Intel 9th Gen CPUs revealed

[Durge Driven]

i would like to know the real world results from your i5 9600k .

Here is real world
9600K@Turbo vs 9700K @5GHz





I know which build I would have

 

[Kek700]

There both scary prices , the 9700 will give me the solid 118 fps i need in AC?
neither will give me a solid 118 in ACC ? ( with settings in game, at a best view option )
i am better off financially, loosing 10 fps in AC , and accepting 75 fps in ACC ( G sync does work
well at providing a smooth image ) to keep both my CPU and GPU working at less than max.
i have already forked out £300 for a second hand over clock water cooled 1080. ( 2025 Mhz )
which i must say was not exactly a startling improvement on my 980ti... ( 1350 MHz )
The thought of shelling out another £700 on a 9700 and mother board for a few more fps
is, for me, a bridge too far. ( i’m sure i will succumb to this though ) hope eternal
i may be a bit “naive” here, but there seems to be a lot of money changing hands for, not really,
overall, a much improved viewing experience. ( i am alway referring to AC or ACC. )
there is the hope that ACC will be better optimised in the future.?

 

[RobertR1]

You will probably see perf optimizations for ACC post release.

You’re a bit maxed out currently. The next step is unfortunately a full overhaul. There are no single upgrades to be had. You’ll just end up with an imbalance.

Also, an overclocked 980ti is a decent contender to the 1080 so you should not expect a massive jump. It’s the 1080ti and above where you get a major jump but then the bottleneck will shift to your other parts.

 

[Kek700]

Unfortunately, I think I have just come to that
conclusion.
The 1080 is a fair bit better than 980 ti, but mainly due to being clocked to 2025 MHz.
I get “about” an extra 30 FPS with it in ACC. But nothing with AC. So I am happy about that.

 

[David Turnbull]

@Durge Driven

That first system you posted is near exactly what I bought, I have the 9600k, the Corsair Vengeance Pro LED 3200mhz ram and the Gigabyte z390 Gaming X, I added a Corsair H100x cooler and an MSI RTX2070 8gb.

Here in the UK the 9700k is at least another £150 over the price of the 9600k so I couldnt justity it when I bought it, but having the 9600k now gives me upgrade options for the future, I could either go to the 9700k or the 9900k when prices fall in a year or two.

 

[Turk]

There both scary prices , the 9700 will give me the solid 118 fps i need in AC?
neither will give me a solid 118 in ACC ?

They're two completely different things. I don't think you need a 9th gen CPU to run AC at 118fps. I have a i7 8086 with a 1070 and I can run it at maxed settings in VR with 2x SS.

ACC is another story but I'm holding out hope that ACC performance will get much better with optimisation. I'm at medium settings and if it would get just a little bit better and smoother I'll be happy with that performance.

 

[Kek700]

I have thought about an 8086 ( very attractive buy ) but it is still £400 plus £200 mother board.

 

[RasmusP]

"you don't need a 9th Gen cpu" - you just need a special 8th Gen edition that's basically identical performance wise to the 9th Gen if not better due to hyperthreading...

You don't need a 2080, a 1080ti is also fine

 

[RobertR1]

Like a halo laptime, I'm quite happy to have broken the 16k barrier on time spy after lots of tuning. I should spend a 1/10th of this much time learning car setup tuning but I find this more fun

I'm on Air cooling for GPU and CPU so I don't have much more to extract at this point. With water, I could possiblly do more.

https://www.3dmark.com/3dm/33358882
Score 16 006
Graphics Score
16 684
CPU Score
13 014

There's a lot of noise that overlocking newer gen of hardware is not really worth it. I think quite worth it. This is a what a stock 9900k/2080ti combo nets you. https://www.guru3d.com/articles-pages/intel-core-i9-9900k-processor-review,22.html

You can see there's a notable delta. Without much effort, I can run daily at a performance level that nets me around 15k points which is still well ahead of an out of the box 13.2k score. For VR, that extra oomph matters especially for keeping your low framerates up.

 

[Kek700]

@RasmusP
@Martin Fiala
or any one that can help????

Msi afterburner
Brands hatch GP Gt3 ( not that it matters )
23 Ai
AC
i5 6600k
Using FPS & Ai to load CPU.

Limited to 60fps all cpu cores at about
65% load

Removing FPS cap....
Now 1 Ai.....
FPS at 290

All 4 cores at 90 + % .. sometimes railing at 100%.

My point is, all cores are in unison with one
and another.
From what has been hinted at , I expected
only one core to be loaded, but when one
core railed , all 4 cores railed also.

It’s like I have a 1 core 4 core processor.
It almost makes me think that if I replaced
my i5 with a i9 9600k assuming they a similarly clocked all I would have is extra cache; all 6 cores
will be in unison.
I can see the gain in that , as long as 1 core in the 4 core equals 1 core in the 6 core.
It’s like the processor is using all it’s cores
irrespective of the software requirements ?
Which leads onto more questions
I am obviously missing something, can you
give a corrective explanation .
I think an explanation has priviously been
forwarded to me but unfortunately I sm still
confused.


