Ryzen vs i7 (Mainstream); Threadripper vs i9 (HEDT); X299 vs X399/TRX40; Xeon vs Epyc

about the announcement;


Edit; TBH this throws the ball into AMD's court. On the lower end HEDT it makes the offerings much more competitive. If the x399 is not supported with the new TR then the 10980x might be tempting according to price points. I have already shown where the 16 core is way more than I need.

 

IDK how interesting this is going to be to you, it seems like a bunch of Intel marketing BS to me - I've had it on in the background - but some of you might find it interesting being current Intel international marketing. Notice how few are in attendance, Intel's loosing popularity?, bad Marketing Organization - not getting the word out? I don't understand the title - must be Australian Slang?:

Museum of Contemporary Art Australia
Intel 10th gen paunch Australia
Alex On Tech
Started streaming 40 minutes ago

 

The screen capture is current (Oct 2), the cost of this 'old' chip is no longer tenable, what is its relative worth now, new ... $800? The entire 7xxx 9xxx ownership must have let out a collective gasp today, as Intel snaps its finger like Thanos & their over-priced chips lose approx. half their worth in an instant

I felt a great disturbance in the Force, as if millions of voices suddenly cried out in terror & were suddenly silenced. I fear something terrible has happened -Obi Wan Kenobi-

Wow. Ty AMD & Dr Lisa Su. Price drops similar to this were clearly possible last year. That $2000 chip that turned into a $1000 chip overnight will live in infamy as the testament to their undeniable anti-consumer greed. When Gen10 hits the market the 7xxx 9xxx prices will nose dive as the market floods & a race to the bottom ensues to get what u still can ... a slow-motion train wreck if you're in possession of one of these chips ... ... ...

 

I agree as it has to just be a bad situation if you dropped $2000 on a 9980.

 

Before this announcement, Intel fans talked crap on AMD's new chips each year dropping the value of the old chips so much that it wasn't worth getting AMD, ignoring that prices are not set on intrinsic value.

Now, just like Intel did for the 6950XE or whatever, Intel undercut it's existing line so hard that anyone stuck with the old chips practically cannot resell for anywhere near what they paid.

In other words, Intel just adopted the generational price abrogation seen on AMD chips, eliminating any idea that they are worth the value.

But, when AMD releases Zen 2 TR, just like mainstream Intel chips are going against both zen+ and Zen 2 on pricing and performance per dollar, Intel will be facing cheap Zen+ chips and competitive performance Zen 2 chips. That means Intel is in for some hurt. If nothing else, this says the amount they cut from the product was all margin to begin with, as nothing between 14nm+ and 14nm++ radically changed levels of production. As such, it really shows how much Intel has been gouging their consumers, especially with the 6-series followed by the 7920X, then this.

The only reason a company can charge so much for something not worth it is because consumers buy it, not lack of competition. Competition does give consumers alternatives. But it is when consumers stop buying, not the presence of competition, that prices drop. The problem with high prices is consumers. AMD, according to mind factory, has been selling 2:1 compared to Intel on volume for HEDT since Zen's release. Now, with the mainstream 12-core competing well with Intel's 10 and 12 core HEDT, Intel had to cut prices to sell their HEDT. That means less margin moving forward (the shareholders will feel that force in Q4 and Q1, along with now having to write down aging inventory due to the price leaks).

Now if consumers would only figure that out with nvidia's Titan and XX80 Ti lines...

 

In the end it comes to fruition, AMD has drastically changed the CPU landscape. On the higher end of HEDT Intel has nothing to compare for the performance. TBH 16 core AMD is more than enough for me at ZEN 2, I would have preferred the 24 core but I will take what I can get for the x399. If I had to give up the x399 I would probably go AM4, again though I would have to see all the prices.

 

That's the tough choice, get the 16 hand picked cores included with the AM4 3950x *and* get another generation of upgrade for the CPU in 2020, or jump to the TR5(?) with either 4 or 8 channels of memory?

