Core 9000 series

[mmj_uk]

I'd take 8c/8t over 6c/12t every time.

Perhaps dropping Hyperthreading is a response to ongoing security concerns?
https://www.theregister.co.uk/2018/06/20/openbsd_disables_intels_hyperthreading/

As detailed in this mailing list post, OpenBSD maintainer Mark Kettenis wrote that “SMT (Simultaneous Multi Threading) implementations typically share TLBs and L1 caches between threads.

“This can make cache timing attacks a lot easier and we strongly suspect that this will make several Spectre-class bugs exploitable.”

 

[Panos]

So, the normal question - which would be better - i7-8700k 6-cores/12-threads or i7-9700k 8-cores/8-threads?

I believe 8/8. The HT is consuming the left over perf of the cores, and in the case of Intel isn't as efficient as the AMD solution.

 

[4K8KW10]

I believe 8/8. The HT is consuming the left over perf of the cores, and in the case of Intel isn't as efficient as the AMD solution.

It is very likely the 12 threads will be equal or faster in well-done multi-threaded applications. In lightly threaded applications, it should be the same performance plus-minus the frequency difference.

I base this observation on the fact that the 8700k shows 43% higher multi-threaded performance than a 8600k.
So, 8600k + 2 cores?
http://cpu.userbenchmark.com/Compare/Intel-Core-i7-8700K-vs-Intel-Core-i5-8600K/3937vs3941

 

[Panos]

It is very likely the 12 threads will be equal or faster in well-done multi-threaded applications. In lightly threaded applications, it should be the same performance plus-minus the frequency difference.

I base this observation on the fact that the 8700k shows 43% higher multi-threaded performance than a 8600k.
So, 8600k + 2 cores?
http://cpu.userbenchmark.com/Compare/Intel-Core-i7-8700K-vs-Intel-Core-i5-8600K/3937vs3941

If I wanted multithread perf I would buy a Threadripper imho. For gaming and basic stuff they are both the same with or without HT.
Already contemplating the idea of a 2950X tbh or at least a 2920X.

 

[IT Troll]

So, the normal question - which would be better - i7-8700k 6-cores/12-threads or i7-9700k 8-cores/8-threads?

The performance increase offered by hyperthreading varies greatly depending on the application. At the extremes it can range can range between 0 - 33%, but 5 - 20% is more typical. This should mean that when averaged across a range of software, 8 cores will beat 12 threads. In highly threaded applications performance could be pretty close and will depend on final clock speeds and ability to hold boost.

 

[DragonQ]

So, the normal question - which would be better - i7-8700k 6-cores/12-threads or i7-9700k 8-cores/8-threads?

Depends what you're doing. Anything that uses maybe 1-10 threads (i.e. most cases) will likely be faster on the 9700K, whereas something maxing out the chip (e.g. video encoding) may well be faster on the 8700K. It might be the first time that peak performance actually drops on a new product since the Core i series began.

IS this basically intel saying they cannot make proper silicon cope with the heat and power demands of a SMT core at the really high clock speeds they now have to eek out of a crap process tweaked to within an inch of its life?

Just the thought of doing this a couple of years back would have had you laughed out of the office at Intel.

That's a good point, SMT does definitely add to the maximum thermal output of the CPU so disabling it for even moar MHz would make sense.

 

[X-Mikuru_Beam-X]

If this i9 9900K turns out to be soildered and a decent clocker I may finally be tempted to drop my 4770K....

 

[X-Mikuru_Beam-X]

the i7 9900k apears to be soldered! thats good news!.

Yes i did write i7 9900k. Because thats what it is.

It is whatever intel decided to call it. The marketing department made up the i7 label so they can make up the i9 label should they so choose..

 

[Perfect_Chaos]

If this i9 9900K turns out to be soildered and a decent clocker I may finally be tempted to drop my 4770K....

With the way i hear many people pronounce it in videos I'm pretty sure they will change the spelling to sordered.

 

[X-Mikuru_Beam-X]

With the way i hear many people pronounce it in videos I'm pretty sure they will change the spelling to sordered.

A little piece of me dies when the yanks on you tube pronounce it as “soder” or something to that effect

P.s I spelt it wrong myself

 

[LePhuronn]

A little piece of me dies when the yanks on you tube pronounce it as “soder” or something to that effect

"sawder". It drives me up the wall. It's not a quirk of language or a tweak in pronunciation over time, it's flat-out wrong and ignorant. There is no conceivable way "sol" can be pronounced "saw"

 

[X-Mikuru_Beam-X]

It's not a quirk of language or a tweak in pronunciation over time, it's flat-out wrong and ignorant.

