Confirmed - i9 9900k will have soldered IHS, no more toothpaste TIM

[epsilon84]

Sarcasm? If not AMD indirectly dictates Intels pricing by nothing more than setting their pricing.

9900k seems expensive until you look at pre Zen pricing. Although they set the price I'm sure Intel isn't happy with it. They got Zen'd!

Not at all, if Intel pricing was dictated by Zen we would have sub $400 9900Ks and then people will need something else to complain about other than the price

If anything, the top end 'K' SKUs have increased in price the past couple of years. Look at the 7700K 8700K 9900K prices, although technically the 9900K is a 'new' tier aka i9 but that's all marketing anyway

 

[skaertus]

Not at all, if Intel pricing was dictated by Zen we would have sub $400 9900Ks and then people will need something else to complain about other than the price

If anything, the top end 'K' SKUs have increased in price the past couple of years. Look at the 7700K 8700K 9900K prices, although technically the 9900K is a 'new' tier aka i9 but that's all marketing anyway

Intel’s prices are not dictated, but they are somehow influenced by AMD. The I7-9700K will cost in the same range than the i7-8700K, but the specific configuration of such model may vary based on the competition.

Plus, this time I am not sure how much of Intel pricing was driven by competition. Intel is struggling to produce enough 14nm chips, so demand is higher than supply. Which means that Intel can manage to charge higher prices and vendors may still get a premium. Once the situation is normal again, Intel will feel more compelled to compete in price.

 

[AtenRa]

Proof of your last statement? At 95 watts, you may not get 4.6 ghz all core, but I also highly doubt it will be limited to "base clocks".

From Intel

Example, Core i7 8700K, just click the question mark next to the TDP.

https://ark.intel.com/products/126684/Intel-Core-i7-8700K-Processor-12M-Cache-up-to-4-70-GHz-

TDP
Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload. Refer to Datasheet for thermal solution requirements.

Also from the above link, In the Package Specifications section ,
In the Thermal Solution Specification, Intel reference heat-sink for the Core i7 8700K (95W TDP) is the PCG 2015D which is rated at 130W.
This heat-sink is recommended in order for the CPU to be able to reach the advertised Turbo clocks.

 

[AtenRa]

I wish it was a bit cheaper (was hoping for $450) but for those saying it should be $300, you gotta be kidding right?

If I had a product that was a clear 25% (or more) better than the competition I surely wouldnt price it at the same level unless I was desperate for marketshare.

Plus we are talking about Intel here, they don't let AMD dictate their prices.

For the last 5 years Intel CPUs had more than 25% higher performance than the competition and yet Intel never increased the Mainstream CPU prices above $350-380.
Today with fierce competition at both the mainstream and HEAD platform here comes Intel launching a Core i7 without HyperThreading (first time ever) and also increasing the top mainstream SKU price to $488.

 

[frozentundra123456]

From Intel

Example, Core i7 8700K, just click the question mark next to the TDP.

https://ark.intel.com/products/126684/Intel-Core-i7-8700K-Processor-12M-Cache-up-to-4-70-GHz-



Also from the above link, In the Package Specifications section ,
In the Thermal Solution Specification, Intel reference heat-sink for the Core i7 8700K (95W TDP) is the PCG 2015D which is rated at 130W.
This heat-sink is recommended in order for the CPU to be able to reach the advertised Turbo clocks.

So you are saying for instance the i5-8400 with a base clock of 2.8 ghz will only run 2.8 ghz at all core turbo? Seriously?
Just look at the turbo frequencies from anandtech linked in post 547. 9900k for instance has a base frequency of 3.6 and an all core turbo of 4.7.

 

[AtenRa]

So you are saying for instance the i5-8400 with a base clock of 2.8 ghz will only run 2.8 ghz at all core turbo? Seriously?
Just look at the turbo frequencies from anandtech linked in post 547. 9900k for instance has a base frequency of 3.6 and an all core turbo of 4.7.

The Core i5 8400 at 65W TDP yes it will only run at 2.8GHz all cores for an infinity amount of time. (PL1/PsysPL1)
All-Core turbo of the Core i5 8400 can reach up to 3.8GHz (PL2/PsysPL2) for short periods of time as described by the Intel Package Power Control and Turbo Boost 2.0

https://www.intel.com/content/www/u...core/8th-gen-core-family-datasheet-vol-1.html

 

[coercitiv]

9900k for instance has a base frequency of 3.6 and an all core turbo of 4.7.

