Intel 9900KS = 5.0 GHz on ALL cores

[leadeater]

9 minutes ago, Drak3 said:

That makes the assumption that only now Intel is binning them this strictly. That might not be the case. It's possible that there were some chips that binned higher than exceptional 9900Ks but not enough to push them as their own pool of chips yet.

There isn't any assumption, the dies can only be as good as the fab that is making them and since they are of the same architecture, use the same fab equipment they are of all the same stock, even if they get made in Israel or US etc. They were in the pool of 9900Ks, 9900KS doesn't make them better but this can make the 9900K worse in that respect (if every single 5.0 cable die is only ever sold as a 9900KS).

 

If you actually cared enough to get a good die Silicon Lottery was the place to go, if these are sold for more than $590 then it's a bit of a kick to the balls.

[Drak3]

Just now, leadeater said:

There isn't any assumption, the dies can only be as good as the fab that is making them and since they are of the same architecture

There is a huge assumption:

 

There wasn't any binning beyond separating the 9900K from the other octocores on Z390. Nor will overclocking remain the same between batches.

[leadeater]

8 minutes ago, Drak3 said:

There is a huge assumption:

 

There wasn't any binning beyond separating the 9900K from the other octocores on Z390. Nor will overclocking remain the same between batches.

That's still irrelevant to the point, they were in the pool of 9900Ks. I didn't say they weren't binning them before, it didn't matter they were all 9900Ks so how would you know? Silicon Lottery was the place to go to find out, now we have the 9900KS. Nothing has changed on the die or fabing, they aren't better they just now have a new label on them.

 

The top percentile possible 9900K will still be exactly the same as the now top percentile possible 9900KS, the upper percentile hasn't changed, it can't change if there is no change. Names and labels don't actually impact the physical properties of the item.

[Dabombinable]

52 minutes ago, S w a t s o n said:

I mean we're getting into semantics here. The PL2 of the 9900k is over 200W no matter how you look at it. Yea it's kind of my point. The 9590 was an overclocked 8350.

It was easy to see around when the FX9590 was being binned for. New FX8320 and 8350 suddenly stopped hitting 5GHz any where near as easily.

[Dabombinable]

4 hours ago, Zodiark1593 said:

Probably for older CPUs then, or certain boards. My i5-4590 has a maximum 3.7 GHz turbo, and my board lets me lock that frequency across all cores (I can certainly verify this). From what I've read, the board has control over turbo frequencies. Maybe this is no longer the case with newer CPUs?

I can confirm that it's the same for my i5 4440 on my old Asus H87M Pro, and Gigabyte H97 HD3 (which is voltage locked-no 4.8GHz 4790K, unlike with the older Asus board).

Edit:
I just remembered, it is definitely the same for newer CPU, tech reviewers (eg. tech jesus) have done videos and articles about motherboards running Intel's CPU out of spec with default settings.

[Drak3]

20 minutes ago, leadeater said:

Nothing has changed on the die or fabing, they aren't better they just now have a new label on them.

 

The top percentile possible 9900K will still be exactly the same as the now top percentile possible 9900KS, the upper percentile hasn't changed, it can't change if there is no change.

Again, you're making a huge assumption that it's the same binning. It might not be. There is enough deviation in silicon for the potential of the 9900K to end up not being the best example of the silicon.

[leadeater]

14 minutes ago, Drak3 said:

Again, you're making a huge assumption that it's the same binning. It might not be. There is enough deviation in silicon for the potential of the 9900K to end up not being the best example of the silicon.

"As of 3/16/19, the top 8% of tested 9900Ks were able to hit 5.1GHz or greater."

 

Come back to me when 9900KS can do better than 5.1GHz.

 

There is zero change to the product, it's label change. 10 days ago the best possible 9900K was able to achieve a certain clock, when the 9900KS comes out the best possible will still be this same certain clock. Unless you're trying to make the argument a label does improve performance?

 

Edit:

Percentiles matter, not how wide it is but the performance of the upper percentile. 9900K might only be 8% and the 9900KS might be 16% but this is about performance not quantity.

