Intel Core i9-9900k 8c/16t, i7-9700K 8c/8t, i7-9600k 6c/6t 2nd Gen Coffee Lake CPU's + Z390

Most likely AMD CPUs will use 7nm and 7nm EUV for their processors for a while. 5nm is most likely a half node so probably going to be a mobile only, just like 20nm and 10nm. They will probably wait until 3nm for the next step.

After Intel's 7nm and TSMC's 3nm, they will either have found new materials for smaller nodes by then or would need to go 3D stacking to further the progress.

 

From 14nm to 12nm was a smaller jump than 7nm to 5nm, AMD won't be skipping 5nm, or 3nm, they are all significant nodes with plenty of investment that will need to be paid back.

Intel's 7nm desktop CPU's will be lucky to be in production by 2023, and by then TSMC 3nm should be well into it's first year of production.

From 2016, considering 10nm full production in desktops in 2020, add 3 years (or more) to the estimates below, making Intel 7nm desktop CPU production in 2023 :

https://twitter.com/tweaktown/status/689999127204200449

 

Nvidia and AMD only jumped on next nodes with at least double the density from previous nodes, From 28nm to 14nm (12nm and 16nm are basically based on same nodes) to 7nm. Barring any significant delays to TSMC's 3nm, they will be more likely to move straight to 3nm from 7nm.

 

AMD went from 14nm Zen 1 (+ Polaris / Vega) to => 12nm Zen+ (+ Polaris) to => 7nm Zen 2 (+ Navi / Vega / Polaris? / Arcturus?) to 7nm+ Zen 3 (+ Arcturus...)


There are no commitments to 5nm or 3nm yet, but as long as the node provides useful power and performance improvements, why not produce products on the new nodes?

Back when AMD and Global foundries were still planning for their future products together, there was a mention of Zen 5 on 3nm instead of 5nm, but there was no mention of Zen 4... which would be the logical first AMD product candidate for 5nm - or spend another generation on 7nm++.

AMD Zen 5 To Be Based On 3nm Process As GlobalFoundries Plans To Skip 5nm
By Ahmad Hassan, May 15, 2018
https://segmentnext.com/2018/05/15/globalfoundries-skip-5-nm-amd-zen-5/

But, that was based on Global Foundries already getting cold feet about investing in nodes 7nm and beyond, and GF soon announced they were bailing out and stopping at 12nm/14nm.

I don't think TSMC / AMD are going to skip 5nm, or 3nm, as they are all for progressing forward with any performance and power improvements each node provides.

TSMC is already well into construction of it's 5nm fab, and I am sure they will offer the first 5nm production capacity to AMD one of their best clients, and AMD is going to want to be first on 5nm in 2020, way ahead of Intel.

Construction of TSMC factory in southern Taiwan fastest in world: Tainan Mayor
TSMC's new Tainan factory has 5,000 workers laboring around the clock to make it operational in 2020
By Scott Morgan, Taiwan News, Staff Writer, 2018/10/18 15:04
https://www.taiwannews.com.tw/en/news/3555224

 

First, 12nm is basically refined 14nm, just like what Intel does with 14nm 14nm+ etc. It takes significant engineering effort and cost to implement design to a new node, and AMD has already made significant effort to implement their designs on 7nm. They are not going to move to 5nm prematurely before recouping their investment on 7nm process. 5nm doesn't provide enough improvement to warrant years of engineering effort when 3nm would be better long term investment. Of course, if TSMC's 3nm is significantly delayed, 5nm could end up being next major node instead, but that is another big maybe.
If AMD ever decides to move to 5nm, it won't be until 2021 at the earliest. 2020 would be for mobile part and usually high performance parts comes out 6-12months later.

 

This is ignorant. All it does is further segment the market and confuse consumers. Studies have shown beyond 3 choices and consumer confusion increases with each additional choice. It is BS logic of failed economics.

This is why AMD disclosed the reason they stopped doing a product at each segment was to target the primary areas and price points at which consumers buy products, hence why, generally, AMD doesn't have the products at each point that Intel or Nvidia do. There is a cost to that, though, as a person may step up to a price point in between those targeted by AMD at which AMD is not competing. There are tons of examples for this, so I won't belabor the point.

Intel doesn't cut costs, and even facing competition from AMD, Intel doesn't cut prices. So higher prices is NOT coming from fewer chips, unless you are talking not binning and selling defective silicon, which then the cost of that wasted silicon must be included in the price of the non-defective silicon, which would increase the price. It isn't retailers that are greedy though, it is Intel. Stop the blame shifting.

Papusan is right in the first part, wrong in the second part. When starting up new lines, you can see increased defects. These can be separate from manufacturing refinements, but coffee lake itself had yield issues a year ago, and Intel doesn't publish yields, other than the rare time they did showing **** yields when they first went to 14nm, where the problems were vast. Took awhile to get it under control, which they did during Skylake, then further under Kaby. But coffee at 14nm++ was a step back on yields until they got the kinks worked out. Opening new plants likely will take a bit of time to get that straightened out.