It will definitely not stay at base clocks for 95W package power in non-AVX loads.

The reference for power consumption should be Coffee Lake 6c/12t @ 4.3Ghz = 95W in a heavy, non-AVX load. That equals 125W @ 4.3Ghz if considering the jump to eight cores on the same process. It will obviously be better than that, since the process is improved and the chips are likely more aggressively binned (at that price they better be). My ballpark estimate would be 4Ghz.

 

[arandomguy]

As with Nvidia's Turing release it seems the conflict is due to individual expectations of how much tech you should get versus time. Particularly in this case Intel offered a huge jump with 7xxx to 8xxx series. 7700k -> 8700k had barely any price increase for 50% more CPU resources and a mere 100mhz stock clock drop but higher OCed clocks. So in terms of the CPU consumers got something like around 40% more tech (conservatively) for the money. 8xxx to 9xxxx by comparison is a huge disappointment.

We'll have to see where the 8700k to 9700k actually lands in that respect. 6700k to 7700k was 10% while 1700x to 2700x was about 15% (or more in some scenarios) for the dollar.

I'm also guessing some folks are predicting (or hoping at any rate) that Zen 2 will bring 30% more for the dollar.

For me personally what I hoped for in terms of plausible expectations and would have preferred was a i7-9700k at 8c/16t at $399 (still price increase) without the clock speed boost. Will be interesting to see where i7-9700 lands in terms of clocks and pricing.

9700K really has no 8th gen counterpart. The 8700K might keep up in some multi thread benches, but I kind of doubt it. I think overall the true 8 cores will probably stay ahead of 6 cores plus HT.
Also the 9000 series have some hardware fixes for meltdown and spectre.

My one issue at the moment with the 9700k is that it isn't an actual universal improvement due to 12 threads vs 8 threads. While in general the 9700k will outperform the 8700k (even in fully MT usage in terms of total throughput) the lack in threads may rear it's head in terms of latency/consistency issues in some cases due to how thread scheduling works if you do fully saturate all the threads.

For example in a gaming + streaming encoding scenario this could result in frame drops for encoding or frame time spikes even if the avg FPS is higher. The 8700k vs 8c/16t Ryzens had this issue. Even though the latter may have been slower in terms of avg throughput the extra threads hid those "load spikes" in heavy multitasking scenarios in which latency is also a factor.

Albeit the above could be addressed via manual tuning in terms of thread prioritization but that's an inconvenience.

Of course we'll want to see power numbers as well to see how it compares to the 8700k.

 

[IntelUser2000]

The Core i5 8400 at 65W TDP yes it will only run at 2.8GHz all cores for an infinity amount of time. (PL1/PsysPL1)
All-Core turbo of the Core i5 8400 can reach up to 3.8GHz (PL2/PsysPL2) for short periods of time as described by the Intel Package Power Control and Turbo Boost 2.0

That depends heavily on the application. If you are running Intel Linpack where it uses the AVX unit, you may be right. If you are running non-AVX, you can run at 3.8GHz for pretty much most of the time.

 

[AtenRa]

That depends heavily on the application. If you are running Intel Linpack where it uses the AVX unit, you may be right. If you are running non-AVX, you can run at 3.8GHz for pretty much most of the time.

Nope, with a 65W TDP Thermal Solution (cooler) you will only have 3.8GHz all core turbo for UP TO 100s as described in figure 5.1 above.
Increasing the TDP to 95W TDP will allow you to have an all core turbo (or base) at up to 3.7GHz as described in table 5.5 above. (for a 6-core SKU).

 

[IntelUser2000]

Nope, with a 65W TDP Thermal Solution (cooler) you will only have 3.8GHz all core turbo for UP TO 100s as described in figure 5.1 above.
Increasing the TDP to 95W TDP will allow you to have an all core turbo (or base) at up to 3.7GHz as described in table 5.5 above. (for a 6-core SKU).

