Intel's 9th Generation Core Family - Coffee Lake (Refresh)

Juanrga, that is a whole lot of history fabrication, even for you.

Of course Intel had this and that planned - just never below $1000 until Ryzen came. So enough of the empty "Intel planned" comments.

"Zen architecture was inferior than expected" What?? By who?!

At least you admit SkylakeX was a reaction. A rushed, mess of one too. Was the same-priced 9000 series refresh a reaction to TR2 or just Intel sticking it too their hardcore loyalists?
 
Verado said:
Uhm... ryzen was available in numbers first couple of weeks of march.
8700k was a paperlaunch and werent available until december/january if I recall correctly.
Lots of people ordered and literally waited for months, and bitched about it on these forums.
Click to expand...
We were talking about hexacores. The 1600X came out in April. The 8700K came out in August.

Regardless of the actual timeline and availability, no CPU manufacturer is going to release a new line of processors in less than a year in a reactionary response to a competitor.
 
juanrga said:
No. Skylake-X was in the pipeline as replacement for Broadwell-E. You didn't pay attention to what I wrote.
Click to expand...

Specifically the 12+ core parts I was referring to. You give seem to throw in that 12nm is just rebadged 14nm, but at least there are some clock improvements. This is no more deceptive than renaming your HEDT parts from 7xxx to OVER 9000 when it is just a paste to solder change.
 
We have every reason to believe, based on Intel's trends, that an eight-core 'Ice Lake' following Skylake would have been priced similarly, perhaps with a small bump early on, but certainly in the <US$500 range that they've kept the top consumer socket parts in for the last decade.

That's different from their trend in the PIII / PIV era, and different from what AMD did the last time they were competitive in the CPU space.

If AMD had a clear advantage in performance, we very well might be looking at >US$1,000 sixteen-core consumer socket Ryzen CPUs and ~US$500 hexacores.
 
Nightfire said:
You give seem to throw in that 12nm is just rebadged 14nm, but at least there are some clock improvements. This is no more deceptive than renaming your HEDT parts from 7xxx to OVER 9000 when it is just a paste to solder change.
Click to expand...

12nm is a relabel of 14nm+. The plus means it gets some extra MHz on top, of 14nm. Relabeling it as 12nm, when all the metrics are identical is deceptive

7c02bb7b9df96f40c1ffae935087eead5ff59a804c563082d37c4f040bd6d9c4.png


In the other hand going from 7000 series brand to 9000 series brand is not deceptive, because the name of the series can be anything. AMD named 2000 series to the Zen APUs and I don't recall you complained then.
 
You guys see the latest from Intel? 10%-15% price reduction on their CPUs. Not sure when that happens.
 
Armenius said:
We were talking about hexacores. The 1600X came out in April. The 8700K came out in August.

Regardless of the actual timeline and availability, no CPU manufacturer is going to release a new line of processors in less than a year in a reactionary response to a competitor.
Click to expand...

It's a romantic thought to think Intel was forced to make a six core after seeing the acceptance of Zen... but that's not even possible. Simple stepping changes take months to design, validate, and produce. A new chip with more cores? Completely out of the question.

Now of course AMD is absolutely applying pressure to Intel. However, in the 8700k time-frame that would result in pressure to pricing, not the presence of the chip at all.
 
PhaseNoise said:
It's a romantic thought to think Intel was forced to make a six core after seeing the acceptance of Zen... but that's not even possible. Simple stepping changes take months to design, validate, and produce. A new chip with more cores? Completely out of the question.

Now of course AMD is absolutely applying pressure to Intel. However, in the 8700k time-frame that would result in pressure to pricing, not the presence of the chip at all.
Click to expand...

The only thing I can say is that I'm sure that AMD pressured the Intel timeline for release for both the 8700k and the HEDT X299 platform. There wasn't widespread availability until January the next year even if there was a "release" in August (I think I remember more of a "preview" with an October release).
 
kirbyrj said:
The only thing I can say is that I'm sure that AMD pressured the Intel timeline for release for both the 8700k and the HEDT X299 platform. There wasn't widespread availability until January the next year even if there was a "release" in August (I think I remember more of a "preview" with an October release).
Click to expand...

Very possible, but I would just indicate they must have had designs already in progress for any pressure to be useful. It could not have been a flat-footed reactionary response. They can accelerate (to some degree - this is really not a simple thing to speed up, you need massive validation suites and production statistics), and you can price differently. But you can't just introduce something new as has been at least somewhat implied. Apologies if I'm misinterpreting assertions / implications made here.

In the longer term - you bet your bippy AMD's pressure is affecting roadmaps and pricing very significantly. We're back into a good period of competition.
 
juanrga said:
In the other hand going from 7000 series brand to 9000 series brand is not deceptive, because the name of the series can be anything. AMD named 2000 series to the Zen APUs and I don't recall you complained then.
Click to expand...

Solid rebuttle, though I did complain about it, with the 3400g being an even bigger sin. (My 2400g actually run clocks and memory more akin to 2nd gen, perhaps just luck)
 
I have zero issues with the performance of the 9900k. I am not sure why people are after “more” for day to day use right now.
 
Keljian said:
I have zero issues with the performance of the 9900k. I am not sure why people are after “more” for day to day use right now.
Click to expand...

People said the same thing about the 2600k for the past 8 years...

I think it has more to do with people who will be upgrading in the future rather than pushing another upgrade on people who are already on a 9th gen part. If the 3700X/3800X comes in at near the same performance as the 9900K and costs less, people are going to consider it. Realistically, people who are on an 8700k or better don't NEED to upgrade to anything for a little while yet.
 
The 9900KS was not the last of the 9th gens. We now have the Xeons to add to the fray with up to 16 threads:
https://www.guru3d.com/news-story/i...ssors-based-on-9th-generation-core-chips.html

The new e2200 series look to match their desktop siblings in speed (at least base and turbos, though doubtful they have MCE) with added features available only to Xeon.

Price wise, the 8/16 parts seems to ask for a nice premium ($500-$540), though the 6/12 and Xeon-exclussice 4/8 parts offer a rather good value.

The 2246g is a 6/12 part for $311 running respectable clocks.
The 2244g is a 4/8 part for $270 with nice clocks as well.
 
More details from anandtech on boost freqs depending on cores. They will be reviewing these soon.
https://www.anandtech.com/show/1505...-2200-series-for-servers-8-cores-up-to-50-ghz

Nerdtastic info pulled from the Intel data sheet showing a 28 second tau ie the 2276g could run 4.6 ghz for up to 28 seconds. Since it is an 80w class 6 core, it would be running at 112w, which is very respectable. 8086g-like
Screenshot_20191104-225249_Drive.jpg


The 210w PL2 value of the 8 core cpus doesn't look so great, though.

Ambient OC_575px.png


9900K OC_575px.png
 
drescherjm said:
A 8700k has 12 fast threads already. It will be years before this becomes a serious bottleneck for most multithreading applications.
Click to expand...

Would be great if everyone ran just one app. That's just not the case.
 
DooKey said:
Pretty much everyone running an OC'd 8700K is gaming and not doing anything else.
Click to expand...

Move the goal posts and so what. Please, Tell me more about your gaming intel example and nothing else. Because that's all we get as a counter. Ever. Please continue on your 10fps games only sidetrack.
 
Pretty much everyone running an OC'd 8700K is gaming and not doing anything else.
Click to expand...

Tell that to the people who purchased business desktops from Dell.

For us Dell is the approved vendor. And we are only allowed to purchase a select list of products (which does not Include anything at all good from AMD).

Although thankfully there has been added a loophole to this and I am looking on replacing on my work desktop with a Ryzen 3600 or 3700 in the $1000 US price range. I can't just build a PC because we are not permitted to purchase windows and software licensing does not have it for sale only an upgrade which requires an existing licence.
 
Last edited:
DooKey said:
Pretty much everyone running an OC'd 8700K is gaming and not doing anything else.
Click to expand...

I mean, the reason to build a system around an overclocked 8700K today would be for gaming, but the system would be great at any other desktop stuff too.

[sauce: use an overclocked 8700K for gaming and other stuff]
 
drescherjm said:
A 8700k has 12 fast threads already. It will be years before this becomes a serious bottleneck for most multithreading applications.
Click to expand...

Each logical thread is only 30% the speed of a physical thread.
It does reduce stuttering on programs that actually tries to use more threads than the # of physical cores you have though.
 
Falkentyne said:
Each logical thread is only 30% the speed of a physical thread.
Click to expand...

...on average.

It can be 100%. It can also be -20%, if there is a high contention for already employed resources.

Since game threads tend to be different in composition to background threads, SMT tends to work out pretty well when it comes to reducing frametimes. Not always, but almost always.
 
The numbers don't roll on vR20 as compared to v15 so I'm deleting.
 

Attachments

  • H8gOA.png
    H8gOA.png
    8.5 KB · Views: 0
  • WhHYH.png
    WhHYH.png
    8.8 KB · Views: 0
  • 250817_H8gOA.png
    250817_H8gOA.png
    21.4 KB · Views: 42
  • 250823_WhHYH.png
    250823_WhHYH.png
    21.5 KB · Views: 49
Last edited:
Mega6 said:
Two Logical Threads > one physical core. Otherwise, Hyperthreading would be a waste in every scenario.
Click to expand...
That actually is what he's saying. 100% for the 'physical' core and then another 30% for the 'logical' core, so 130% core performance with SMT2 versus no SMT.
Again, the only problem with that is that it is much, much more complicated in reality, and can vary significantly depending on the exact hardware and the software being executed.
 
The numbers don't roll on vR20 as compared to v15 so I'm deleting.
 
Last edited:
Mega6 said:
The example shows you get 90% x 2 or 180% with two logical cores. >> 130%

Probably a best case scenario for Threading given the app. Never the less, more accurate and based on actual data. A 130% with two logical cores doesn't seem very realistic.
Click to expand...

I'm sorry but your math is very wrong.
Can you explain how, in Cinebench R15, which scales perfectly on threads, a 9900K with hyperthreading disabled (apples to apples since L3 cache is higher than on 9700k) gets about 1750 CB (hyperthreading disabled) at 5 ghz, while a 9900K HT enabled at 5 ghz gets 2240 CB? (All OS security mitigations disabled for both).

I see a 30% increase right there.
What's the percentage increase of 1750 to 2240 ?

Your math is incorrect, I'm sorry.

If you still don't believe me, disable hyperthreading on your 9900K and check for yourself, then enable it and check these scores.

BTW a 9700K gets about a 3% higher CB Score than an 8700K at same frequency.
 
30% has been a good estimate of SMT performance for a long time (aggregate number). Yes, some hw and apps will push the values around, but if you're looking for napkin math - this is a good number to use.
 
The numbers don't roll on vR20 as compared to v15 so I'm deleting.
 
Last edited:
PhaseNoise said:
30% has been a good estimate of SMT performance for a long time (aggregate number). Yes, some hw and apps will push the values around, but if you're looking for napkin math - this is a good number to use.
Click to expand...

And 6 cores with 12 threads = 8 physical cores essentially using that math. Which gets back to the point that upgrading from an 8700k to a 9700k is questionable at best. Unless you really do work where the 30% is lower then you'd want more cores. And conversely if you do work that is highly threaded you would want more threads. Either way I don't see the 8700K as being an irrelevant chip anytime soon.
 
kirbyrj said:
Which gets back to the point that upgrading from an 8700k to a 9700k is questionable at best.
Click to expand...

This would never have been a good idea, though I'm sure a few buyers tossed between the options when starting new. I can't say that I'd easily be able to pick one over the other myself.
 
PhaseNoise said:
30% has been a good estimate of SMT performance for a long time (aggregate number). Yes, some hw and apps will push the values around, but if you're looking for napkin math - this is a good number to use.
Click to expand...

Seems about right eg an 8700k = (6×1)+(6×0.3) or 6+2 which is why it is similar to the 9700k in many benches.
 
kirbyrj said:
And 6 cores with 12 threads = 8 physical cores essentially using that math. Which gets back to the point that upgrading from an 8700k to a 9700k is questionable at best. Unless you really do work where the 30% is lower then you'd want more cores. And conversely if you do work that is highly threaded you would want more threads. Either way I don't see the 8700K as being an irrelevant chip anytime soon.
Click to expand...

Nor I. It’s a fast chip.

I think many dramatically overestimate the average saturation of the cpu, and how parallel their normal daily load really is.

Now if someone just wants to own a beast or do truly render/encode all day (or time=money), that’s cool, buy the latest beast. But the 8700 will not be a poor performer for many years.
 
Yep, I would rather have 6 fast cores (r5 3600 / 8700k) than 8 medium speed cores (r7 1700) even if the later gets slightly higher CB scores. That will not change in 5 years, gaming or otherwise.
 
Agreed. I sold my 5960X for a 8700K. I did like having one of the first 8-core consumer chips, but I found the faster 6-core to be better for everything I do.
 
Nightfire said:
These prices seem a bit ... obnoxious. Intel looks to be charging around $33 for their iGPU.
Click to expand...

On the low end, they look marginally price competitive; on the high end, they do offer a small single-thread advantage that people seem to be willing to pay for, and since AMD refuses to offer an IGP throughout their lineup, some of that difference is justifiable.

Not all of it of course, but it is Intel who can still trade on their brand with the broader public.
 
IdiotInCharge said:
On the low end, they look marginally price competitive; on the high end, they do offer a small single-thread advantage that people seem to be willing to pay for, and since AMD refuses to offer an IGP throughout their lineup, some of that difference is justifiable.

Not all of it of course, but it is Intel who can still trade on their brand with the broader public.
Click to expand...

If Zen 3 comes in and removes that single thread advantage over 4th Gen Skylake then what? A price cut? Let's face it. The IGP customers on the high end are few and far between. If it is more than 2-3% of overall sales, I'd be surprised honestly. I agree with the trading on brand though, people recognize the brand, the Intel Jingle, and buy accordingly.
 
On the high end, the IGP isn't make or break obviously; but for many / most in terms of retail, a discrete GPU simply isn't needed or perceived to be, while more CPU may, and Intel is it there until AMD gives their APUs more cores.

As for single-thread performance, if AMD can finally close the gap that Intel created with the release of Core 2 all those years ago, then great!
 
IdiotInCharge said:
On the high end, the IGP isn't make or break obviously; but for many / most in terms of retail, a discrete GPU simply isn't needed or perceived to be, while more CPU may, and Intel is it there until AMD gives their APUs more cores.

As for single-thread performance, if AMD can finally close the gap that Intel created with the release of Core 2 all those years ago, then great!
Click to expand...

They don't need to close the gap they have to create their own gap, what intel did with all their lakes and bridges the last decade was according to everyone meh, not worth upgrading too, 5% here, 7% there that's peanuts, now AMD is going up in single digits and it's like the best thing ever.

The only thing they got going is the tons of cores which for the gamers among us is pretty meh as they are unused or clock down the CPU so far it's a wash.
 
You must log in or register to reply here.
Back
Top