As to the part on 30% removed = die run hotter, that is not an absolute. While removing the silicon that sits atop the die would lessen the amount of surface contact and diffusion of heat through the extra silicon when the iGP is not present, it is more a question of transistor density for transference to the IHS and to the heat sink. So, there is some truth here, but more explanation is needed.

As to it not being a good idea this gen, why? Because it runs hot AF?

Then you show you don't have a true grasp of the shortage. Analysts said that the profits enjoyed will not be hit until after Q4 of this year. Meanwhile, the largest system's integrator has said it doesn't expect Intel to alleviate the shortage until half way through next year. That is about 6 more months, while TSMC has extra 7nm capacity during the first half of next year, meaning if AMD is smart, they will be pushing Epyc 2 and mainstream Ryzen 3000 out hard and fast at steep price cuts to increase market share and get people on the platform.

Meanwhile, the whole demand is up thing has been debunked. Yes, there was stronger than average demand, but Intel's supplies were lower than historical. This was due to capacity and 10nm causing a traffic jam on 14nm, causing them to give production of SoC chips and other processors to TSMC 12nm. More on that in a moment.

How are you calculating density? Because using two out of three methods, excluding Intel's new method which redefines the calculation, TSMC wins on density. So you cannot say that as an absolute, especially with no figures on actual density seen in their laptop 10nm CPU, which performs worse than the 14nm++ chip it replaces, nor the density changes from abandoning cobalt over active gate (at least that is the rumor), or the other changes made to 10nm that have been documented, which is why it looks as though they made it less dense in order to get working silicon. Go the the Ryzen vs Intel thread and go back to my comments during the summer through September for a more detailed accounting (think I did that before October). I know I have some after Intel was reported as waiting on EUV to mature before adoption, which was expected for 7nm late 2021. Funny you don't mention that.

But, back to density. If it were true that is all that matters or that Intel is so much more advanced, why are they using TSMC 16nm fourth iteration which is 12nm? Makes no sense if it would produce inferior chips. But seems it doesn't so Intel is using it. Funny....

But you are correct, 7nm TSMC is roughly in line with 10nm Intel chips. How you calculate density is a different matter, but they are about equal. Intel's 7nm is roughly around 3-5nm, better than TSMC 5nm density from reported numbers, but considering Intel, once in production, doesn't tell us accurate information on real counts and density, it is all marketing. But, even with that, 5nm, yes, 3nm, not so sure Intel's 7nm matches it. That is 2021 for Intel and 3nm is 2022-23 for TSMC. Samsung is doing their 3nm GAAFETs in 2021, and AMD did use Samsung's 14nm process licensed to GF, so, considering that and the joint venture to create GAA transistors likely using nanowire or nanosheets at 3nm (even TSMC is saying has to make that switch at 3nm, as Intel must also leave finFETs to go to 7nm, except for a hybrid the market is working on to do finFET-esque combinations with GAA designs with SOI implementation, but that is deeper than relevant for this conversation). As such, and without Intel holding the patent on GAA, there isn't a handicap like was seen going to 22nm and 14nm as the market waited for the finFET patent to expire, which Intel only used for about 2 or 3 years before expiration.

As to Sunny Cove, it is the ice lake architecture they have had since 2016. The fact you are falling for the marketing BS is disappointing to say the least. Instead of doing a backport of the design to 14nm, it has sat on a shelf with minor tweaks since Kaby came out. So Sunny doesn't say **** about Intel getting 10nm under control for production in the slightest. period.

And, Holiday 2019 for mainstream chips is over 6 months after AMD's 7nm is on the market. When you have dips like Papusan saying AMD has no answer when AMD is on the exact opposite release schedule by 6 months from Intel, you can't agree with his comments, if you do agree with them, and hold your current statement of belief.

And I agree Intel's woes won't carry over to 7nm. I've said since summer of 2017 that Intel would face issues at least for a couple generations. I also said I didn't expect coffee to be refreshed one year later, which seeing the production pile up, neither did Intel. Their 10nm designs are going to be an issue on performance uplift due to node. This is not to say the uarch is problematic, and those redesigns will give better IPC and other performance increases. But, the market is narrowed considerably during these couple bad years for Intel, and if you don't see it, it is because of willful blindness at this point.