[Drak3]

Just now, leadeater said:

There is zero change to the product, it's label change, 10 days ago the best possible 9900K was able to achieve a certain clock, when the 9900KS comes out the best possible will still be this same certain clock. Unless you;re trying to make the argument a label does improve performance?

I'm saying that you're working off an assumption, a speculation.

 

We don't know yet if the KS is just a binned K, or if the K and KS are at different levels of binning. Wouldn't be the first time Intel has done absurd binning, they binned the fastest core in Broadwell-E for Turbo Boost 3.0 and per core overclocking, and have pushed drivers for Windows so that single threaded workloads operated off of that core.

[leadeater]

30 minutes ago, Dabombinable said:

I just remembered, it is definitely the same for newer CPU, tech reviewers (eg. tech jesus) have done videos and articles about motherboards running Intel's CPU out of spec with default settings.

I watched that one as well and I think you're correct. From what I remember MCE does work on locked chips, you can't set the multiplier above the boost but you can force all cores to that multiplier instead of only having single core boost.

[leadeater]

15 minutes ago, Drak3 said:

I'm saying that you're working off an assumption, a speculation.

 

We don't know yet if the KS is just a binned K, or if the K and KS are at different levels of binning. Wouldn't be the first time Intel has done absurd binning, they binned the fastest core in Broadwell-E for Turbo Boost 3.0 and per core overclocking, and have pushed drivers for Windows so that single threaded workloads operated off of that core.

Please explain how the 9900KS die can be better than a 9900K of old. What exactly have they done to make it better, where's the new fab line, where is the new fab node?

 

Binning is binning, it has no relation to the manufacturing of the die only the testing to find out how good it is, binning does not make it better. That's not assumption that's reality.

 

You're the one that assumed something off what I said originally, and it's pretty obvious to me that your assumption of what I said was incorrect.

[Drak3]

1 minute ago, leadeater said:

Please explain how the 9900KS die can be better than a 9900K of old. What exactly have they done to make it better, where's the new fab line, where is the new fab node?

 

Binning is binning, it has no relation to the manufacturing of the die only the testing to find out how good it is, binning does not make it better.

Once again: The 9900K might not be the best sampling of the silicon. Intel could have very well binned the chips and held back the top x or .x% to release later.

[leadeater]

Just now, Drak3 said:

Once again: The 9900K might not be the best sampling of the silicon. Intel could have very well binned the chips and held back the top x or .x% to release later.

Well I was actually waiting for this argument, they sell way too many, produce way too many for this to be a thing for anything other than short term. There is enough existing 9900Ks out there to know the upper point already. Even if Intel started binning 6 months ago you're making the assumption that they bin every single die and none of the good ones ever made it out to market.

 

And they have to wait a long enough period of time and manufacture enough product to even know what the upper binning point actually is. Intel couldn't pre-know this information and hold back dies from day one, it actually takes time find out these percentiles in regards to binning above standard product spec.

[Taf the Ghost]

1 hour ago, leadeater said:

New standard ones, these dies were already in the existing 9900K pool so you're net gaining nothing and lose out on the 9900K pool of dies. The best new 9900KS won't overclock more than the old best 9900K.

 

1 hour ago, Drak3 said:

That makes the assumption that only now Intel is binning them this strictly. That might not be the case. It's possible that there were some chips that binned higher than exceptional 9900Ks but not enough to push them as their own pool of chips yet.

Intel has rolled out a new stepping for most of the 9000 series parts. Some with solder and some with TIM. (I saw a spreadsheet with the details.) R0 I believe. Started showing up a few weeks ago, so they should have new binning information available.

[Drak3]

1 minute ago, leadeater said:

Well I was actually waiting for this argument

Bullshit. That was the 3rd time I've made the argument.

 

1 minute ago, leadeater said:

There is enough existing 9900Ks out there to know the upper point already. Even if Intel started binning 6 months ago you're making the assumption that they bin every single die and none of the good ones ever made it out to market.

 

And they have to wait a long enough period of time and manufacture enough product to even know what the upper binning point actually is. Intel couldn't pre-know this information and hold back dies from day one, it actually takes time find out these percentiles in regards to binning above standard product spec.