Why don't you go buy the 8400 and try it yourself? Many applications are not demanding enough that it can do 3.8GHz indefinitely even with 65W TDP.
https://www.notebookcheck.net/HP-El...op-Review.274722.0.html#toc-energy-management

A 7820HQ with PL1 set at only 30W is running Unigine Valley stress for an hour at 3.4GHz, and the GPU at 1050MHz.

i5 7500 is running at 3.6GHz with consumption only at 32W in Prime95: https://www.notebookcheck.net/Lenov...op-Review.248294.0.html#toc-energy-management

A 35W setting is enough to keep Skylake i5 mobile CPU at 2.7GHz and integrated HD 630 graphics at 900MHz running Furmark and Prime95 simultaneously.
https://www.notebookcheck.net/Lenov...ok-Review.160923.0.html#toc-energy-management

Hades Canyon NUC shows 65W TDP is enough to run Prime 95 at 3.8-3.9GHz, and Prime 95 + Furmark with CPU at 3.5-3.8GHz, and the RX Vega M GH GPU at 1190MHz.
https://www.notebookcheck.net/Intel...agement-more-demanding-than-a-gtx-1060-laptop

 

[coercitiv]

Why don't you go buy the 8400 and try it yourself? Many applications are not demanding enough that it can do 3.8GHz indefinitely even with 65W TDP.

Let's put this thing to rest: I disabled HT on my 8700 and limited power draw to PL2 = PL1 = 65W.

The result was sustained 3600Mhz while in Prime 95 v293 with Small FFTs. Even with lower silicon quality the 8400 won't have much trouble sustaining 3800Mhz in non AVX loads. For reference, moving from Prime 95 to Cinebech R15 is probably a 25% power drop.

Spoiler: screenshot

 

[coercitiv]

Hardware Unboxed is a bit annoyed by Intel's independent performance testing.

 

[Abwx]

Why don't you go buy the 8400 and try it yourself? Many applications are not demanding enough that it can do 3.8GHz indefinitely even with 65W TDP.

The "65W" 8700 consume as much as a 8700K, the allegedly 35W 8700T is measured at roughly 65W at stock and in Cinebench, at some point you have to realise that Intel s definition of TDP is extremely elastic..

8700 TDP is here :

https://www.computerbase.de/2017-11/intel-core-i7-8700-i5-8600k-test-auto-oc-ddr4/3/

And for the 8700T :

https://www.computerbase.de/2018-06...ition-cpu-test/2/#abschnitt_leistungsaufnahme

 

[The Stilt]

Hardware Unboxed is a bit annoyed by Intel's independent performance testing.

Not a surprise really, as to me it seems HWUB has become pretty openly a pro AMD site, more or less recently.
Basically they complained that Ryzen's memory was not overclocked nor the timings weren't optimized manually, and stock speeds and timings were used instead.
All of the systems were running at the maximum officially supported memory speeds and according to HWUB that equals to gimping AMD and favoring Intel.
Despite the fact that Intels were running at lower memory speeds (2666MHz vs 2933MHz).

 

[DrMrLordX]

By branding Skylake-X Refresh as 9xxx I think it confirms that Cascade Lake-X has been cancelled.

Really? When is Cascade Lake supposed to hit the Xeon market?

Yea, I agree. What is intel supposed to do, roll over and die because Zen 2 is coming out? They will certainly have the mainstream desktop lead in both single and multi-threaded performance until Zen 2 comes out, but I guess that is not enough somehow. And I am waiting for real benchmarks to see if Zen 2 matches or beats the 9900k, in contrast to the premature assumptions from the AMD camp.

Uh, well, see, the problem is that Intel . . . ah who am I kidding, this isn't an AMD thread anyway.

What I do know is that this is the faster chip Intel will release for consumer desktop for a looooong time, unless you expect a 9950k-like refresh between now and wherener the hell IceLake hits the consumer desktop market. Which could be late 2019, or even 2020. So um good luck Intel! Hope things work out for you folks.

Correct, that was another big fail for mainstream.

No it wasn't. The 1800x was competing with the 6800k on HEDT and matching/beating its performance for half the price. Never mind that an OCed 1700 could do the same . . .

Regardless, within the context of March 2017, it was a resounding success at $499. Now it looks silly, and AMD has lowered prices accordingly.

I love the aggressive Turbo Bins. That TSMC process better be as good as touted.

I would not expect anything more than 4.5-4.6 GHz out of that TSMC process.​

488 bucks for the 9900k. Dayum, that's not bad.

Yeah, about that . . .

Ouch - £600 for the 9900k.

https://www.overclockers.co.uk/inte...ocket-lga1151-processor-retail-cp-65j-in.html

Eep.

Looks like Newegg's got the auto-gouger going... $579 for the 9900K (albeit OOS), $419 for the 9700K and $279 for the 9600K. Amazon had the 8600K last week for $239.

$579 AND sold out. Good luck getting one folks.

Hopefully prices settle down in a couple weeks.