Not most likely, AMD confirmed they are using 7nm in 2019 and 7nm+ in 2020, whereas Apple intends on using 5nm for their phones in 2020. But, Apple products release in the fall, so there was speculation on Apple using 5nm first, as they got a majority of fab time with TSMC for the second half of 2018. Turns out, that is incorrect assumption. Huawei has declared it will be the first customer utilizing the full EUV 5nm process from TSMC in the first half of 2020. This would leave apple to follow that fall, followed by AMD in 2021, unless AMD jumps over to Samsung for 3nm in 2021, which Samsung, GF, and IBM developed GAA in a joint venture and IBM and Samsung both have their independent processes for it. This is needed to go past 10nm for Intel and 5nm for TSMC. Plus, costs on building the fab and when time of completion is expected, although TSMC mentioned their 5nm capacity may be able to switch for their 3nm lines already. So, many bits out in the ether there.

As for 3D stacking, it isn't new, and AMD already has a plan to do it once the cost of the active interposers go down for the nm node the interposers need for construction for packaging. Meanwhile, Intel had to push 14nm designs to the 22nm node, meaning they now have fabs they couldn't decommission when planned. That means not keeping it at capacity bleeds money. Enters 22nm active interposers to the rescue. Mature node, low defect density, and fills out the empty fab time. It was a business decision heralded as more than it is.

You are explaining it in a way that misses the critical part that @JasonLLD also gets wrong: it is a half node. They left the base footprint on those layers the same, but utilized smaller transistors, which decreased density, thereby allowing more thermal headroom. Intel actually made Kaby and Coffee less dense than they made Skylake, which aided in the extra thermal headroom leading to higher clocks, which is why AMD got around 15% higher clocks from Zen+.

As to Intel's 7nm, unless you got some inside info you want to share, late 2021 volume production is still the roadmap, which is products in 2022. This is assuming all goes right, though, which is in question with Intel's recent record.

This is ignorant. Things are changing and AMD has a long history of going to new nodes before others. Just so happened that TSMC scrapped the 20nm node and went for 16nm. Nvidia usually waited a year to get the bugs out of a process node before adopting it, which is different than shooting for a density. If you cannot accurately describe WHY the industry went from 28nm down to 16nm/14nm nodes, then stop talking. It is easier.

Now, the question is why not use 5nm in 2021 for the Zen after Zen 3 and use 7nm+ again, all to wait for 3nm in 2022 or 2023 for volume production, roughly same timeline as Intel does 7nm designs, when instead the 5nm gives something, even if small, and is not significantly different from 7nm+ EUV and is the last node a finFET can effectively be used, unless there is a problem doing finFETs in regards to yields on 5nm, or why not go to Samsung 3nm GAA in 2021 instead? Especially since GAA continues below 5nm for Samsung and TSMC.

Your argument needs fleshed out. Not saying you are wrong, but you lack certain details to make a compelling argument.

It is called a partial node. You only shrink certain parts of the chip on transistors, not the footprint. Doesn't mean it is EXACTLY just a refined node. TSMC 12nm is more of just a refined node and is technically the 4th iteration of 16nm. There is a difference. AND Intel is using that 12nm to produce chips. Also, TSMC 10nm beat Intel's density on 14nm over a year ago, on Intel's own terms at the time.

Also, you are NOT familiar with TSMC's new partner plan for accelerated node development, cutting the time per node from 3 years to 1.5yrs for development, saving on costs. Also, with EUV on 7nm+ and 5nm helping to cut on the quad patterning that has plagued the industry, the move is LESS than what you currently make it out to be, although it is worth noting making a single design on 7nm costs about 3X that of the 14nm node. Good thing AMD is using only one core design that scales from server to mainstream. And as they separate more complicated elements off, such as the I/O chiplet, then they are given more flexibility to push ahead with cutting edge nodes and recouping costs while reducing the defect rates of unusable silicon dies from the wafer, etc. But I don't need to tell you this, do I?

You seem to follow the industry well enough, but you are missing critical details in your comments which change to total mix of information. Make more compelling arguments!

AMD:
2019 = 7nm
2020 = 7nm+
2021 = 7nm+/5nm+/Samsung 3nm
2022 = Samsung 3nm or TSMC 3nm (may bleed into 2023 waiting on TSMC 3nm, hence the shift to Samsung makes a compelling argument).

Intel:
2019 Q4 = 10nm desktop
2020 = 10nm Server and HEDT/14nm products continue
2021 = 10nm refinement (previously called tiger lake)
2022 = 7nm Saphire Rapids/Granite Rapids

Show me Intel's process lead. Looks like it is a microarchitecture fight, not a process fight anymore. And both companies have amazing uarchitects. So, I don't see where you wind up where you do at ALL....

 

Intel i7-9700K Review: Hyper-Threading's Value vs. 8700K
Gamers Nexus
Published on Dec 26, 2018
We benchmarked Intel's i7-9700K vs. the 8700K, 9900K, 2700, and pretty much every other major processor, including the i7-2600K. Overclocking included.
Article: https://www.gamersnexus.net/hwreviews/3421-intel-i7-9700k-review-benchmark-vs-8700k-and-more
The Intel i7-9700K received ample criticism at unveil for its core/thread assignment, feeling like a departure from a decade's worth of hyperthreaded i7 CPUs. The oddest aspect of the change was not the 9700K itself, but its predecessor: Intel had just moved to 6C/12T on the 8700K, a fierce and competitive upgrade from the 7700K's 4C/8T configuration, but has now regressed to an 8C/8T part. It's a mid-step. With two more physical cores and a boost in frequency, the part will undoubtedly do well in gaming -- the question is just whether the value is there, especially with a price-hike to over $400 on the new i7, up from the $350 range on the 8700K. Further still, the 9700K drops four threads, which have proved advantageous in some workloads.

