Intel 9900KS = 5.0 GHz on ALL cores

[xg32]

1 minute ago, TrigrH said:

My freebie 8086k can to 5.2 full avx so does that make it better? I guess 5.3/5.1 is a more responsible way to clock it?

if it can do 5.2 why at decent temps why not, thats a good chip though *flex*

[TrigrH]

3 minutes ago, xg32 said:

if it can do 5.2 why at decent temps why not, thats a good chip though *flex*

speaking of flex, i need to fix that, my cpu PCB is quite bent. XD

[Ankerson]

I suppose I need to put my 8086K back in a machine.

 

Running an 8700K in the other gaming machine currently, need to drop the 8086K in that one.

 

Mine does 5.0 GHz all cores at 1.248V.

 

 

 

[JoostinOnline]

9 hours ago, Ankerson said:

and I would bet that they will OC quite a bit higher than the standard 9900K's.  

Why? It's just pre-overclocked, there's nothing special about it.

[Drak3]

Just now, JoostinOnline said:

Why? It's just pre-overclocked, there's nothing special about it.

Again, we don't know that yet.

[JoostinOnline]

1 minute ago, Drak3 said:

Again, we don't know that yet.

We know it beyond a reasonable doubt. Unless Intel lied about using binned 9900K's, then it should perform identically.

[Drak3]

Just now, JoostinOnline said:

We know it beyond a reasonable doubt.

No, we do not.

 

1 minute ago, JoostinOnline said:

Unless Intel lied about using binned 9900K's, then it should perform identically.

I've already covered this. You're operating under the assumption that they're at the same level of binning. Hell, Taf even brought up that there is a new stepping.

[Kisai]

4 hours ago, Taf the Ghost said:

Intel will call it a 135w part. This is going to be 180w or higher under most loads.

 

This launch review is suddenly relevant again.

 

https://www.techspot.com/review/1730-intel-core-i9-9900k-core-i7-9700k/page3.html

 

 

 

So, Intel is going to charge probably 599USD or more for this and you're going to need at least 100USD in cooler to even think of not throttling it under core-heavy load.

See this is where all the "Extra" performance has been coming from. We haven't really gained gross performance, rather we've gained net performance by allowing the energy creep rather than lowering it with a die shrink.

 

It's kinda criminal to release a desktop CPU or GPU that consumes more than 100 watts full power IMO. A laptop tops out at 220 watts. Given the right cooler, you get a faster CPU, but it still costs you more to run. A 600 watt GPU and a 250 watt CPU is already past where any reasonable UPS can handle it (1200 watts tops, and those are $1200 data center UPS systems, 865 watts is about $400, but not sine-wave.) I already killed the inverter on my UPS when it kicked in while running a game with a Geforce 1080.

[Drak3]

1 minute ago, Kisai said:

It's kinda criminal to release a desktop CPU or GPU that consumes more than 100 watts full power IMO.

Even though that's pretty typical and has been for a long time in the high end?

The 88W TDP 4790K can draw 120W at load.

The 84W TDP 4770K can just over 100W.

The 77W TDP 3770K can draw 140W

The 95W TDP 2700K can draw 145W.

The 125W TDP Phenom II X6 1100T can draw 180W

 

https://www.tomshardware.com/reviews/ivy-bridge-benchmark-core-i7-3770k,3181-23.html

https://www.tomshardware.com/reviews/core-i7-4790k-devils-canyon-overclock-performance,3845-9.html

[JoostinOnline]

21 minutes ago, Drak3 said:

No, we do not.

 

I've already covered this. You're operating under the assumption that they're at the same level of binning. Hell, Taf even brought up that there is a new stepping.

Seems like you just don't understand what you're talking about and really want there to be some special sauce involved.

[Drak3]

2 minutes ago, JoostinOnline said:

Seems like you just don't understand what you're talking about and really want there to be some special sauce involved.

No, I'm just not jumping to conclusions. You might be right.

 

Then again, you might not be.