I'm predicting a month as best-case before you can get one for $488. And even then, that's a bit much. Worst-case, we may not see supplies normalize until December or later.

 

[LTC8K6]

The "65W" 8700 consume as much as a 8700K, the allegedly 35W 8700T is measured at roughly 65W at stock and in Cinebench, at some point you have to realise that Intel s definition of TDP is extremely elastic..

8700 TDP is here :

https://www.computerbase.de/2017-11/intel-core-i7-8700-i5-8600k-test-auto-oc-ddr4/3/

And for the 8700T :

https://www.computerbase.de/2018-06...ition-cpu-test/2/#abschnitt_leistungsaufnahme

Intel's definition of TDP is crystal clear, actually. It's not hidden or anything. Yet people keep acting like they have no idea how Intel defines it.

Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload. Refer to Datasheet for thermal solution requirements.

 

[IEC]

My i7-8700K overclocked to 4.7GHz, no AVX offset with -50mV undervolt hits 130W+ package power under heavy load.

There will be a price to pay to hit 4.7GHz+ all-core with 8c/16t.

 

[Abwx]

Intel's definition of TDP is crystal clear, actually. It's not hidden or anything. Yet people keep acting like they have no idea how Intel defines it.

Their CPUs factory settings do not run at the stated stock frequencies, so their definition do not apply, i m curious for the reviews of the 8C CFL, Computerbase gave a hint and that s apparently 150W with AVX2, for a 95W official TDP wich is accurate number with Cinebench, though...

 

[LTC8K6]

Their CPUs factory settings do not run at the stated stock frequencies, so their definition do not apply, i m curious for the reviews of the 8C CFL, Computerbase gave a hint and that s apparently 150W with AVX2, for a 95W official TDP wich is accurate number with Cinebench, though...

Is there anyone knowledgeable about computers who thinks published TDP numbers are the power draw numbers that you will see in the real world? Is there anyone who does not know that AVX loads are hard on CPUs?

Is there anyone out there who looks up the cooling specs for CPUs to see what cooler specs are necessary for the intended use?

 

[Asterox]

Intel's definition of TDP is crystal clear, actually. It's not hidden or anything. Yet people keep acting like they have no idea how Intel defines it.

Yes it is clear, or clear as "child marketing overclocked 28 Core Xeon+water chiller".

 

[coercitiv]

All of the systems were running at the maximum officially supported memory speeds and according to HWUB that equals to gimping AMD and favoring Intel.

Official speed yes, stocks settings only on AMD though (see methodology here). XMP was enabled on the Intel system, leading to lowered memory latency on the kit they used. When using XMP you go from 3200 15-17-17-35 @ 1.35V and scale from there based on manually chosen frequency. When using stock timings you go from 2133 15-15-15-36 @ 1.2V and scale from there based on manually chosen frequency.

On top of that they used a 4 x 16GB memory configuration... wonder why?

 

[LTC8K6]

Yes it is clear, or clear as "child marketing overclocked 28 Core Xeon+water chiller".

https://www.anandtech.com/show/1344...75x-28core-processor-for-extreme-workstations
It will be interesting to see what it can actually do with reasonable cooling.

 

[Vattila]

Not a surprise really, as to me it seems HWUB has become pretty openly a pro AMD site, more or less recently.

Nah — they are just throwing some criticism towards Intel (as PCWorld also does in a recent video, with participants from two other YouTubers). The issue is the publishing of an Intel-commissioned review before the NDA deadline for independent reviews. I'm pretty sure you will get a fair and thorough review from Hardware Unboxed. But it will, like other reviews, I expect, probably show a smaller lead for i9-9900K.

I think Intel could have been more conservative and avoided any backlash. They are already winning in gaming — no need to overstate it.

EDIT: The commissioned review actually puts Ryzen in 4-core mode. See Hardware Unboxed's discussion at Patreon: https://www.patreon.com/posts/21950120.

Regardless, within the context of March 2017, it was a resounding success at $499. Now it looks silly, and AMD has lowered prices accordingly.

Yeah. We need to remember that Threadripper wasn't announced until May 2017, so it was mightily impressive that AMD suddenly had a mainstream chip, albeit at $499, that competed with Intel's HEDT.

 

[dacostafilipe]

It's even worse than we thought, Principled Technologies are running the 2700X as a quad-core. Updated results here: https://www.patreon.com/posts/21950120

Source: https://twitter.com/HardwareUnboxed/status/1049646385186951168