Or both?

 

No, I'd rather wait on mainstream. Hopefully with the coming changes, all mainstream chips will get a boost in PCIe lanes in 2021, especially as HEDT moves to 8 channels next year (at least on AMD, whereas Intel will have either 6 channel HEDT or 8-channel, depending on process used and decisions between now and then, but Intel's server chips next year seem to be heading toward 8-channel, so)...

With increasing mainstream core counts, server getting 80 cores next year and potentially 128 cores in 2021, meaning the HEDT core count may increase again in 2021 (I'm not betting on AMD going over 64 cores for HEDT until then), there is a chance that what once was considered HEDT (like Intel's 28-40 PCIe lanes and potential memory channels) might find itself filtering down to mainstream, which would move more back to mainstream from HEDT except for those that need it. It also would further knock Intel's offerings at a time they could ill afford it.

But until mainstream gets more features, there really isn't a need.

Also, EposVox is going from a 7980XE to Epyc, as a fun side note.

 

Sooo, you are waiting until 2021 to upgrade your Ryzen set up? Seems like a long wait...

 

First, I was talking if considering mainstream rigs.

Second, if the price or performance on the 24-core isn't right, then I may. If they had a 16-core TR x399 compatible Zen2 chip for around $800-850 (sliding in around Intel's 14-core pricing and under the 18-core pricing), then yes, I'd guarantee I'd upgrade.

As it is, with what is publicly known, I have to wait to see the 24-core value.


What has happened with Intel's pricing on the HEDT segment just put the entire lineup between $570-$970. That means AMD is uncontested for their 24 and 32 core CPUs in the $1000-2000 range, including smashing the value of the 28-core Xeon whatever.

But, now you can get more memory channels and PCIe slots buying Intel HEDT with good, possibly higher, single thread performance and good multi threaded performance (waiting to see on mitigations) than AMD offers.

So, your options are to go mainstream on either for about $500 with an 8-12 core CPU, depending on vendor, soon a 10-core from Intel up to $600, or get AMD's 16-core for $750 with only 2-channels of memory and like 20 or 24 PCIe lanes with no other offerings until you reach over $1000, or you go Intel HEDT with 10, 12, 14, or 18 cores, with like 48 PCIe lanes and quad channel memory for between $570 and $970.

I may talk of Intel being in trouble, but at those price points, Intel's HEDT is compelling, all while no rumors exist of any AMD chip that is comparable.

This is why I said AMD should have an entry 16-core threadripper, because that thrown in the mix for $800-850, and maybe reducing mainstream 16-core to $700, would competitively knock Intel's offerings even at the new pricing. It's like pouring salt on Intel's nuts when the sack is already split open, then grinding your heal on the wound. AMD destroyed Intel's HEDT platform. Now that Intel is trying to move to where it has value, that is when you give no quarter. Then you use the new workstation platform to destroy workstation Xeons entirely.

Now, with TSMC expanding money to 5nm fans, they likely expect more AMD growth and more demand from them. It will be fun.

But we have to see what AMD will do...

 

I thought they were pricing for the 16 core to go heads up against the 14 core Intel now at $784 and if they price the 24 against the $999 to the 18 core variant? This makes AMD still the value chip. The 3900x 12 core is still a value then over Intel 10 core at $589. This could be a reason Intel is panicking they are also trying to drive the AMD chips cost down too before their launch. If it comes to low profit margins which company can hold out longer?

 

And nowhere in that statement is there proof AMD will go from $1400 per 24-core to $1000. As I said, they have the $1000-2000 market Locked up. But a 16-core on mainstream, even of it is good, doesn't have the mem bandwidth nor the PCIe to go against Intel's HEDT.

Edit: also, look at the cost per die at 7nm. AMD increased their profit per chip with it. So having to tweak prices isn't going to hurt them like it will Intel, whose yields drop off a cliff the higher the core count goes.