 

Intel i7-7700K Revisit: Benchmark vs. 9700K, 2700, 9900K, & More
Gamers Nexus
Published on Dec 30, 2018
We're revisiting the Intel i7-7700K, which feels a whole lot older than it is given the pace of Intel's product launches. This benchmark comes in at the end of 2018 to compare against the 9700K, 8700K, et al.
Article: https://www.gamersnexus.net/guides/3423-intel-i7-7700k-revisit-benchmark-vs-9700k-2700-9900k
The Intel i7-7700K got the short straw when Intel was drawing for products, launching in January of 2017 and being superseded just 9 months later by the objectively and notably superior i7-8700K. The 7700K was still a strong overclocker, so we wanted to revisit this CPU with a 5.1 and 5.0GHz OC applied (depending on test) to see how it benchmarks versus the i7-8700K, 9700K, 9900K, R7 2700, and more.

 

Ha, I got double-burned, I bought my 6700K in Nov 2016, while 7700K launched in Jan 2017 (the difference between 5.0Ghz and 4.7Ghz for likely max overclock), while 8700K launched just a few months later! My 6700K is still good though for what I need it to do, and their conclusions in that article were that 7700K is still good for gaming. If another fps multiplayer title comes out which I get into and if that requires a beefier CPU than my 6700K then that would be the time to buy a new motherboard and CPU, so I can't say for certain that my 6700K will last out another GPU upgrade - I don't think it will if I want to be certain to stay at 144fps minimum. I could put a 7700K in this motherboard, but we're only talking about 7 or 8% performance gain, which could translate to the same 8% performance gain in any future CPU bound games, but that's only the difference between 130 to 140fps, not worth the hassle. Might be quite likely that I'll upgrade to the "9900K" of whatever the next Intel generation is gonna be, unless AMD's upcoming CPUs will be able to push 144fps without compromise.

Didn't get too burned though, at least I built my PC before RAM prices doubled, and while GPU prices were sane before the mining craze, not to mention getting on the Pascal bandwagon in the first year (which was a good move).

 

We analysed the Z390 socket!
der8auer
Published on Dec 31, 2018

 

Unannounced Intel Core i9-9900KF, i7-9700KF, i5-9600KF and i5-9400F CPUs Listed
by Anton Shilov on December 28, 2018 5:00 PM EST
https://www.anandtech.com/show/13750/intel-core-i9-9900kf-i7-9700kf-i5-9600kf-i5-9400f-cpus-listed

"The products in question are the eight-core Core i9-9900KF and Core i7-9700KF, as well as the six-core Core i5-9600KF and Core i5-9400F. These devices have been listed by retailers Data-Systems.Fi, Newegg, and distributor Synnex (see screenshots below).

Spoiler: Details...

According to Intel’s existing nomenclature, the CPUs with model numbers ending with F, like 9400F lack integrated graphics, so we suspect the new processors will primarily target higher-end systems featuring discrete graphics. This will mark the first time that Intel has launched integrated graphics-free processors in its mainstream family at the high-end for many, many years.

The higher-end Core i9-9900KF, Core i7-9700KF, and Core i5-9600KF look set to run at the same frequencies and feature the same cache configurations as their non-F colleagues. As for the Core i5-9400F, this six-core chip runs at 2.9/4.1 GHz, well below the clocks of the i5-9600K, but will have a TDP of 65 W. All of these parts, according to the listings, will be able to be used in current 300-series motherboards.

Intel has not officially confirmed existence of these CPUs, or mentioned plans to release them. In the meantime, listing of the Core i5-9600KF by Newegg and the Core i5-9400F by Synnex Thailand indicates that their launch is likely imminent. Avid readers will remember that CES 2019 is taking place in early January, so the question is whether Intel starts to sell these CPUs more or less quietly ahead of CES, or if it will announce them publicly at the trade show.

In any case, if Intel proceeds with the launch (or rather when), it may broaden availability of its latest eight-core and six-core CPUs both in terms of physical availability and in terms of pricing (i.e., the i9-9900KF will hopefully cost less than the i9-9900K). In the meantime, one has to remember that Intel has high demand issues in general, so the effect of the launch is something that remains to be seen. It could be that the company will not focus on the Pentium/Celeron parts this time around, but instead make these higher-performing (and higher margin) offerings more regular.