[JoostinOnline]

Just now, Drak3 said:

No, I'm just not jumping to conclusions. You might be right. 

 

Then again, you might not be.

Explain to me what could reasonably be special about these if they are  the same silicon as 9900k's.

[Drak3]

Just now, JoostinOnline said:

Explain to me what could reasonably be special about these if they are  the same silicon as 9900k's.

 

28 minutes ago, Drak3 said:

Taf even brought up that there is a new stepping.

 

 

And if you want a really quick history lesson, we've seen in the past that batches of the same chip can overclock differently. The 5960X was not as good of an overclocker as the 5930K when both were introduced. Later into the production runs, the J batch of the 5960X came out hitting the same 4.7-4.8GHz mark as easy as the 5930K. That wasn't the case with prior batches.

[leadeater]

14 minutes ago, JoostinOnline said:

Explain to me what could reasonably be special about these if they are  the same silicon as 9900k's.

Steppings are rather minor changes and can be done for a host of reasons, typically when clocks are involved it's to improve the yields and bin ratios but it's rare for the upper point to be increased. You can see this on the 8700K vs the 8086K, the upper limit didn't change however the stepping as far as I know wasn't actually changed.

 

Steppings are more commonly used to fix bugs/errata and from the information that exists for R0 this stepping is going to be used on multiple (30+) products so it isn't a specific stepping for the purpose of increasing upper bound overclocking. All the major lineup for 9th gen is going to R0 stepping, down to i5-9400, because you don't maintain and produce two different steppings of the same product. Fortunately Intel documents errata and stepping changes so we will be able to find out what changed. If I were to bet on it and due to the cost of a stepping I would be looking in the area of security (Re: Intel recent news) with some changes to widen the upper percentile of dies to help out with the 9900KS.

 

So if this supposed increase in upper point for overclocking is coming from R0 it's also coming to the 9900K and 9900KF. Then of course you could say every single chip is binned and only the dies that can do 5.0 all core will be used on the 9900KS, which is a bold claim and one that relies on the sales of the 9900KS to be high enough to even allow that.

 

With all the 14nm shortages Intel withholding large stocks is highly unrealistic, highly capable dies have always made it in to lesser products because more of those sell.

[SeraphicWings]

All cores on 5GHz base?

You probably will need a nuclear reactor to power this thing and the entire Antartica to cool it, I think?

[Drak3]

6 minutes ago, SeraphicWings said:

All cores on 5GHz base?

You probably will need a nuclear reactor to power this thing and the entire Antartica to cool it, I think?

It's not going to be substantially different from other octocores on Skylake derived architectures.

[StarsMars]

Introducing the $700 Intel space heater.

EZ bake oven

toaster

irrelevant product

heh heh..

[Kisai]

2 hours ago, Drak3 said:

Even though that's pretty typical and has been for a long time in the high end?

The 88W TDP 4790K can draw 120W at load.

The 84W TDP 4770K can just over 100W.

The 77W TDP 3770K can draw 140W

The 95W TDP 2700K can draw 145W.

The 125W TDP Phenom II X6 1100T can draw 180W

 

https://www.tomshardware.com/reviews/ivy-bridge-benchmark-core-i7-3770k,3181-23.html

https://www.tomshardware.com/reviews/core-i7-4790k-devils-canyon-overclock-performance,3845-9.html

I have a i7-4770. and a Geforce 1080. Right know the UPS says *drumroll* 145 watts, this is basically just idle.

now if I jump over to cpu-z stress, it jumps to 193w tops, so let's say the difference between full load and idle load is about 50 watts. If I switch over to GPU-z and run the render test, it goes to 270 watts. Both 301 watts.

GPU-Z says it tops out at 115 watts. Yet... it tops out at 390 watts in FFXV. Energy creep is real.

