocking) the chip in real time, much more effectively/efficiently than even a manual overclock? There were lots of noise made here when Intel announced the 9900k. People condemned Intel for factory-overclocking the chip and leaving little room for overclocks. Well, let's see the performance of your overclocked 3900x against the "stock" 3900x then.
It doesnt need overclocking, if you had payed attention you would had noticed that Computerbase not only used stock speed on RAM but they also kept the latency at its stock value for AMD, and the result is a 2-3 % difference, frequency wise they let the 9900k run wild..

What i m saying is that from 2666 to 3200 (CL 14...) the 9900K has exhausted most of the improvement from the RAM and has no room left since the fequency potential is already maxed out, hence the desperately announced 5.2GHz...

Btw, AT updated their review, you should check the results in games..
 
Feb 23, 2017
600
544
106
How can you say the 3900x is "stock" when PBO is actively maximizing (overclocking) the chip in real time, much more effectively/efficiently than even a manual overclock? There were lots of noise made here when Intel announced the 9900k. People condemned Intel for factory-overclocking the chip and leaving little room for overclocks. Well, let's see the performance of your overclocked 3900x against the "stock" 3900x then.
PBO isn't even working. It is supposed to push up power and current limits to allow for higher boost clocks, yet we've not seen a single review where Ryzen went above its own listed clocks. Hitting it's clocks is a function of PB2 and XFR2, which are stock settings; limited to 142w and 95a whilst thermally constrained.
 
Apr 27, 2000
12,376
1,325
126
I want to see his reaction when those 5.2Ghz 9900ks pop up in benchmarks! Heh
And just how many people can actually run a 5.2 GHz 9900k? These so-called "self overclocking" Matisse chips can behave the way they do 100% of the time. Plus as @PotatoWithEarsOnSide indicated, PBO doesn't even appear to be working on the higher-clocked SKUs.
 

Dave3000

Senior member
Jan 10, 2011
901
4
81
Are you suggesting the 9900K should ONLY be run with DDR4-2666 because that is the 'official' spec? Who buys a $500 CPU to gimp it with slow memory, especially since DDR4-3200 isn't really that much more expensive these days? You're literally paying $10 - $20 more for a 3200 kit instead of 2666, for a ~10% increase in performance on a $500 CPU, thats a no brainer. So Im not really buying that argument. If anything, I would think that most 9900K owners (and 3900X as well) would be running high end RAM kits with 3600 or even 4000 speeds. If you're gonna go high end you go high end with ALL components, not super expensive CPU with dirt cheap RAM.
I think I'm going to be upgrading to a 9900k soon. Am I better off buying non-XMP JEDEC spec'd DDR4-3200 CL 22 1.2v or DDR4 3200 CL 16 1.35v that is only through the XMP profile but has only 2400 speeds for the JEDEC profiles. I seen reviews on DDR4-3200 and one of them showed CPU-Z showing the memory's profiles in the SPD and it was 2400 for the JEDEC profiles and 3200 for the XMP profile. Would memory like that be considered factory overclocked memory and not real DDR4-3200 and am I getting ripped off if I buy memory that has it's advertised speed only through XMP profiles but much slower speeds through it's JEDEC profiles? Also what are the chances of buying a 9900k that won't run DDR4-3200 RAM reliably since the 9900k officially supports up to DDR4-2666?
 
Last edited:

Zucker2k

Senior member
Feb 15, 2006
809
106
136
And just how many people can actually run a 5.2 GHz 9900k? These so-called "self overclocking" Matisse chips can behave the way they do 100% of the time. Plus as @PotatoWithEarsOnSide indicated, PBO doesn't even appear to be working on the higher-clocked SKUs.
It is working.

Only problem is that there is no more power/thermal headroom. The chips are maxed out from the factory.

Edit: From The Stilts musings from the link above, I believe Asus engineers tried to add an additional 27watts on top of AMD's spec in order to help PBO boost pass the measly 25Mhz boosts they were getting on their particular boards..
 
Last edited:
Feb 23, 2017
600
544
106
Which part of PBO overrides all AMD set limits is hard to understand?
ASUS throwing 27w beyond the 142w is neither here nor there; it is just an attempt to alter a PB2 and XFR2 limit, not PBO. PBO has no power or current limits other than those mobo VRM ones that incidentally can handle upwards of 400w through the CPU socket.
 

Zucker2k

Senior member
Feb 15, 2006
809
106
136
Which part of PBO overrides all AMD set limits is hard to understand?
ASUS throwing 27w beyond the 142w is neither here nor there; it is just an attempt to alter a PB2 and XFR2 limit, not PBO. PBO has no power or current limits other than those mobo VRM ones that incidentally can handle upwards of 400w through the CPU socket.
So what's holding PBO back, in your opinion? I'm sure AMD did not release Zen 2 with a non-functioning PBO, if that's what you're implying.
 

rbk123

Senior member
Aug 22, 2006
609
93
116
LOL how am I cherry picking when I specifically referenced GN and was using it as my source for my argument? Think about that for a sec, please. I've said many times in this thread that the 3900X is very close to a 9900K in gaming, but that should be obvious to everyone already.

However, if you actually had to rank the CPUs in terms of their gaming performance, then IMO the 9900K/9700K is 'tier 1' due to the higher stock clocks, and the 8700K/3900X just slightly lower at 'tier 2' mainly due to the lower clocks. If you overclock then the 9900K/9700K/8700K basically all perform the same since the clockspeed ceiling is about 5.0 - 5.1GHz for all these chips, and at this point there is a clear distinction between it and say a 3900X (or any 3000 series Ryzen chip) that is limited to around 4.3GHz.
Yawn. Sorry, no matter how much you drone on, when you use loosey-goosey terms like you purposefully did, and for obvious reasons, then you have to add the 3900X into the group. When it can beat the 9900K in some games and come in a few frames behind, it qualifies for your own "can hang with".
 
Feb 23, 2017
600
544
106
So what's holding PBO back, in your opinion? I'm sure AMD did not release Zen 2 with a non-functioning PBO, if that's what you're implying.
Something within the BIOSes that is preventing those limits from being overridden. Even Anandtech said in day dot review that BIOS was not allowing PPT to be overridden, but since everyone seems to skim reviews just to browse flawed gaming benchmarks, no-one pays attention to the implications of what is being said.
 

Zucker2k

Senior member
Feb 15, 2006
809
106
136
Something within the BIOSes that is preventing those limits from being overridden. Even Anandtech said in day dot review that BIOS was not allowing PPT to be overridden, but since everyone seems to skim reviews just to browse flawed gaming benchmarks, no-one pays attention to the implications of what is being said.
We shall revisit this response in the coming weeks.
 
Feb 23, 2017
600
544
106
We shall revisit this response in the coming weeks.
By all means.
Opinions aren't always correct.
I'm stating what I'm seeing from what little actual data gets posted.
If PBO doesn't do anything, why bother invalidating your warranty by turning it on?
AMD sent reviewers the clock scaling for PB2, and Robert Hallock also kindly produced a video detailing exactly what PBO does. If I see reviews that conform to PB2 behaviour despite PBO being switched on, then I'm entitled to believe that PBO isn't working properly. Unless, and until, we see a clock above the listed spec, the rational assumption is that something isn't working.
 

epsilon84

Senior member
Aug 29, 2010
978
158
136
Yawn. Sorry, no matter how much you drone on, when you use loosey-goosey terms like you purposefully did, and for obvious reasons, then you have to add the 3900X into the group. When it can beat the 9900K in some games and come in a few frames behind, it qualifies for your own "can hang with".
Talk about pedantic with wording haha. I'll try to be more concise in future so your feelings aren't hurt about the 'gaming champ' 3900X that came out by 1% or whatever in a particular review ;)

I think this particular topic is 'more or less' done so... :p
 

CHADBOGA

Golden Member
Mar 31, 2009
1,829
102
136
It doesnt need overclocking, if you had payed attention you would had noticed that Computerbase not only used stock speed on RAM but they also kept the latency at its stock value for AMD, and the result is a 2-3 % difference, frequency wise they let the 9900k run wild..

What i m saying is that from 2666 to 3200 (CL 14...) the 9900K has exhausted most of the improvement from the RAM and has no room left since the fequency potential is already maxed out, hence the desperately announced 5.2GHz...

Btw, AT updated their review, you should check the results in games..
AT has some of the worst gaming benchmarks on the net.

They really need to do a major overhaul of their gaming tests and video card used.

Hopefully other sites will soon update their gaming benches with the new bios or whatever it was that has extracted more single core performance.
 

epsilon84

Senior member
Aug 29, 2010
978
158
136
AT has some of the worst gaming benchmarks on the net.

They really need to do a major overhaul of their gaming tests and video card used.

Hopefully other sites will soon update their gaming benches with the new bios or whatever it was that has extracted more single core performance.
I believe Anandtech still tests gaming with a GTX 1080 (!), not even a 1080 Ti. That's basically an upper mid range card these days, so their gaming tests aren't really useful IMO because they are so GPU limited.
 

CHADBOGA

Golden Member
Mar 31, 2009
1,829
102
136
I believe Anandtech still tests gaming with a GTX 1080 (!), not even a 1080 Ti. That's basically an upper mid range card these days, so their gaming tests aren't really useful IMO because they are so GPU limited.
Hopefully if a 5900XT or 3080 Ti comes out in the next year, AnandTech adopts this card as their one for their gaming benches, in addition to adding a few more games.
 

rbk123

Senior member
Aug 22, 2006
609
93
116
Talk about pedantic with wording haha. I'll try to be more concise in future so your feelings aren't hurt about the 'gaming champ' 3900X that came out by 1% or whatever in a particular review ;)

I think this particular topic is 'more or less' done so... :p
No problem Shaggy. As long as you want to stick with cpu's that can hang with the 9900k, I'll continue telling you the 3900K can hang right in there. Don't feel so bad about putting your foot in your mouth, feel bad that you keep changing feet. You'll get better, though, I have faith in you.
 
May 27, 2002
12,446
38
91
I'm still waiting for the promised 9900K price drop of $75 to $420... any chance its coming soon?
 

TheELF

Platinum Member
Dec 22, 2012
2,864
128
126
Something within the BIOSes that is preventing those limits from being overridden. Even Anandtech said in day dot review that BIOS was not allowing PPT to be overridden, but since everyone seems to skim reviews just to browse flawed gaming benchmarks, no-one pays attention to the implications of what is being said.
Der8auer already checked it,you need some crazy ass low temps for ryzen to clock higher.
You need 30° just to hit 4275Mhz.
For games it's a bit less demanding on cooling but not all that much.
 

TheELF

Platinum Member
Dec 22, 2012
2,864
128
126
What I disagree with, is that the 3900x comes with a decent cooler, and the 9900k comes with NOTHING, but they don;t add a decent cooler price before they calculate those numbers. NOT FAIR !
Yeah!Cry about it being unfair towards AMD when they show 3 gens ,or cycles or series, of AMD CPUs but only the most expensive 9ths gen CPUs.
Also the $255 i5-9600k hits the same 135 FPS avg of both the $330 3700x and the $500 3900x....
Those minimum $75 gets you a pretty decent cooler, me thinks.
 

Similar threads



ASK THE COMMUNITY