 

Is Fast RAM A Waste? Unleashing the Core i9-9900K with DDR4-4000
Hardware Unboxed
Published on Jan 1, 2019

 

Great. Since he didn't need to touch vcore even with nearly half the pins taped - it seems there is zero effect on stability too

 

I wonder why Intel would change the pin count/layout then? It makes the new CPUs incompatible with Z170/Z270 unless you make modifications. It would sell new motherboards, but I don't think Intel has a vested interest in increased motherboard sales? If anything, the ability to have been able to drop in a new 9900K into a Z170/Z270 would probably stimulate 9900K sales a little bit, as it's an easier & cheaper upgrade rather than buying a whole new motherboard. I'm not sure on Intel's motivation here?

 

the test itself is fine but things were left out. was 9900k running at stock? if uncore and core frequency were overclocked, then it can take advantage of faster memory by another few %. not to mention they didnt test any avx2 workload as they should, those hogs memory bandwidth and a jump in bandwidth will yield much better results.

overall the video isnt doing any justice for faster ram. even consider the diminishing gain (which still exist, just higher gains than whats shown in the vid) they need better software to test the gains. its kinda like saying no point buying ryzen 16 core cpu and then test ryzen 16 core cpu with only single threaded softwares, you get my point.

 

They do care about selling motherboards, though. MB MFRS. Are partners. You need both to build a PC. By artificially limiting a socket, you drive sales keeping partners happy. By doing so, you can cause events like the MB shortage when Ryzen released in 2017, as well as MB partners blowing off doing a decent UEFI, followed by MB partners blaming AMD when it was obvious Intel pushed Z270 at that time, plus Chinese new year, causing them to say **** off AMD. You could not do that if AMD was selling more (like now) and you weren't the primary driver of their income by releasing parts driving sales.

That is why the 8700K with the z370 socket 9 months later chapped people, followed by Z390 which does nothing to drive sales comparatively over the Z370.

But AMD isn't helping MB MFRS either in allowing long socket compatibility. Now, that may change with the new PCIe 4.0 or 5.0 in the next two years, or with DDR5 support in 2021 (should be noted, this is after the EOL timeline for the socket for DDR5 support). If AMD can do backwards compatible, but continue to drive sales for MB partners through new features, like precision boost overdrive, better VRMs, etc., then they won't need to pull an Intel to get support.

Sent from my SM-G900P using Tapatalk
Yeah, but the channel is focused on gaming, and this shows gaming doesn't care about BW as much. There are many limits involved, and it shows big, flashy ram is a diminishing returns scenario for average people.

I paid $400 for 4x8GB 4133 ram in 2016 or around there. You won't see those prices today, still. Now, it also gives the mistaken impression bandwidth doesn't help much at all. He also chose some rendering benchmarks that just fit into cache, not actually testing the memory in some scenarios. Shame.

But, it will make a product with HBM2 or HBM3 on package harder to sell in the future. That has 10x the bandwidth of DDR4 dual channel memory. Yes, you have to practically use an active interposer to get latencies to acceptable ranges, but if you need bandwidth, that is a huge jump (and pricing for 16GB is in the ballpark of 16GB 4133+ DDR4 right now).

 

i know its for gaming, they still went ahead and did a bunch of synthetic tests though. sandra, aida and cb15 etc. 7zip and maybe winrar would be something that can take advantage of it, use ramdisk/vmware so its good to have that too.

 

I know i have posted several times on my 9900K but wanted to give a quick rundown of what i found with my chip. This is with no AVX offset.

-At Stock with XMP enabled and Turbo CPU Loadline Calibration i was able to get 1.21 Vcore in the bios stable. This is at 4.7 uncore. I have not updated to the Gibabyte new bios but is seams they have changed the stock uncore to 4.3 now.

-At 5.0 with XMP enabled, Turbo CPU Loadline Calibration and C states disabled I was able to find stability at 1.345 Vcore in the bios. This is at 4.7 uncore.

-At 5.2 with XMP enabled, Turbo CPU Loadline Calibration and C stated disabled i was able to find partial stability (did not crash but did not do as much testing) at 1.435 Vcore in the bios. This was also at 4.7 uncore.

Testing done was Aida 64 blend test overnight, Prime95 for an hour or more, Intel Burn test on standard, very high and maximum, OCCT for 4 hours and many hours of gaming. 5.2 was limited to Aida 64 for 10 minutes, Cinebench numerous times, Intel Burn Test on standard, very high and maximum, OCCT for 4 hours and a few hours of gaming.

At Stock settings above i am seeing max temps at around 55C CPU and 43C GPU in Gaming.

At 5.0 settings above i am seeing max temps at around 66C CPU and 45C GPU in gaming and CPU temps maxed out around 83C during other intense testing.

At 5.2 settings above i am seeing max temps at around 78C CPU and 48C GPU in gaming and CPU temps maxed out around 93C during other intense testing. My normal everyday is 5.0 as these temps and Vcore are higher than i am willing to go.