 

Basically what I'm saying is that these are not data center CPU's, they should not consume that much more than a laptop's CPU full tilt. Like I expect 42-45 watts from a Haswell dual-core, because logically it has half the cores, yet somehow when we get 8 cores in a i9-9xxx CPU, at half the nm process, you'd expect the energy consumption to be the same at full tilt. Yet not the case

Source: https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/21

 

That full load i7-6700K and 7700K increased 12 watts. Now take the note I mentioned above about my CPU idle/full load difference being about 45 watts difference. Now look at the i7-9700K and i7-8700K, remember these are ALL the same 14nm process node (5xxx,6xxx,7xxx,8xxx,9xxx).

 

So:

6700K, stated 95w, reality 83w (4C) (-12w)

7700K, stated 91w, reality 95w (4C) (+4w)

8700K, stated 95w, reality, 146w (6C) (+51w)

9700K, stated 95w, reality 125w (6C) (+30w)

for completeness

i9-9900K, stated 95w, reality 169w (8C) (+74w)

i7-8086K, stated 95w, reality 101w (6C) (+6w)

 

The closest any of these are to their target TDP is the 8086K and 7700K. Now you'll also notice that that +51 and +30 bump added two cores, which you'd think would be expected and the CPU's marked as such, but no, we'd have to have the clock speed dialed down to hit that TDP, thus making them not as attractive. We also see this with the 9900K, so it goes from 4 cores in the 6700K/7700K to 8, so it's power consumption should double, and yet it looks like is double... of the previous 22nm process.

 

Don't get me wrong, this would not be a good enough reason to avoid buying such a thing, but it would be something that needs to be taken into account if you need to keep a power budget.

[williamcll]

Can't wait for 6Ghz Intel chips. 

[Drak3]

7 minutes ago, Kisai said:

you'd expect the energy consumption to be the same at full tilt.

I wouldn't.

[porina]

5 hours ago, Kisai said:

It's kinda criminal to release a desktop CPU or GPU that consumes more than 100 watts full power IMO.

High end GPUs have been around 250W for many generations and that would seem to be the practical limit for consumer facing stuff. I think everyone is used to it. CPU wise I think around 200W (overclocked) is a practical limit. That's still manageable with air cooling.

Quote

A 600 watt GPU and a 250 watt CPU is already past where any reasonable UPS can handle it (1200 watts tops, and those are $1200 data center UPS systems, 865 watts is about $400, but not sine-wave.) I already killed the inverter on my UPS when it kicked in while running a game with a Geforce 1080.

What are you running to get 600W through a GPU?

 

I haven't ran a UPS in a long time, but a typical high end non-OC gaming system can comfortably run off a 550W PSU which isn't too bad for a UPS.

[leadeater]

12 minutes ago, porina said:

What are you running to get 600W through a GPU?

 

I haven't ran a UPS in a long time, but a typical high end non-OC gaming system can comfortably run off a 550W PSU which isn't too bad for a UPS.

My OC 4930K and dual 290X OC only just pushes mid 850W, must be a Vega 56 with power play tables and a rather epic OC to be hitting 600W single GPU.

[RejZoR]

These high clocked CPU's are what keeps Intel in the game as they are great for gamers and gamers are the ones doing them the loudest PR literally by paying Intel for it when they buy these CPU's. As a whole Intel is a bit weak, but for gaming, they are still the best if performance is all it matters and you don't care if it's a furnace in your PC.

[JoostinOnline]

6 hours ago, Drak3 said:

 

 

 

And if you want a really quick history lesson, we've seen in the past that batches of the same chip can overclock differently. The 5960X was not as good of an overclocker as the 5930K when both were introduced. Later into the production runs, the J batch of the 5960X came out hitting the same 4.7-4.8GHz mark as easy as the 5930K. That wasn't the case with prior batches.

That's far too minor of a change to consider it special. It's just the higher end of a spectrum.

[mr moose]

3 hours ago, Kisai said:

The closest any of these are to their target TDP is the 8086K and 7700K.

 

The anandtech test was at all turbo frequencies, not base.  So the their power draw results with regard to TDP spec is not a meaningful observation.