 

It's funny, some are seeing Intel's new price for their 18 core as forcing AMD to come in under, but I see it as Intel figuring out AMD's pricing - or hoping to have figured it out - and are preempting their doom by dropping their pricing before AMD comes out with a better product for a lower price next month.

Intel is either trying out lower pricing before AMD launches to see if they can catch some flounder before AMD releases their better performing and better price competitive models, or Intel has lost their nerve.

AMD had 80%+ of the market sales last month (see my other post) so Intel has to catch their prey between AMD's announcements - a short window of sales where their customers can justify not waiting for AMD to announce - then it's back to Intel's 20% market sales Gulag.

AMD made Intel flinch before the Boo! has been released.

 

No, this is why I said price points. Up to $400, AMD is a recommend. At $500, the 8-core 9900K still takes some spots compared to AMD's 12-core. With a 14-core on a better platform coming around the mid-$700 point, Intel is poised to take the $500-1000 market. Period. Above $1000, AMD takes it.

So it is about price points. And for every chip sold in that price point to Intel is a guarantee they won't get a 24-32 core sale trying to step people up next year, supposing they even keep any quad channel offerings for Zen 3.

This should be Intel's least X299 chip, so it isn't about keeping them on the Intel platform, rather keeping them from going to the AMD platform.

This is why AMD this month needs to come out and lay out their socket compatibility plans for HEDT/workstation market.

 

The thing being AMD is starting to see black where Intel has been there for quite some time. With SSD's and other ventures they have elsewhere to fall on as well. My look at it is Intel may be finally ready to fight a war over it where they had not had too before. Since all the preemptive chip releases to combat prior performance improvement have no been made (ie. 5GHz chilled performance etc.) now it is time to throw other marketing tactics out there. Intel is struggling to, as always, make the first strike.

Agreed, they should come out and state the chipset plans for TR ASAP.

 

new TR3 boards will likely last 2 gen and may support zen3. if only a laptop with TRX80 comes out, i think its time to go desktop. hopefully these will work with optane dimm too can't wait to use those for OS drive.

 

That makes no sense. TRX80 is not coming this year. That is known. A rumor of no backwards compatibility is out there. That is known. The 24-core chip is coming in November. That is known. No motherboards have been shown off for TRX40/80 or the workstation boards, even though B550 boards are already being shown marketed toward the upcoming APUs. All of that is known.

Taken together, it suggests just the opposite. Zen2 would likely work on X399 and the refreshed TRX40 boards. But if the rumor of no backwards compatibility with the TRX80 boards and chips are true, if they are moving the entire line to that standard, which may make sense if the I/O chip is also to now be manufactured on 7nm in 2020, then you would not get two generations out of those boards. You would get one.

Cascade-X you also would get only one year if not on X299 already, so not like it isn't similar for both platforms. That is why I focused in pricing.

This is why I said it is time, before November and the release of cascade-X, for AMD to become transparent as to what is going on with the socket and chipset changes. This would clear up this speculation and allow for planning for platform changes. Your assumption of Zen 2 and Zen 3 support is no more probable than my statements, and that is the problem. It is uncertain what they are doing and they need to come clean to the market on their plans.

As to optane DIMMs, no one cares about them. What is more important is whether the memory controller is up to supporting 32 and 64GB DIMMs, especially since 32GB is now mainstream and due to the ram prices coming down, optane DIMMs are overpriced and don't deliver on what they are. Until Intel gets the price down, they are worthless. Instead, fill up on high capacity ram and make a ram disk or use the ram to cache for a disk. Cheaper and just as good. If you are worried about the volatile nature of ram, then you need to examine your energy infrastructure, including your PSU and looking into a UPS.

And TRX80 is never coming to a laptop. Just be happy that possibly 6 and 8 core Ryzens are coming, along with potentially a 12-core 3900 for DTRs. But the power consumption of TR chips alone, and thereby needing more cooling, suggests it will never happen. Laptops are barely working with Intel's 8 cores. Considering no solid information on a 16-core TR on Zen 2 is out there, that means you are talking a 24-core+ chip being put in a laptop. That just isn't happening any time soon.