During all the testing i have been adding radiators so some temps may actually be lower now if i was to rerun tests; however, with my newest radiator i have reached a point where it did not matter on the tests i run. I am now able to run the fans very low which is nice.

 

I think you made the right choice about running 5.0Ghz as your daily driver. I also think it's good that you say at the end there that you can run your fans on low rpm now - to me that's what good cooling is about, the ability to get low or decent temperatures while running at quiet/silent fan speeds while running a healthy overclock at the same time! (Yep, that's a lot of radiators you have there!)

 

That is why I went TR: memory bandwidth and cores. It is hard with what is out there to show impact, unless showing server work or Linux. Hell, adobe is still just single core optimized which is why Intel trounces AMD in a lot of its workloads. Or, many of the workloads that could use the bandwidth were already moved to GPUs making the CPU memory bandwidth where critics would say it is unrealistic to test using that, even though it is still done on other OSes, or in specific workloads.

It also says that programmers, in certain instances, could try to change that so that bandwidth is better utilized. We will likely see that after DDR5 comes to mainstream around 2021.

What I want to see is AMD, once they use an active interposer, just charge $250 more for 16GB HBM2 or $500 more for 32GB HBM2 (or HBM3 if available at the time for roughly the same cost) on their chips. If the latency on active interposer can be 30-60ns, while doing 10x the bandwidth for mainstream, 4-5x on TR, then doing the dual interposer for 1TB bandwidth for server chips, while allowing that to be fed by DDR4/5, it really would be night and day.

I feel the same way about Intel's Foveros which uses an active interposer. Granted, that would make the 9900K $750 for 16GB of integrated HBM2, but how is that different than buying 4000 MT or higher ram speeds? Just need workloads that can use the 10x bandwidth.

Sent from my SM-G900P using Tapatalk

 

I started off with 2x240mm (480mm) which was great with just the CPU and then added the GPU into the mix. I then added a 360mm as my CPU temperatures came up. With the 3 radiators i was seeing a temperature difference between the fans running at 1800rpm to 1100rpm of around 5C which is really not a great amount. I then added a fourth recently with some Christmas gift cards and i have not tested extremely low yet but have made it down to 900rpm and am still seeing the same temps i was seeing at around 1800rpm with the three radiators. I am going to test the fans even lower in the future but with my old ears i hear the pump over the fans so it is pretty silent to me at 900rpm. Kind of overkill but the rad's should last a long time.

 

Trying to get into HEDT cooling territory? At some point, consider the mora3-420. But you are closing in on my 3x480s! Lol.

Sent from my SM-G900P using Tapatalk

 

This case is pretty heavy now. 240 mm PE in the front was really about as big as i could go unless i went with a thicker one. 420 mm CE (45 mm thick) on the side, could have gone with a 480 mm but had 3 140 mm fan's that fit the build already. Top has a 360 mm XE (60 mm thick) and a 240 mm PE on the top as well. The 240's came from another case so did not worry about upgrading them. One side on the top allows for a thick rad but the other side would have to go with a thinner longer rad if i ever did go bigger due to the video card. Case will fit up to 2 480's in the top, one on the side and one 240 mm in the front but i think i am good with the ones i have!!

 

yep going TR is good, those extra cores is worth it and we only need to upgrade the CPU no less. DDR5 and pcie4 wont be that much of an upgrade unless very specific thing like sequential SSD bandwidth or memory bandwidth is involved, quad channels solves most of the problem already with ddr4.

i wanted to go zen2 real bad because I use ram disk, like i use A LOT of it constantly and i start to notice if i do editing (video/audio which uses lots of ram) while having cache files for those video/audio in ram again, it slows my system down, even with my latest clevo.

i'd love to have those extra L3 cache and quad channel, less in memory and more in L3 cache the better, it'll at least take some stress off memory for things like browsing which are chunks of small files, stay in L3 more pls.

 

That's a coincidence, I run all the fans (Noctua 140's and 120's) in my PC at 900rpm maximum too which is 75% fan speed. I configured it through motherboard fan control curves so that at CPU temperatures above 40 degC all case fans run at 75% speed, while at idle temperatures below 30 degC CPU temperature I have them set to about 40% (450rpm) - this way it's silent during idle and just 'quiet' during max loading/gaming. I found 75% was the best compromise between cooling performance and fan noise, while for PC idle fan speeds I just selected the minimum percentage fan speed that would be just enough to trigger the fan to spin.

 

I also run them all through the bios. Custom fan curve sounds nice I will give that a try. Had them manually set to 800rpm last night for 3 hours of gaming and still no difference in temps. I like the idea of the fan curve though and will probably give it a shot the way you have yours set up.