Yes and what’s worse they “could care less” about how to pronounce Solder the correct way

 

[CAT-THE-FIFTH]

Its interesting that AMD dropped a little tidbit today - 7NM Epyc is sampling in 2018 for introduction next year. This means 7NM Ryzen is probably sampling this year too.

So does that mean 7NM Ryzen 3 against 14NM CFL MK2?? Or is Intel going to be able to release a 10NM CPU in time??

 

[LePhuronn]

...they “could care less”...

Don't. Please don't. You'll make me cry.

 

[pastymuncher]

So the latest rumour suggests that the 9900k and the 9700k will be soldered while all the rest have crappy thermal paste. They could at least extend the soldering to all of the unlocked K series cpu's!! It makes perfect sense as they charge extra for a unlocked cpu but of course they won't do the right thing.

 

[smilingcrow]

Its interesting that AMD dropped a little tidbit today - 7NM Epyc is sampling in 2018 for introduction next year. This means 7NM Ryzen is probably sampling this year too.

I think they announced that they are moving EPYC to TSMC for 7nm so all bets are off with regard to Ryzen 3 as that may be with GF in which case the two are unrelated.

 

[X-Mikuru_Beam-X]

Don't. Please don't. You'll make me cry.

I feel your pain

Back on topic, I’m pretty excited for the 9900K it could be what I’ve been hanging out for. So pleased I didn’t jump on that 8086k malarkey

 

[Panos]

"sawder". It drives me up the wall. It's not a quirk of language or a tweak in pronunciation over time, it's flat-out wrong and ignorant. There is no conceivable way "sol" can be pronounced "saw"

Like router and route has become raouter and raout. Is cringing.
Especially when is new thing of last 10-12 years.

 

[Panos]

Its interesting that AMD dropped a little tidbit today - 7NM Epyc is sampling in 2018 for introduction next year. This means 7NM Ryzen is probably sampling this year too.

So does that mean 7NM Ryzen 3 against 14NM CFL MK2?? Or is Intel going to be able to release a 10NM CPU in time??

10nm Intel is a troublesome process.
Their only 10nm CPU they made is a dual core with a non working IGP, whilst still the IGP is in the chip.
If we were going to see 10nm CPUs from Intel next year, they should have started producing their server chips already, and not spend money to improve 14nm to 14nm++.

 

[CAT-THE-FIFTH]

So the latest rumour suggests that the 9900k and the 9700k will be soldered while all the rest have crappy thermal paste. They could at least extend the soldering to all of the unlocked K series cpu's!! It makes perfect sense as they charge extra for a unlocked cpu but of course they won't do the right thing.

It will be hilarious if that happens and I hope it will be. Remember the 1001 reasons on forums Intel couldn't use solder,which continued even after AMD used it on a finfet process. People are even defending the thermal paste even now!!

I think they announced that they are moving EPYC to TSMC for 7nm so all bets are off with regard to Ryzen 3 as that may be with GF in which case the two are unrelated.

Apparently the GF process has been designed in such a way transitioning between the processes is less difficult:

https://www.anandtech.com/show/12831/globalfoundries-gives-7-nm-capacity-update-mulls-skipping-5-nm

According to Mr. Patton, since the Fab 8 produces leading edge chips for all of the company’s clients, its 7LP capacity may be limited, which is why AMD may not get enough capacity. The good news is that GF’s 1st Gen 7LP is similar to TSMC’s CLN7FF (which also does not use EUV) and therefore clients with significant demand will be able to port their designs from one foundry to another with some ease. Under the wafer supply agreement signed in 2016, AMD has rights to use foundries other than GlobalFoundries in certain situations. Though whether this capacity-constrained situation actually comes to pass, and whether or not AMD and other partners of GlobalFoundries will have enough time to develop all the designs they need for both foundries should it occur, remains to be seen.

I posted the TSMC article in another thread. So that means AMD already has had samples on GF 7NM too(primary partner),but apparently GF is capacity constrained,so AMD is hedging its bets with both foundries.

So in my viewpoint,I suspect Ryzen 3 won't hit 5GHZ,but maybe there will be smaller increases,as something has to give to make it work on two nodes,but the extra density can be used to save power,add moar cores,improve cache latencies,core IPC and the memory controller.

Also,unless GF/TSMC 7NM nodes hit some issues or Intel manages to get 10NM working quicker,we might see AMD possibly get to 7NM for desktop products a bit quicker than Intel. Its only happened once before for AMD to get to a new node quicker!

10nm Intel is a troublesome process.
Their only 10nm CPU they made is a dual core with a non working IGP, whilst still the IGP is in the chip.
If we were going to see 10nm CPUs from Intel next year, they should have started producing their server chips already, and not spend money to improve 14nm to 14nm++.

Yeah,according to the Intel CEO in June they said 14NM would be for "leadership products" for the next 12 to 18 months.

Next year is going to be interesting!!