 

optane dimm is a huge deciding factor for me. maybe not for you but having 1500MB/s random read and random write as OS drive is like everything running on DDR4 ram disk, except its storage. and if i dont wish to sacrifice quad channel memory for high core count then i'll need 6 to 8 channels and intel won't have that until few yrs later.

so i will need the TRX80 from AMD and that will need to last and allow OC. if not then it won't be a good purchase especially given intel dropping their prices on HEDT, can prob just go for intel HEDT in 2-3 yrs time.

 

Just get a NVME pcie 4.0 drive and be done with it. https://premiumbuilds.com/guides/best-pcie-4-0-nvme-ssds-for-ryzen-3000/

 

He want high 4K random Read/write speed. PCIe 4.0 NVMe SSDs 4K speed ain't much faster than what we already have from PCIe 3.0 drives. See... 4k for OS, sequential for moving large files.

 

And the practical impact of that on performance is minimal. What, you want fast boot and shutdown times? Literally, it is throwing away money you could use on so many other things that could benefit your productivity more, like 4xPCIe4.0 nvme.

It's just like how optane was sold as the best new storage, then turned out to either be a cache drive for HDD or only needed in niche markets like 8K video rendering, etc.

If this came out two years ago, I'd still be on board, like I was back then. I still support optane drives as cache drives, but using like 128-256GB striped as a tier in HDD raid storage.

As for DIMM optane, hard pass with ram prices and DDR5 in 2021.

 

same thing can be said about TR 24/32 cores, the practical impact on consumer front for performance is minimal. i'd have more use for optane than i'd have for 32 cores as my video editing, vmware and extract/compress are literally the only thing i use and they make up maybe 2% of my daily workload. where as optane, no matter what you do you get gains. gain could be big to minimal and that is entirely on software optimization for the hardware.

both optane dimm and TR will cost big money and both rely on software optimization, i just dont see why you'd argue with the double standard.

if i can use ddr5 as a boot/storage device when that becomes available i'll go that route instead of optane.

 

But AIC optane does not read 4k files at 1.4 GB/s, more like 600K where good NVME pci-e 4.0 drives are now, so..

Below are not modern PCIe drives just showing optane speed is all.
https://www.servethehome.com/intel-optane-hands-on-real-world-benchmark-and-test-results/

 

If 4k speed is you bottleneck you haven't much other choices than jump on Optane. All is up to your usage. And PCIe 4.0 NVMe SSDs won't give you any benefits vs. todays PCIe 3.0 NVMe SSDs (Only if you moving huge files).

 

So you want a year or two for that, while paying more than the cost of a PCIe optane card for the DIMMs with less storage capacity on the hopes that it will help. Cool. Then why are you running the OS on it. That is the least of your worries (except on VM) when it could be used for that video render scratch disk (which I cited as an advantage for 8K which fills up memory and overflows to the disk, which then speeds up the render significantly).

Also, even if only 2% of your work, if it is the largest portion of where you can speed it up, then it makes sense. I say that for the 24-32 up to 64 core chips or for the optane DIMM. There will be niche places where it is needed. Not wanted because numbers, but needed.

What I'm asking you, which you don't have to tell me your work as this is meant for your benefit, not mine nor my satisfaction, is whether there are alternative changes you could make for the amount of money it would cost, which is about the cost of a 24-32 core AMD TR, that would make a greater improvement in QoL or productivity. If there is, spend the money there. If not, and you have the cash, don't let my critique stop you. But think critically before you spend the money on it as you may be underwhelmed if your use isn't a niche use that truly gets great benefits from it. Hell, picking up a spare 18-core makes more sense after cascade-x drops.