 

Yep, I just set a linear relationship of increased fan speed between 30 degC and 40 degC and 40% and 75% respectively. The only reason I chose those 2 temperatures: CPU is always below 30 degC at idle, CPU is always above 40 degC during gaming. That way during gaming I can be sure I always have the full 75% case fan speed so that my GPU can get plenty of air. (My CPU fan speed is not temperature related, that's just set at a constant 900rpm, which is Noctua NH-D14 air cooler standard behaviour when you have the Low Noise Adapter plugged in.)

 

Sounds like a good setup you have! I am using all EK Vardar fans except for the 140 mm fans on the 420 mm radiator which are Phanteks High Static Pressure as i had those on hand and did not want to spend more on fans. They are all white though so at least they all match in color. Either way i can set up a curve in the bios, just have to do different percentages. Nice to have them completely silent doing normal tasks and in gaming I don't notice the fans as i use a headset.

 

i9-9900K In A Laptop?!
Jarrod'sTech
Published on Jan 4, 2019
This laptop has an Intel i9-9900K desktop PC CPU inside! So 8 cores / 16 threads and it can be overclocked, but just how well does it perform?

Spoiler: Details...

Given the 9900K (8c/16t) is a hot chip we’re expecting some thermal throttling with it cramped inside a portable machine. In this video I’ll cover thermal testing, overclocking, game benchmarks and a number of CPU specific workloads like Adobe Premiere, Handbrake, Blender and more to give you an idea of how it performs.

I've also compared it against a 9900K (8c/16t) in a desktop PC to get an idea of the difference in performance, and against a typical i7-8750H (6c/12t) to see how a regular modern laptop compares.

The laptop is the Clevo P750TM1, sold here in Australia through Metabox, get subscribed for the full review!

Check pricing (Australian dollars): https://www.metabox.com.au/store/b234...

Apparently it's an updated chassis, as he says it's not as thick, has a swappable / removeable battery(!), and only uses 1 power adapter vs 2 used with the previous 8700k model (didn't mention the adapter wattage on the 9900k model).

 

Same old friend but in same old suit and configured with 9th gen Core i7/i9. Delivered with single 330w psu. Put it in here... http://forum.notebookreview.com/thr...np9155-p751tm-g-owners-lounge.809622/page-132

 

Nah he got it wrong, that's the smaller 15 inch P750 Clevo chassis. Metabox call them all "Prime-X" which is probably the source of confusion. The one he tested the 8700K in last year (a full six months before the weaker 6 core laptop BGA chips came along) was the P870TM, 17 inch SLI beast, which with the TM revision has a revised, bigger cpu heatsink.

All three of the Clevo LGA chassis (P750, P775, P870) are being advertised around the place with CPU upgrades to 9900K. There's even listings on ebay from some German retailer of earlier Z170/Z270 Clevos with these CFL-R chips in them.

 

The P75xTM1 also got an updated heatsink for the GTX 1080 (not the 9th gen CPUs), which he shows in the video. But yikes, those temps. Australian summer isn’t doing a machine already on the edge of the thermal cliff any favors, although 25C ambient isn’t that bad. Maybe the only environment it won’t thermal throttle in is @Papusan’s igloo.

 

It could mean less power draw against the package limits + more thermal headroom due to the silicon being "dark", but those are side-effects, not likely Intel's primary intention for coming out with these CPU sku's.

It would appear that Intel is having fab issues with a higher than usual percentage of iGPU region failures, and Intel's reject barrels are filling up with these otherwise good CPU's.

Intel "needs the dude's" - shortages of shippable CPU's - so Intel is making sku's to move this product out the door.

[I am responding here to keep the CES 2019 thread uncluttered with specific discussions like this, I hope you agree it's better to discuss this here in the subject matter thread rather than the CES announcement thread. ]

 

Alienware Area-51m: an exclusive look inside
The Verge
Published on Jan 8, 2019
The Alienware Area-51m might be the ultimate portable desktop: a 17-inch laptop that adds interchangeable graphics cards and swappable desktop processors for the first time in years. This one fits up to an Nvidia RTX 2080 and an Intel Core i9-9900K, the fastest parts in a notebook yet. Alienware designed the Area-51m to be easily rebuilt so we tore it down to see what’s inside.
http://forum.notebookreview.com/thr...ts-video-articles.826767/page-8#post-10844035

 

"Alienware got fed up with thinner and thinner gaming laptops and the restrictions they impose on the hardware and cooling"

THIS FROM ALIENWARE?! SERIOUSLY?!?!?! LOL

on a more serious note: WOW! i wouldnt have expected such a step from dellienware, this is truly amazing! naturally, they decided to go for a completely proprietary gpu form factor so one would HAVE to buy the upgrade directly from them, but still! this is a HUGE step into the right direction!

Alienware, welcome back to your roots

@Mr. Fox

 

Now THAT's really scary! Let's wait when the relaunch has started and the first reviews are up.

