Monday, October 8th 2018

Intel Officially Launches 9th Generation Processors Including the 8-Core / 16-Thread Core i9-9900K

Anand Srivatsa, Vice President of Intel, officially announced their all-new 9th generation of core processors in today's live stream. While the Coffee Lake refresh has certainly been no secret, a few facts were confirmed today. The Core i9-9900k will be Intel's first broad volume 5 GHz processor and is their first mainstream 8 core, 16 thread offering. In order to facilitate better overclocking results for enthusiasts, the company also confirmed that they will use solder TIM for the whole range of products, which should result in not only better overclocking potential but much lower thermals as well.
In a performance demonstration on stage, Intel showed a Core i9-9900K equipped system running two virtual machines with two games running, both streaming, at the same time. The demonstration showed not only that there was enough CPU performance to run multiple games, but it also had enough grunt left over to stream them both at the same time. While not a realistic workload by any means, it does demonstrate the multi-threading capabilities of the processor to some degree. While further data was not revealed, in Intel's own words, having tested 19 of today's games the Core i9-9900K came out on top of anything else, including AMD. This is not surprising considering the 5 GHz boost clock it ships with. However, reviews will be needed to confirm just how big a performance bump actually exists.

In regards to actual specifications, what we do know is the Core i9-9900K will be an 8-core, 16-thread processor with a base clock of 3.6 GHz and a boost clock of 5.0 GHz. The Core i7-9700K which will likely be the sweet spot for most gamers features eight cores, eight threads, due to Hyper-Threading having been disabled. Clock speeds are similar with the base clock being the same as the Core i9-9900K and the boost clock topping out at 4.9 GHz.
Finally the Intel Core i5-9600K comes in with six cores, six threads with hyper threading also disabled. The base clock sees a small jump here to 3.7 GHz while the boost clock tops out at a far more modest 4.6 GHz. All three have the same 256 KB of L2 cache per core, however, the L3 cache sizes vary between them. The Core i9-9900K will have 16 MB of L3 available, while the Core i7-9700K drops to 12 MB and last and with the least comes the Core i5-9600K with just 9 MB. All three processors feature the same 95 W TDP and 16 PCIe lanes. When it comes to the Integrated Graphics, Intel has stuck with the same GT2 configuration as seen on their previous 8th generation processors.
While Z390 boards are launching as well, Intel's Coffee Lake refresh will be supported on the older Z370 chipset. Making the latest generation a drop in upgrade, however, a BIOS update will be necessary. Those interested in purchasing a 9th generation processor wont need to wait long, as Intel also stated that pre-orders are now available with the 9th generation core processor family being made available globally on October 19th.

Update Oct 8, 22:00 UTC:
Intel has released more information as to the performance uplift enthusiasts can expect from the Core i9-9900k. They have also specified a bit more information as to the latest platforms maximum supported PCIE lanes and pricing.
  • Enabling more than 220 FPS on three of the most popular global PC game titles. (up to; 309 FPS in Rainbow Six Siege, 224 FPS in Fortnite, 440 FPS in CSGO and 221 FPS in PUBG.)
  • With Intel Turbo Boost Technology 2.0 delivering up to 5.0 GHz single-core frequency, gamers can enjoy their favorite games with up to 10 percent more FPS across popular gaming titles compared with the previous generation, as well as up to 37 percent more FPS compared to a 3-year-old PC.
  • Capture game play, transcode and stream while playing. With this type of megatasking, gamers can enjoy up to 11 percent more FPS compared with the previous generation and up to 41 percent more FPS compared with a 3-year-old PC.
  • Video editing improved on Adobe Premiere- Up to 34 percent faster video editing compared with the previous generation and up to 97 percent faster compared with a 3-year-old PC.
  • Overall system performance improved up to 15 percent better as compared with previous generation and up to 40 percent as compared with a 3-year-old PC.
  • Overclock these new processors with the Intel Extreme Tuning Utility (Intel XTU), a precision tool for advanced overclocking, and take advantage of the additional thermal headroom with solder thermal interface material (STIM).
Update Oct 9, 11:00 UTC:
We took a closer look at the performance results published by Intel (testing conducted by an external company), and noticed that many things don't add up regarding how the AMD test systems were set up, which in our opinion led to a significant performance disadvantage for the AMD rigs. More details here: www.techpowerup.com/248355/intels-9th-gen-core-gaming-benchmarks-flawed-and-misleading

Update Oct 19:
Our review of the Core i9-9900K is live now.
Add your own comment

114 Comments on Intel Officially Launches 9th Generation Processors Including the 8-Core / 16-Thread Core i9-9900K

#51
GlacierNine
notbAnyway, we've seen both Intel and AMD launching their products with a "50% better than competitor" slogan.
AMD used a synthetic benchmark for a best-case multi-thread performance, because they had a core count advantage.
Intel used a particular setup that pronounces their advantage in single-thread performance, but they used a more real-world test (games).