9th gen product did not get sold the second the first few CPUs rolled off the manufacturing floor. Intel builds up stock first, and the time it takes to do that is enough to see if there is a potential for a higher binned product down the line.

 

No company worth its salt started selling products as soon as they start rolling out of production. They always build up stock.

 

 

And until the KS has been in the hands of overclockers long enough to have a large sample size, your assumptions are as valuable as the 16c/32t/5GHz Ryzen 3000 speculation.

[leadeater]

25 minutes ago, Drak3 said:

Bullshit. That was the 3rd time I've made the argument.

No it's the first, you never started that Intel was holding back dies from the start. But ok I've had enough. You can think what ever you like but as I said come back to me when the 9900KS does better than what the 9900K can, Silicon Lottery will be testing and supplying them so when they have 5.2Ghz bin on their site I'll concede that the 9900KS does indeed OC better.

[Drak3]

Just now, leadeater said:

No it's the first, you never started with Intel was holding back dies from the start.

 

46 minutes ago, Drak3 said:

you're making a huge assumption that it's the same binning. It might not be.

1

28 minutes ago, Drak3 said:

We don't know yet if the KS is just a binned K, or if the K and KS are at different levels of binning

2

23 minutes ago, Drak3 said:

Once again: The 9900K might not be the best sampling of the silicon. Intel could have very well binned the chips and held back the top x or .x% to release later.

3

 

3 times. The same exact statement, made 3 times: the binning might not be the same.

[leadeater]

2 minutes ago, Drak3 said:

3 times. The same exact statement, made 3 times: the binning might not be the same.

And where do you say that they held these back and didn't get on to the market. Number 3 as I said, first time.

 

Seriously you're arguing for the sake of arguing, enough.

[xg32]

it's very likely that it's just the same chip, binned, just probabilities.

 

[Drak3]

5 minutes ago, xg32 said:

it's very likely that it's just the same chip, binned, just probabilities.

 

There's no doubt that it came from the same production lines. The only question is if it is a different binning than the 9900K.

[xg32]

3 minutes ago, Drak3 said:

There's no doubt that it came from the same production lines. The only question is if it is a different binning than the 9900K.

if it's anything like the 8086k, they'll take the top 9900ks and make it the S, the 8700k overclocked like shit after the 8086k was released.

 

tdp goes up to 160-175w instead of 135w, done.

[Drak3]

1 minute ago, xg32 said:

the 8700k overclocked like shit after the 8086k was released.

From what I recall of the 8700K and 8086K, the 86 was a fairly consistent 100MHz increase over the 8700Ks prior to it, and that there wasn't much change for the 8700K.

[TahoeDust]

Unless they are doing INSANE binning, or changed something, this chip will be pulling 220w+ plus stock running Prime95 Small FFT.  

 

[leadeater]

9 minutes ago, xg32 said:

if it's anything like the 8086k, they'll take the top 9900ks and make it the S, the 8700k overclocked like shit after the 8086k was released.

 

tdp goes up to 160-175w instead of 135w, done.

Sadly the historic information is a bit lost due to the products not being sold on the Silicon Lottery site anymore but they have the stats for them, the product pages have the as of testing date which might have given us that information.

 

 

This doesn't tell us how the 8700K changed after the 8086K came out, 5.3Ghz was the top for both tough. I don't remember how SL priced them either.

[xg32]

1 minute ago, leadeater said:

Sadly the historic information is a bit lost due to the products not being sold on the Silicon Lottery site anymore but they have the stats for them, the product pages have the as of testing date which might have given us that information.

 

 

This doesn't tell us how the 8700K changed after the 8086K came out, 5.3Ghz was the top for both tough. I don't remember how SL priced them either.

there's an excel sheet out there somewhere, i recall someone posting it, though iirc the 8700k 5.3s started higher than 4%

[TrigrH]

10 minutes ago, leadeater said:

This doesn't tell us how the 8700K changed after the 8086K came out, 5.3Ghz was the top for both tough. I don't remember how SL priced them either.

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