Then we can determine if Dellienware truly did a step in the right direction, ha ha!

Edit: Did u see the paste job on this one? Geeez! Maybe they turned back and went where they came from, BUT it's still the same paste-Job Johnny! At least one tube of toothpaste! Hilarious!

Gesendet von meinem CLT-L29 mit Tapatalk

 

lol thats totally fine, the stock paste jobs on Clevos are equally as horrible using tooth paste haha. would already be pure awesomeness if alienware provided a machine that had the potential to be great once properly tuned by the user, just like clevo machines

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

It's intrigueing to say at least and if I understood correctly it might be not that pricey -> entry level at two and a half grands!

Well, lets see how this one goes....

 

not too shabby, get the lowest tier model and upgrade the components urself using cheaper aftermarket prices for the hardware (except for the gpu of course...)

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Sure http://forum.notebookreview.com/thr...enware-15-and-17.826796/page-18#post-10844664

You mean this is the roots?

 

Ah, the second deep dive look! Wasn't aware of this thread already ripping this one apart....

It reminds me of the Acer Predator 21x where expectations were raised high initially before it turned out to be the priciest piece of underwhelming performance which would have been okay for 8 Grands less.

But yes, lets see how this goes! I have to admit that I do like the retro look of this one!

 

https://www.amazon.com/Intel-i9-990...?ie=UTF8&qid=1547062468&sr=8-1&keywords=9900k

https://www.newegg.com/Product/Prod...tion=9900k&cm_re=9900k-_-19-117-957-_-Product

9900K Price drops. $499 and free Call of Duty BLOPS 4 game included.

 

aaah and there goes the hope of this machine having the potential to be great in the hands of the right users... was nice while it lasted though, for like 12 hours lol

nevertheless, this instills at least SOME hope that there are still some true enthusiasts left at alienware. plus, it does its, albeit small, part in keeping the desktop LGA / MXM laptop space alive

 

For me... Lasted for like 5 sec. At max Just look at the roots, bruh

 

i understand what ure getting at buddy, but im talking about the time when alienwares were the nonplusultra of DTRs, around the sandy bridge area. im aware its been quite a while, though!

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Get less, pay ... erm, the same?


Tom's Hardware: Save Zero Dollars By Opting for Intel's iGPU-Crippled CPUs.
https://www.tomshardware.com/news/intel-f-series-9th-gen-processors-price,38434.html

 

HA! i KNEW it! just defective silicon where the igpu is disabled. and that at THE SAME PRICE as the regular 9900K. whatever, intel!

and since its defective silicon, u can bet ur ass it will NOT clock better than existing 9900Ks

 

It doesn't seem like it will clock any better. But in the same vein, it won't be any worse either.

From the article - "The disabled graphics unit doesn't impart any performance advantages, such as longer boost duration or higher overclocking capability. Simply put, you can expect the same amount of compute performance from these chips as their normal iGPU-equipped equivalents, meaning the only advantage would be that you might actually be able to buy the processors if the normal chips are out of stock. "


This quote is interesting too, "it remains to be seen if the new F-series processors will land at a lower price at retail outlets... That means the F-Series processors could theoretically retail at lower prices if their equivalents are harder to source, but by setting the recommended pricing at the same level as the full-featured models, Intel has (perhaps inadvertently) given retailers license to mark the lesser models up to the same pricing we see with the processors impacted by the shortage. "

 

Yeah, Intel just can't help it can they? Intel, what a bunch'a screw up's

I don't see how there are going to be more of these than the regular CPU's, since "rejects" are supposed to be a *small* percentage of the total, and otherwise working except for failed iGPU's would be an even smaller percentage, so there outta be about 1 barrel of these to sell - worldwide?

I guess someone will buy them for the wrong reason and be disappointed, then return them, so there will be "rejected" "rejects" on the discounted sale table everywhere.

Intel is an embarrassment these days, selling the floor sweepings at this point just to get some more sales to keep the doors open.

Save Zero Dollars By Opting for Intel's iGPU-Crippled CPUs
Submitted 20 hours ago by bizude
https://www.reddit.com/r/intel/comments/afnep4/save_zero_dollars_by_opting_for_intels/
https://www.reddit.com/r/hardware/comments/afnaeb/save_zero_dollars_by_opting_for_intels/
https://www.reddit.com/r/AyyMD/comments/aftlb6/what_a_deal_0_off_for_a_cpu_without_an_igpu/

Tony49UK 191 points 4 days ago
"Also worth noting that the normal chips support 128GB of RAM but the gimped versions only support 64GB."

Dasboogieman 93 points 4 days ago
" Also minus TSX-NI instructions. => 9900k TSX-NI Yes
This is a pretty big deal for multicore scaling efficiency."

 

LOL intel trying hard to max profit b4 amd drop the bomb

too bad, only tards would buy these KF cpu if they are priced the same. not even from a higher binned chip.