Let's be honest: Intel simply asked a company to make a very specific product test that will show it's strengths. It's not an objective review or scientific research. It's just marketing. It's no different from how antiperspirant makers "prove" that it works for 24h and washing powders remove 99% of stain types.

One thing I'd have to mention is the XMP issue, i.e. Ryzen results in this comparison being lower due to bad memory setup.
A situation when a casual PC user has to care about things like memory frequency and timings is just repulsive.
You're massively misrepresenting what intel did.

AMD did pick a test that advantaged it's product, yes. But they didn't intentionally hobble the Intel Hardware to exaggerate that result.

Intel did. They picked a set of tests they already knew they would win, but then they also:

1 - Used 4 sticks of memory instead of 2, to hobble the AMD platform's timings.
2 - Didn't apply XMP timings to AMD's sticks when they did to their own.
3 - Used superior cooling to AMD (A Noctua NH-14S versus a Wraith) in order to reach their maximum boost clocks, when we all know that the boost TDP of that chip (Which is not 95W, as Intel measures TDP at base clocks, not boost) would overwhelm a solution that was actually equivalent to a wraith, leading to throttling.

Additionally, Intel's performance numbers were presented as if an objective review of the product, despite being released weeks ahead of the NDA for real reviews.

This isn't the same thing as AMD deciding to only promote the results of it's best event. This is Intel making AMD run the 100m sprint uphill when Intel ran down.
Posted on Reply
#52
TheGuruStud
notbNeither of these CPUs is made "for most people".

Intel is better and more expensive. Simple as that. And the difference should be easily noticeable.

Similar on average. HT is giving around 40-50% boost.
That's why we're not getting 6C/12T anymore in desktops.

Not made for idealized benchmarking? So what is it good for? :p
No, it's supposed to be made to represent exactly what you see it doing, multimedia work. It's only now that since AMD excels at it (especially due to superior multithreading) that it's called into question. For the rest of history it's been sponsored and fine tuned for intel, so suddenly, it's now a BS benchmark. In reality, it's always been BS.
Posted on Reply
#53
GlacierNine
notbOne thing I'd have to mention is the XMP issue, i.e. Ryzen results in this comparison being lower due to bad memory setup.
A situation when a casual PC user has to care about things like memory frequency and timings is just repulsive.
I think maybe you're just too used to how easy overclocking is on Intel platforms after 10 years of it being done exactly the same way, and so many little things being introduced to make it easy - on top of the fact the ring-bus architecture (Possible due to intels monolithic dies, which are very much hitting their limits in terms of practicality, hence the wider move to MCM or chiplet designs of CPU) - so many things are done for you on Intel to get performance, from being able to simply enable XMP to being able to turn on MCE in the BIOS to overclock even a locked CPU. Auto voltages and even voltage/frequency curves in automated OC utilities. Hell, EVGA are shipping built in OC tools and stress testing in their latest BIOS.

It was never this easy in the past, and all platforms have quirks that users need to account for to obtain the best performance. Most of the time we just learn, get used to it, and move on. It becomes considered "common knowledge" after a while. Take Dual vs Single Channel Memory for example. That didn't used to be a thing, and when it became a thing it was a quirk like memory performance affecting CPU scores is a quirk. What about on X58, where we could potentially run Triple channel? That quirk came and went and was unique to that platform. I remember back in the day when people didn't understand Dual Channel. These days any idiot knows that if your RAM sticks are in adjacent slots you're running them suboptimally (Well, unless you're making a video for The Verge...)

What about Netburst architecture, where your overclock was determined by the combination of your FSB and Multiplier, and your FSB was actually really important in determining your final clocks? These days if you OC, you don't need to touch the equivalent BCLK at all, and if you do you rarely change it more than 2 or 3MHz at the most extreme. Back in the Netburst/K8 days, FSB overclocking was the norm and you had to fiddle for ages, figuring out if you could reach your next CPU speed step at a given FSB or if you had to try and reach it at a lower FSB with a higher multiplier, etc.