I do realise that I am looking at a snap shot from
Msi afterburner every few hundred millisecond.
But at this point in time, I cannot see that that matters

 

[Sim Strata]

You don't need a 2080, a 1080ti is also fine

Isn't that true though? I read the 20 series GPUs maybe enter legacy status sooner then any prior GPU family.

 

[RobertR1]

@RasmusP
@Martin Fiala
or any one that can help????

Msi afterburner
Brands hatch GP Gt3 ( not that it matters )
23 Ai
AC
i5 6600k
Using FPS & Ai to load CPU.

Limited to 60fps all cpu cores at about
65% load

Removing FPS cap....
Now 1 Ai.....
FPS at 290

All 4 cores at 90 + % .. sometimes railing at 100%.

My point is, all cores are in unison with one
and another.
From what has been hinted at , I expected
only one core to be loaded, but when one
core railed , all 4 cores railed also.

It’s like I have a 1 core 4 core processor.
It almost makes me think that if I replaced
my i5 with a i9 9600k assuming they a similarly clocked all I would have is extra cache; all 6 cores
will be in unison.
I can see the gain in that , as long as 1 core in the 4 core equals 1 core in the 6 core.
It’s like the processor is using all it’s cores
irrespective of the software requirements ?
Which leads onto more questions
I am obviously missing something, can you
give a corrective explanation .
I think an explanation has priviously been
forwarded to me but unfortunately I sm still
confused.


I do realise that I am looking at a snap shot from
Msi afterburner every few hundred millisecond.
But at this point in time, I cannot see that that matters

When you took the FPS limit off, the GPU was able to exercise it's potential which meant it was asking more and more of the CPU.

Change in the game, windows OS, cpu architecture, PC hardware can all have an impact on which resources are used and which becomes bottlenecked.

Perfect example would be, on my 9900k at one point turning hyper threading off was faster but then I made some changes and enabled HT which gave me fewer dips.

There's no prescriptive answers. You just need to find a balance given your scenario. That's the beauty and curse of PC gaming.

 

[RobertR1]

Isn't that true though? I read the 20 series GPUs maybe enter legacy status sooner then any prior GPU family.

Nvidia's 7nm at the high end won't be ready until 2020 so the you're probably looking at 1.5years between the 2080ti and next iteration.

 

[RasmusP]

Isn't that true though? I read the 20 series GPUs maybe enter legacy status sooner then any prior GPU family.

Yes. My point was that a 1080ti and a 2080 basically have the same performance.
Same for the special 8th Gen i7 and the 9th Gen i7.
I tried to be ironic and make a joke about the stupid statement I quoted.

 

[RasmusP]

@RasmusP
@Martin Fiala
or any one that can help????

Msi afterburner
Brands hatch GP Gt3 ( not that it matters )
23 Ai
AC
i5 6600k
Using FPS & Ai to load CPU.

Limited to 60fps all cpu cores at about
65% load

Removing FPS cap....
Now 1 Ai.....
FPS at 290

All 4 cores at 90 + % .. sometimes railing at 100%.

My point is, all cores are in unison with one
and another.
From what has been hinted at , I expected
only one core to be loaded, but when one
core railed , all 4 cores railed also.

It’s like I have a 1 core 4 core processor.
It almost makes me think that if I replaced
my i5 with a i9 9600k assuming they a similarly clocked all I would have is extra cache; all 6 cores
will be in unison.
I can see the gain in that , as long as 1 core in the 4 core equals 1 core in the 6 core.
It’s like the processor is using all it’s cores
irrespective of the software requirements ?
Which leads onto more questions
I am obviously missing something, can you
give a corrective explanation .
I think an explanation has priviously been
forwarded to me but unfortunately I sm still
confused.


I do realise that I am looking at a snap shot from
Msi afterburner every few hundred millisecond.
But at this point in time, I cannot see that that matters

On holidays from next Monday ongoing. I'm gonna make a comparison with and without hyperthreading and with 3 vs 4 cores (and ht on/off) so you'll be able to see the differences.
Simple answer:
Let's say a game runs on 1 thread.
You have 4 cores.
1 thread can't be split so the theoretical maximum cpu load is either 1 core at 100% or all cores "shuffling" and showing a load of 25%.

I think up to this point everything is clear?

Now this "shuffling", that windows does makes the load on cores going higher to some extend to let's say all cores loaded up to 40% since sharing and preparing the cache or whatever the cpu does internally becomes more efficient than just loading one single core to 100%.

So from 1x 100% you went to 4x 25% and due to efficiency increased you end at 4x 40%.

Now this effect of increasing efficiency multiplies with more threads. So with 2 threads you'll probably see it going from:
2x 100% to 4x 50% and then efficiency increasing to 4x 75%.