Of course, this is depending on the size of the optane DIMM you are talking about (I'm assuming one of the larger sizes, not the smaller ones here, which the 128GB is $695, or the cost of a 12-core Intel CPU or nearly that of the 14-core or AMDs 3950x, while the 256GB is $2595, or the cost of building a 32 core rig almost from AMD or nearly Intel's OCable Xeon, while the 512GB is $7,816, or more than the top end 64-core Epyc, where you could grab the lower P single processor for the $4000 range, then build a server around that for that price; hell, even pick up another GPU if your software scales rendering with additional GPUs). That pricing is going off of announced pricing in April in a Tom's hardware article, so pricing may differ, but point remains.

 

My point again is at 600k for 4k files transfer rates NVME drives can do that too for much less and larger storage too.

 

sure, but with the difference being that nvme drive do those numbers under ideal conditions for a short while, optane doesnt care about the type of workload, it just stays constant.

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

 

Once again, is it worth it for the price of a 2080 super, a 12 or 14 core cascade-x, or a 3950X for a 128GB optane DC DIMM?

 

haha that is a question that everyone has to answer for him/herself i guess

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

 

If you need that type of performance you need enterprise hardware and pay accordingly.

 

Hell, you could even get a PCIe 900P optane drives 480GB for that, or the 380GB m.2 905P, or a 2TB PCIe 4.0 for $450, or 4x500GB PCIe drives for $130 each, then raid them for massive throughput (like 15-16GBps on 2TB of storage). And that is just compared to the 128GB DIMM. With the 256GB DIMM, you could raid 4x2TB PCIe 4.0 drives and still have hundreds to spare.

 

not to mention that flash performance dips like crazy when doing read and write which is basically, everyday workload. where as optane barely changes at all.

it is worth it, i'll take a 3070 and call it a day. zen4/5 TR 24 cores or 18 core tigerlake HEDT + 240 or 480GB optane dimm, with either 4070 or big navi 6700/7700, with 128gb of ram.

 

As long as it is right for you. Just wanted others here to realize it's use case on truly benefiting can be limited and there are numerous alternatives that may be more meaningful upgrades, including a larger sized optane PCIe scratch disk. There are definitely use cases for it.

But I would also be remiss if I didn't mention that 2x128GB 2933 DDR4 ECC DIMMs can be had for around $2300, meaning 8x128GB (1024GB or 1TB) 2933 DDR4 ECC DIMMs can be had for $9200 (plus tax). That means going to an 8 channel server chip could put you at more memory to use as a ram drive, which may be an alternative to consider. And when DDR5 comes, I think speeds are stating around 3600-4200.

But, yeah, just trying to give alternatives which may be viable (that 1TB of ECC at 2933 looks better to me than 512GB optane DIMM at close to $8000).

 

i think 8 x 128gb maybe over kill but if i have it, i can probably make use of it. and i would never spend that much just on ram LOL.

i'll probably dig something like 4 x 32GB config and leave additional channels for optane dimm.

 

How about 8x64GB ECC at 2933 for $2700, approximately? That's 512GB of ram much more reasonably priced, and comparable in price to a single 256GB optane DIMM or 4x128GB optane DIMMs (512GB of optane, which I do not know how many slots, etc., can be used with optane). ...

 

7nm Supply Showdown: AMD, Nvidia May Fight for Scraps as Apple Reportedly Ups A13 Production tomshardware.com | Oct 04, 2019

TSMC is Apple's primary fab and is tasked with manufacturing the A13 Bionic. Apple's purported order for another 8 million smartphones probably won't help matters. Nikkei Asian Review said the company was originally conservative with orders related to the latest iPhone models. Demand has apparently surpassed expectations, though, especially for the base model iPhone 11 and iPhone 11 Pro (Apple reportedly told suppliers to make fewer iPhone 11 Pro Max units).

That could pose a problem for TSMC's other customers as they pursue their own 7nm ambitions.

TSMC 7nm process attracts multiple orders for AI processors digitimes.com Oct 04, 2019
TSMC with its competitive 7nm process technology has obtained orders for made-for-AI chip processors from Nvidia, Xilinx and Qualcomm, and a number of China-based vendors such as Huawei, Bitmain and Alibaba, according to industry sources.


From my old post...

 

optane dimm pricing wont be that absurd especially if its next gen we talking about, more capacity and better endurance, higher performance too.

i can't use ram as OS drive and even though i could use it for caching, my experience tells me caching barely helps because software optimization is just so bad. optane would essentially boost everything even with poor optimization, its not that much from ram speed tbh, may be slower in bandwidth than ddr4 or ddr5 ram but should be on par with ddr2-3 already.

restore image, install software, install windows, MS explorer search, gaming and everything else that i havent think of can take advantage of optane. also optane pair up with fastcopy is awesome, files written directly into drive as it has no dram and i can know what true speed really is.

 

Except for the one article mentioning Nvidia which inked a deal with Samsung for 7nm, Apple has always been the largest fab time for TSMC in the fall. But, as mentioned, AMD signed up for more fab time already, which is evident to anyone seeing them change the revenue for the year for Q3, but leaving overall guidance for the year unchanged.

Not only that, those deals are likely why the 2 month lead time changed to 6 months. Has that 6 month time frame changed with the public announcement of this order? If not, this isn't new, just an explanation of what has already been reported, giving more context.

As to the other vendors, TSMC was only at like 80% capacity for 7nm in either the first or second quarter of 2019. At the 6 month lead time, that means those other starts, if not inked and baked in, are taking up that percentage or the extra capacity as TSMC moves lines over.

Once again, without more, this is much ado about nothing, other than AMD shortages continuing unless AMD also already inked a deal for more fab time already and it not having been reported.

I do wonder if Apple's order has to do with the recent news that Samsung phones are fully pulling out of China, meaning Apple is trying to take their market share instead of giving it to XaioMi or Huawei.

 

I haven't seen any evidence about Optane being beneficial for gaming or really much general computer use. To be fair it's been months since I've done any reading on it, and that was the m.2 variant. If you have some good sources which actually demonstrate it, I'd love to see.

 

https://www.extremetech.com/computi...-provide-only-modest-performance-improvements

https://venturebeat.com/2019/09/20/optane-persistent-memory-breakthrough-or-broken-promise/ (take into account this author doesn't mention that Epyc is disruptive, making it possible to replace 4xV4 Xeon systems with a single 64-core Epyc chip while AMD supports up to 4TB of memory instead of Intel's 1TB (unless buying their memory variant for obscenely more money), thereby negating the claims of lowering TCO as it may be lower switching platforms altogether, especially after security mitigations are factored in, lower energy costs, etc.)

https://www.zdnet.com/article/first...s-show-benefits-and-limits-of-intels-nvdimms/

Remember, Intel's first to the NVRAM party, but them trying to force it only on Xeons will backfire. AMD supports NVRAM standard set by JEDEC. But because Intel has locked Optane as proprietary, AMD doesn't support their chips, but will support variants coming from other players in the next couple years.

Only reason this came back up is a week or two ago Intel did a press event saying they have more new versions coming (see vaporware, like Intel on so many products as of late).

It is only for the server community, which if you ask me, why buy a Xeon at this point at all when you could have Epyc, which costs less and crushes Intel's offerings in many workloads.

If you need it, you need it. I find the conversation at this point arbitrary when looking at other benefits once forced into the server realm.

Edit: just saw Intel is doing an X299X workstation variant that supports optane DIMMs. But the only examples they showed were Spec Workstation Maya and handbrake, which you can easily fill memory when using which requires then using virtual memory, which is disk space. Once that happens, as in the 8K video encoding example I pointed to earlier, you get a hit in performance. So this had 4x the performance for Maya and 3X on handbrake in their controlled example, while not telling readers how much memory was available in the non-optane example.

If you look at optane reviews on daily loads, it did nothing. Even looking at SSD sata to nvme, load times are minimally effected. I'm doubting that changes with optane here as well. This is why it is something I consider trivial for most users, especially outside the professional and data center markets.

 

So, looks like Zen 3 will have 8 cores per CCX on a chiplet, instead of two CCX per die! That cuts the latency from going to the I/O die to go to the other 4 cores on the same die. Very exciting.

But Zen 4 may be until 2020 (guaranteed 5nm or 3nm then, depending when 3nm comes online).

So seems like Zen 3 is a buy and wait CPU, as DDR5, and PCIe 5, won't be around until 2021 at the earliest. That also means no CXL or Gen Z interfaces until then.

So, if you haven't upgraded yet, skip both and get next year's offerings. Doubt Intel will be off 14nm for desktop so likely recommend Zen 3, but will be the year of GPUs with a 3-way battle on 7nm or equivalent.

 

they need to first support optane dimm then i'd be happy, which intel prob won't give this advantage up so only hope would have to come from micron's 3dxpoint for their quantx SSDs.

i'd be wanting to see zen3's latency improvement and overall IPC from this change i'd be interesting, and i agree on zen4 probably the better buy.

i'd still drop and bu if something excellent comes to laptop space though, not the regular 10 cores 14nm++++++ that uses 300w in a laptop with a 125w heatsink.

 

Since there is like no commercial advantage of offering Optane support I doubt you will see any hurry to implement it specifically. Since the pci-e lanes can be provided individually and too the CPU if an external Optane device is ever offered you could use that. I doubt there would much of a market for that extra speed at that cost. With such a small market this is why there is no rush to support.

 

I already explained why optane DIMMs don't matter. I even gave articles that mentioned alternatives to optane and Micron's version that are coming out in the next year or two. Third gen optane is in 2021.

Also, for Zen 3, what a single 8-core CCX means is that all 8-cores operate similar to a ring on core to core comms. You still take the hit on memory calls, but they even hinted at possibly increasing the L3 cache more, with each core now having access to the full 32MB of L3, rather than the 16MB that is in their CCX. If you don't understand how that will speed up performance after my lecturing you on memory and cache keeping fed, I don't know what to say. In fact, I'm thinking they have to widen some things further, including new instruction sets, to combat having the L3 that large, possibly AVX-512 or some new instruction set, as when you go larger, the table search of the cache takes longer and you have to overcome that latency on that cache level.

On laptops, go look up Splave overclocking the 65W 12-core 3900. At that TDP, there is a chance of that beast getting put in a laptop. Intel would not really get close if they design the cooling system half as robust as they do for the 9900K systems.

 

You explained squat.

Since you're incapable of understanding why they're important, please stop the lectures of something you don't understand.

In a year or two is when 3rd gen Optane will be available... Hardly a reason to wait for something else that may or may not be better than what current Optane offers today in Intel systems.

 

LoL, go for it then. If it truly will make you money better than what A faster and more powerful AMD solution then go for it and spend away. The market for it may be small but it is still there, so be it. Just understand that 99.999% of users may not agree with you though as their usage will not be that extreme or specialized.

 

Put it the other way... The manufacturers as always design the cooling around the chips TDP. And for an locked down chips forget much cooling headroom (in same way as the comparable locked down Intel chips). If they used the 3900X we would probably see cooling capacity above rated TDP.

 

It's not locked down. You can overclock it, unlike Intel. But using the built in boost won't go as high as the X variant.

 

I expect the notebook manufactures will put in line with locked down Intel mobile chips. Notebook aint desktops Bring the X and they will handle it as the Intel K/HK chips.

 

for laptop with AMD cpu it is better for cooling because AMD's TDP isnt just base clock and much more power efficient, and call it 105w. where as intel's TDP is base clock. if OEM design based off of those value its clear we'd get more performance out of AMD.

if you OC through AC or other means then those don't matter i guess.

i want a 16 core CPU with a fat heatsink in laptop and none of these companies will give us just that. they really pushing us to desktop.