Having to remember to turn on XMP to get the best performance out of your Ryzen Platform is an extremely minor niggle compared to the practical nightmares that have been normal in the past in order to get a system working it's best. Especially when you remember that high end memory overclocking used to mean buying things like THIS to get more voltage through your DIMMs - www.madshrimps.be/articles/article/227/OCZ-DDR-Booster-Review/0

These days the limitation for memory overclocking is how much voltage you can put through the IMC of your CPU. Why? Because we don't have Northbridge/Southbridge in systems anymore, which is another thing that's been simplified since the "Bad old days".
Posted on Reply
#54
Melvis
notbAnd AMD or Nvidia haven't? And what about your favourite brand of washing powder?
Sometimes I do feel like some people on this forum are totally detached from reality. This is an Intel advertisement. It's no different from advertisements made by any other company. It's not fake (that's illegal), but it's usually based on a best case scenario.

Remind me, just how well Cinebench multi-thread reflects typical software run on consumer PCs? :)
Umm this is intel, do you not know how much lies and BS they have passed onto all of us over the yrs? Nvidia and AMD dont even come close! (yes they do mislead but not like Intel) to how misleading and straight out lies that Intel have done, I guess you wasnt around for the 478/939 days which intel fooled alot of people that there P4's where faster then AMD's which where not even on the same lvl. Did you not watch the video? maybe you should watch the video and listen to what Steve has to say..
Posted on Reply
#55
GlacierNine
MelvisUmm this is intel, do you not know how much lies and BS they have passed onto all of us over the yrs? Nvidia and AMD dont even come close! (yes they do mislead but not like Intel) to how misleading and straight out lies that Intel have done, I guess you wasnt around for the 478/939 days which intel fooled alot of people that there P4's where faster then AMD's which where not even on the same lvl. Did you not watch the video? maybe you should watch the video and listen to what Steve has to say..
Let's not forget that Intel are a successful company quite literally due to the fact they welched on the technology sharing agreement they originally signed with AMD.

The profit they reaped from doing that enabled them to drag the court case out for over a decade until eventually they were found guilty and fined a colossal amount of money - but in the meantime they'd made so much money that they were able to simply pay the fine and move on, when it SHOULD have crippled their business if they'd paid an amount proportionate to the deceit.

That plus their agreements with Dell, paying them off to not sell products with competitors processors in - another court case they ended up losing.


Intel's record of lies and deceit is long and well documented. AMDs? Sure, they've pulled bullshit. Nowhere near on the same scale as Intel though.
Posted on Reply
#56
the54thvoid
Intoxicated Moderator
notbA situation when a casual PC user has to care about things like memory frequency and timings is just repulsive.
A casual PC gamer won't build their own system. If you do, you're a hobbyist. If you're a hobby builder, then at least you ought to know what you're getting into.

A casual PC gamer will buy a pre-built, or have one built for them. They don't worry about timings. They'll simply enjoy gaming on a competent platform.

And that's from me, a first generation 1700x owner with 3200Mhz memory at 14 timings. And I'm barely competent. ;)
Posted on Reply
#57
TheTechGuy1337
I just recently did an old 2600k pc rebuild and honestly I'm still impressed with sandy bridge performance for today's standards. Is it the fastest? Hell no, but can I maintain above 60 fps with a medium level graphics card at 1080p and 1440p? Yep. Does a gamer require more then that? Only you as an individual can answer that question. We all have different needs, but add an ssd to sandy bridge, overclock it, and you can still have a great gaming experience dirt cheap.
Posted on Reply
#58
Unregistered
£599.99 for a mainstream processor, you've got to be joking
#59
Sasqui
Xx Tek Tip xX£599.99 for a mainstream processor, you've got to be joking
It's $488 for the CPU and a whole bunch more for the cool looking Dodecahedron box.
Posted on Reply
#60
Unregistered
SasquiIt's $488 for the CPU and a whole bunch more for the cool looking Dodecahedron box.
Pfft. Best part is the free postage, can't beat that AMD :laugh:
#61
Bluescreendeath
R-T-BSoldered TIM across the lower end parts too? Interesting...
Well, these "lower parts" costs as much as or more than the higher parts of the Intel 8000-generation and AND Ryzen 2s.
Posted on Reply
#62
trparky

What the hell Intel? Obvious paid results is obvious.
Posted on Reply
#63
Captain_Tom
LOL I find it funny that they are showing stupidly high framerates. The only way they can pretend the new chips are any better at gaming is if you use ridiculously low settings and resolutions. Yawn.
Posted on Reply
#64
GoldenX
*Obi-Wan's voice* You were supposed to lower the cost of multi-core processors and bring balance, not throw them into the $600 darkness!
Posted on Reply
#66
notb
GlacierNineYou're massively misrepresenting what intel did.
No, you are.