And then, while the games have "only 2 big main threads" for example, you'll have about 20-70 "mini threads" that still take some amount of cpu load.

I'm still wondering about your i5 hitting 100% on all 4 cores though. I'm gonna check this next week while disabling hyperthreading and a core.
But maybe there is just something else eating your cpu? Could you tell us what the "overall load" graph shows? Because when each core shows "95%", which looks like full load on the small afterburner graphs, it would only be
80% overall load.

 

[Sim Strata]

On holidays from next Monday ongoing. I'm gonna make a comparison with and without hyperthreading and with 3 vs 4 cores (and ht on/off) so you'll be able to see the differences.

I tried that with AC. Hyperthreading makes no noticeable difference with a 3770K. Performance was something like 25fps with a single core. Even disabling one core had a hit.

 

[Case_]

I've done some comparison tests between HT on and off back with the i7 2600, thinking you might perhaps get a somewhat better single core performance with HT off, and my conclusion was that you don't, you just further limit your performance by turning off HT.

My point is, all cores are in unison with one
and another.
From what has been hinted at , I expected
only one core to be loaded, but when one
core railed , all 4 cores railed also.

If you expected only one core to be loaded based on the fact that it's been mentioned this or that game only "uses 1 or 2 threads and maxes them out", then it's a common mistake to think of threads in "core" terms. You might think a single-threaded task will only put strain on one physical core, because it kinda seems to make sense that if you run a thread that is maxed out or nearly maxed out, you just dedicate a single core to it and keep it running at that single core. That's not really how it works, though - the system (as in both hardware and software) still spreads the thread load across multiple cores, because it is a more effective way of running it (see Rasmus' explanation). This, in my experience, seems to be even more true with Intel processors - with Ryzen, it seems that I see full (or near full) one core usage a lot frequently than with my previous i5/i7 where it was very rare to see something like that.

So yeah, even if you run a load that is mostly single-threaded and manages to max out that thread, you're still unlikely to see this reflected in core usage numbers. That's one of the reasons why looking at core/CPU usage is a lot less useful on modern multicore systems than you would expect it to be, and a lot less representative of what's actually going on.

1 thread can't be split so the theoretical maximum cpu load is either 1 core at 100% or all cores "shuffling" and showing a load of 25%.

I think saying that "thread can't be split" might be a bit confusing, as you then proceed to talk about basically "splitting" the thread between multiple cores I think what you meant to say was that one thread can't exceed the performance limit (or percentage of overall CPU usage) of a single core.

 

[Kek700]

I can sort of understand all your comments, but when i apply these to what i am seeing it does not compute ( not a pun ), my 4 cores are buzzing away doing thier stuff, at 60 fps. Then i let the fps run free , so now it’s in the hundreds and the gpu is working harder, that means the cpu has to work harder to keep up. Now this is the bit that i find confusing.
( I don't have threads to cloud the issue , just the thread allocated to a particular core )

We are supposed to think that this game is a one core wonder, so my i5 at 4.7Ghz is very capable of providing good single core performance compared to your 4 core 8 thread processor. which as you have shown in the past posts, is the case.
Even though it is at 4.5Ghz ( i think ).
( maybe if I under stood that reason a bit better, I would not be asking this question.)

But what do i see, all four cores moving in a reasonable synchronized fashion to the demands of my increased fps. Not 1 core being stretched, and 3 core keeping up with the extra marginal demands that may incur.

it makes the bit about single core performance seem irrelevant.

I am pretty sure that i am wrong about this, but it just does not seem so, when i am observing in Msi afterburner.

if I load a random programme, the processor will give that to a core and you can see the other cores all raising there game a small amount, but generally just one core keeps the demand going for that programme.

it's really a problem when you know you are wrong but cannot see it with the evidence in front of you, I am such a numpty.

have a great holiday...

the 100% load thing is mainly for ACC. its very obvious in ACC.

 

[Sim Strata]

But what do i see, all four cores moving in a reasonable synchronized fashion to the demands of my increased fps. Not 1 core being stretched, and 3 core keeping up with the extra marginal demands that may incur.

I don't see quite the same thing. I note that with acs.exe there are about 44 threads running. With my eight processes (HT switched on) CPU0 is maxed out and CPUs 1-7 are all chugging away at about 20-30%. i7-3770k@4.7ghz.

 

[Kek700]

@Martin Fiala

wrote the above before reading your response.

it would seem to suggest that a 9700 k with its 8 cores should perform much better than i5 9600k
with it’s 6 and then my i5 6600k with 4 cores
i had just focused on single core performance as far as AC is concerned, it appears it is not
that simple. nothing seems to be simple in this world now. ( except me of course. )
i must then assume that if i loaded AC to a max , and i had a 9700k i would in theory see all 8
cores loaded in a simular fashion to my 6600k with its 4 cores all loaded.