You still don't get this. It was a marketing exercise, not an objective review. Intel asked an external company to give them a number for a "up to $this % faster than 2700X".
It's a "cleans your teeth 62% more effective than other toothpaste" kind of thing.

Everything else is just a result of you misunderstanding what this comparison was about.
1 - Used 4 sticks of memory instead of 2, to hobble the AMD platform's timings.
There are 4 slots on that motherboard. A user may use them. [*]
2 - Didn't apply XMP timings to AMD's sticks when they did to their own.
A user may or may not use XMP. I bet most don't know what it is. [*]
3 - Used superior cooling to AMD (A Noctua NH-14S versus a Wraith) in order to reach their maximum boost clocks, when we all know that the boost TDP of that chip (Which is not 95W, as Intel measures TDP at base clocks, not boost) would overwhelm a solution that was actually equivalent to a wraith, leading to throttling.
9900K user may buy a Noctua (and, let's be honest, he most likely should) and 2700X user may keep the Wraith. It's a plausible situation.
Additionally, Intel's performance numbers were presented as if an objective review of the product, despite being released weeks ahead of the NDA for real reviews.
Just like with the toothpaste.
Problem here is just that PC people are too hooked up on quantitative tests. You need some distance. :)

Intel didn't gimp AMD performance in any way. They didn't disable boost or XFR. They used the bundled cooler (not a bad one) instead of something worse. Once again: it's a plausible situation.
Hence, the conclusion is true:
"Intel’s Core i9 9900K is up to 50% faster than AMD’s Ryzen 7 2700X in games."

[*] I repeat what I said earlier. It's appalling that you have to know so much about RAM tweaking to get competitive results from Ryzen. Not all gamers are PC geeks.

And here's an interesting twist!
As you said: utilizing 4 sticks of RAM takes away some performance of AMD CPUs. But hey... there are 4 slots, so I'd imagine some users will use all of them. Agree?
In that case aren't pretty much all CPU reviews helping AMD?
For example: good gaming performance reviews usually report an average of many runs. So if, for example, 1 in 5 real users has 4 RAM sticks instead of just 2, shouldn't we do that in 20% of runs as well? ;-)
GoldenX*Obi-Wan's voice* You were supposed to lower the cost of multi-core processors and bring balance, not throw them into the $600 darkness!
"Thank you AMD!"
Posted on Reply
#67
GlacierNine
notbProblem here is just that PC people are too hooked up on quantitative tests. You need some distance. :)
I think this line tells anyone reading this thread anything they need to know about how far you'll go to defend your delusional, insane belief that this was somehow OK.

For everyone else, I will simply *also* leave this 30 minute video of Steve from Gamer's Nexus absolutely obliterating Principled Technology's testing methodology bit by bit, as Zyll did.

Posted on Reply
#68
notb
GlacierNineI think this line tells anyone reading this thread anything they need to know about how far you'll go to defend your delusional, insane belief that this was somehow OK.
I just look at this as on the toothpaste thing. And it's OK. They wanted to have an "up to X better" figure.
Come on. CPU is just a product like everything else. Why are you so serious?

Do you have a car? Check what the manufacturer claims about fuel consumption. :)

I admit Intel made a mistake here. They're mostly concerned about business and OEM clients. But since this time they've launched a typical geek product, they should have done it appropriately.
This will hurt them, but not because of some legal issues (because, as I've said, this exercise is 100% legit and correct). But all this bad PR really takes eyes away from the product itself. 9900K is pretty great, but everyone is talking about a poor PR stunt.
Posted on Reply
#69
kajson
notbNo, you are.

You still don't get this. It was a marketing exercise, not an objective review. Intel asked an external company to give them a number for a "up to $this % faster than 2700X".
It's a "cleans your teeth 62% more effective than other toothpaste" kind of thing.

Everything else is just a result of you misunderstanding what this comparison was about.

There are 4 slots on that motherboard. A user may use them. [*]

A user may or may not use XMP. I bet most don't know what it is. [*]

9900K user may buy a Noctua (and, let's be honest, he most likely should) and 2700X user may keep the Wraith. It's a plausible situation.

Just like with the toothpaste.
Problem here is just that PC people are too hooked up on quantitative tests. You need some distance. :)

Intel didn't gimp AMD performance in any way. They didn't disable boost or XFR. They used the bundled cooler (not a bad one) instead of something worse. Once again: it's a plausible situation.
Hence, the conclusion is true:
"Intel’s Core i9 9900K is up to 50% faster than AMD’s Ryzen 7 2700X in games."

[*] I repeat what I said earlier. It's appalling that you have to know so much about RAM tweaking to get competitive results from Ryzen. Not all gamers are PC geeks.


"Thank you AMD!"
Please explain the not gimping in any way when the actual used 2700X results were way worse then the results the factchecker got by using the exact same setup.
Posted on Reply
#70
OutThereSomewhere
8 Cores, I think I'll pass. Not worth the upgrade from my i7-8700k.
Posted on Reply
#71
GoldenX
I'm with notb on this, it's freaking marketing, of course it's going to be biased, it's not like Intel is saying that there are weapons of mass destruction in the Middle East.
The real bad part is settings pre orders with no real reviews available.
And the worst of this, is that it opens AMD the option to rise prices if the performance is close enough. We got the1800x down from 500 at Ryzen's launch, to 350 IIRC, we may lose that.
Posted on Reply
#72
cadaveca
My name is Dave
GlacierNineFor everyone else, I will simply *also* leave this 30 minute video of Steve from Gamer's Nexus absolutely obliterating Principled Technology's testing methodology bit by bit, as Zyll did.

I watched the entire video, and paid close attention to everything that was said. My eyes have yet to roll back forward so I see right. That sounded like my 17-year-old son complaining because I wanted him home at 1 AM on a Saturday night. This guy really needs to work on his presentation skills as he might have some valid points, but he ends up sounding like a whining child, like I find many forums posters do from time to time. That's not how you get people to listen to what you have to say.

A better video would have been one that had a similar-spec'd system, and showed how to set it up properly, gone though the BIOS options, and then showed the difference in benchmarks, rather than standing there and yelling into the camera about how things were done wrong.


A nice black box showed up on my doorstep today...
Posted on Reply
#73
trparky
cadavecarather than standing there and yelling into the camera about how things were done wrong.
The video was meant to rip Intel a new asshole, objective achieved.
Posted on Reply
#74
GoldenX
Bad presentation, totally valid points. I can consider myself lucky if I can upgrade something every 5 years, I didn't know GTA V changes settings after hardware changes, I didn't know the difference in performance for RAM set-up at XMP would be that big with different motherboards.
Someone at Intel is giving jobs to the fired FX marketing people.
Posted on Reply
#75
notb
kajsonPlease explain the not gimping in any way when the actual used 2700X results were way worse then the results the factchecker got by using the exact same setup.
Similar hardware, but different setup. Intel-ordered comparison was using a fairly off-the-shelf AMD setup (no tweaking).
Intel's setup was optimized (using a better cooler etc).

That's how you get the "up to" thing. Seriously, I'm shocked this is so difficult to understand for people here. It's like this was the first time you've contacted a phenomenon called "marketing".

To be honest, I'm quite a fan of the approach. That's how casual PC using should look. I'll keep saying that the amount of PC knowledge one has to possess to reproduce (mostly AMD-based) results from online reviews is way to high.

And the 2 vs 4 DIMM Ryzen issue is just ridiculous.
Actually, I noticed something this morning and I wonder if AM4 "upgrade path" worshipers have thought this through as well.
If 2 years from know I decide to upgrade both my CPU and RAM, am I expected to sell the 2 dies and buy 2 larger ones instead? With that in mind, it doesn't seem such a great deal anymore.

I can't believe we're getting this in 2018.
The last time I had to care so much about BIOS settings etc was 15 years ago (coincidentally, also with AMD CPU).
cadavecaI watched the entire video, and paid close attention to everything that was said. My eyes have yet to roll back forward so I see right. That sounded like my 17-year-old son complaining because I wanted him home at 1 AM on a Saturday night. This guy really needs to work on his presentation skills as he might have some valid points, but he ends up sounding like a whining child, like I find many forums posters do from time to time. That's not how you get people to listen to what you have to say.
IMO his looks and behaviour are intentional. I think that's exactly how he attracts a certain following.
Posted on Reply
Add your own comment
Apr 26th, 2024 13:33 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts