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

Well, businesses must test for software compatibility, work flow, etc. We've talked about test beds a lot recently, but a similar process is used for office deployment on regular workstations (not meaning HEDT workstations needed for specific work, just regular client computers on the network for employees to do their work). Considering no OEM systems were available until the beginning of the month, unless you were buying boutique or from some specific system builder, no one was really looking at commercial deployment of Ryzen. Also, you have to test the exact system considered, which further limited the opportunity for evaluation for deployment.

So, in all honesty, there should be few unless specifically requested to IT for a reason and the company was willing to give that employee what he wanted, but that is a rare occasion. In regards to large companies and offices, that means very little to date. They also potentially passed on evaluation of the new OEM desktops in favour of waiting for Ryzen Pro, which seems to have more features favorable for security and deployment. But that is a ways off before release.

Then comes the lack of iGPU on these chips. Some companies don't need an extra dGPU for what the computer is used for, and that adds to cost. It is somewhat mitigated if the MB integrates lower end graphics into it, but this shouldn't be ignored entirely as trivial. They have to compare the Intel versions of what they have, and what they are considering for deployment next, with the Ryzen equivalent client machines (server client, for clarification).

In that regard, I can definitely agree that your hardware suppliers have not heard much on Ryzen. Considering corporate use and deployment is the largest market (not just server chips, mainstream chips as well), it does cut against wide adoption. But that may change in coming quarters with OEMs now on-board and corporate aimed chips on their way.

But it is a good point to make and a fair point on adoption.



Sent from my SM-G900P using Tapatalk

 

<If one has dipped their toes into it ...> Exactly, *IF*, LOL I've always said and done... I don't want to be another's GP. You learn when you get old as me, HaHa
And if you are on Z170/270. Be there. No reason jump for Z370 mainstream https://www.pcgamesn.com/intel/intel-z370-chipset
I do not trust anything that is on web sites, but Z370 is a dead end, for us who is on Skylake / Kaby. Next year will be better. This year is a mess. At least if you look for Intel(mainstream, as well also HEDT).

 

oh man i thought z370 is a new socket change was it not? that means it'll support coffeelake 6c and possibly tigerlake 6c?

 

If Intel does as usual, there will be 2 processors for socket 1151 v2 (If not, a lot of angry people). But Z370 seems to be short-lived. Coffey will have a short life before the next arrives. If you are on the Z170 / 270 now, I see no big reason to upgrade. 6 core ain't a good enough reason as I can see it. Intel need to pull out better than this 6c Coffey, if they shall compete with Ryzen 1700-1800(X) in mainstream. Expect sooner than later. AMD has turned around the computer market.

 

So, I think there is some confusion. I'll try to clear it out. Intel DOES NOT USUALLY DO MORE THAN ONE GENERATION PER CHIPSET. Mainstream only gets one generation while the HEDT and server lines get two generations.

The reason for the confusion dates back to Haswell. It took SOOOO long to get to Broadwell, they released the 4790K, also known as Devil's Canyon. This brought with it an updated chipset and feature set, but also forward compatibility with the ****e-stain that was Broadwell with the overclocking "c" processor. Then we got Skylake. Because Kaby was an intermediary step because cannonlake was not finished, it was a stepping stone like the 4790K. So they released it with compatibility of both the Z170 and Z270 chipsets, with the new chipset containing the new feature set, like Optane support. We now are looking at the Z370. This is the Coffee lake chipset. Coffee will not work on Z270, unless they recently changed something I'm not aware of. We do not know if it also supports cannonlake.

Now for the CPU generations:

Haswell -> Broadwell -> Skylake -> Kaby lake -> Coffee lake/cannonlake (coffee is cannon on the 14nm++ process; CL on 10nm) -> Ice lake (10nm+) -> Tiger lake (10nm++)

Now, we do not know if Coffee or CL will have IVR like Skylake-X, but according to old reports, Ice and Tiger will have FIVR. Because it is fully integrated on the Ice and Tiger, there is no reason to assume backwards compatibility with Coffee or CL, because at most those are only partially integrated voltage regulators.

Hope this helps!

 

thats right, they usually dont, but they did for skylake -> kabylake simply because there are so little to no improvement no one would bother buying new platform or cpu had they not done so. i'd assume coffeelake might be similar however, the following cpu after that its gonna be 10nm so they might not.

 

I understand this. But if I remember correctly. Ryzen will have more Cpu revisions than 1700/1800 on same socket. If Intel push only one gen mainstream CPUs from now on and you will need a new socket every time, I would think people will look more on AMD. If I have correct, will Intel push a lot of new and old customers over on AMD? Hope you understand what I meant.

 

I get what you mean. But Intel is Intel. The best they could do is coffee and cannon, the ice and tiger. Meanwhile, AMD has suggested n-2, n-1, n+1 socket compatibility moving forward, with the socket layout being the same as AM4.

What I believe this to mean, since Ryzen is used from now through the 7nm refresh, is that they plan on an AM4+ socket. This makes sense as DDR5 standard will be published next year and GloFo already announced the 7nm chips will support DDR4/5, but a slight change in connections might need made for compatibility. This also suggests AMD wants to compete by adopting DDR5 fairly quickly, something we saw lag with them on DDR4 adoption.

But that's speculation. Considering there are rumours of a Z390 chipset and rumours of a chipset different from Z370 for cannonlake, I'm not holding out hope.

Sent from my SM-G900P using Tapatalk

 

People, please keep this on topic, if you want to discuss Intel specific platforms and formats please use the Intel specific thread.

 

I think there at the end @Papusan and I tied it back in. Just had to clarify some information out there before we got to the heart of the question: which company/companies will have multi-generational support on the same socket/chipset.

We already knew AMD in that regard, but had to clarify on Intel to get a better comparison on platform life....

Sent from my SM-G900P using Tapatalk

 

Sorry, maybe this was misunderstood by me. I actually compared platform vs. platform. Because Right now, Ryzen does not have competition and is Nr. 1 on mainstream, as Intel is still with 4 core i7 as best. This is actually Ryzen vs. Intel on mainstream. But Sorry

 

TR gonna be good that large socket a beauty. if zen was to ever catch up to skylake's IPC then intel's lead would be completely over, of course unless reviewers still love to bench intel's way at 5ghz which only intel's cpu is capable of. intel isnt exactly stupid they know full well what and how to make a system faster and take advantage of their better silicon quality, however AMD too is smart by increasing number of cores. on the other hand as core number increase, frequency naturally go down due to power consumption, heat, so even if intel wishes to battle AMD in high core count they cant really use frequency as an advantage anymore, only IPC.

currently, once zen catches up to intel's IPC then intel becoming less and less attractive. however on 18c turbo boost 3.0 with 2 cores still damn good though.

 

So you are saying Zen at 7nm is when your interest shifts? Zen+ or Zen 2 should have 20% IPC increase. That plus faster speeds (GloFo is shooting for over 5GHz for 7nm) should really be good, especially with the 7nm GloFo process being as dense, almost, as Intel's 10nm. By doing multiple chips so large, you have less heat for sandwiched cores, while also spreading it over more space under the IHS seems smart.

So, you may want to skip this year and wait for next. Skylake-X isn't a large improvement over the BW-E. Meanwhile, if they make a second gen board with DDR5, depending on cost of that ram, you could be looking at updating it all again soon.

Sent from my SM-G900P using Tapatalk

 

thats not what i think at all.

what claimed 20% is likely the overall performance boost from shrink, optimization and ipc and "should" and what AMD claims, i'll believe it when it comes out. people would think its possible with glofo etc but do they honestly believe, intel as big as it is and as experienced as they are in foundry business would let others take over, especially they have so much resources. if its me in intel's position i would never let it happen, not to say that mistakes cant happen.

on the other hand, node shrink doesnt provide much IPC improvement that much anymore, mostly it came from optimization now adays to remove unneeded loop in design. zen2 will be based off zen, so firstly ccx issue wont be going away until zen 3, 2nd is that its still the same architecture with a bit of optimization. unless zen has a lot of place to optimize i can't see it happening tbh.

intel on the other hand is very experienced in optimization, just look at what they have been doing for last 5-6 yrs, little to no IPC improvement we can literally call them optimization each year. the skylake we have now is not much different from sandy lol. with added a few features, dmi 3.0 and new extensions thats about it.

 

Links;
http://marketrealist.com/2017/06/whats-amds-strategy-to-increase-adoption-of-epyc-server-processor/

 

Are we sure that Zen's IPC is ALREADY not on par with Intel?
The only advantage Intel currently has is higher clock speeds.
The rest shows us that Zen is maybe 10% behind Intel in certain benchmarks, and most of those numbers were taken with unoptimized software.

 

Wrong! First, you put WAY too much faith in Intel, especially after the BW die shrink and delayed 10nm process!

Second, the biggest fab is Samsung! They produce nand, dram, arm, etc. Their processes are wonderful, which means you don't know enough about the industry. TSMC is the second largest fab and have a 10nm working right now, as does Samsung. GloFo has been working with IBM, in fact taking over their fab, and with Samsung. So, while all other large fabs have their 7nm, Intel is on 10nm and won't be shrinking until the others go to 5nm.

Third, transistor performance for the die shrink is estimated at 40%, if power is kept the same. That means far over 20% ipc claim. Further, 25% of that comes from the speed increase alone, which would be estimated at 1GHz. It also has 50% more transistors per area, which should also give a nice boost. So your calculations are off by a fair amount.

Fourth, this is a new architecture, which means easier for larger IPC refinement. Look at the first i Core to SB. Not as difficult as you make it sound.

Fifth, Intel is now hiring for the icore replacement design team. They know they are at the limits of this architecture.

Sixth, the assertion of CCX makes no sense. That is part of the core operation and design. It is more likely if you are saying IF2 will be ready for Zen 3, not Zen2. Also, it isn't an issue. Look at how optimizations in software have erased most performance differences with the 7700K.

Etc.

Sent from my SM-G900P using Tapatalk

 

wrong, you put too much faith into AMD. with all the stuff you glorified so far, only thing stuff thats available on the table currently is intel's cpu, which has higher IPC, higher frequency, the cons include price for performance and power consumption, which consumption im willing to accept if it gives me a good whopping 4.7-4.8ghz on 10c, AMD can't give me that, neither can Zen2 and i seriously doubt Zen3 could too. but time will tell.

 

I am not putting my faith in Intel, I'm putting my faith in IBM, a company that regularly produces 5GHz+ processors. I'm putting my faith in adapting the architecture to that process. I'm putting my faith in getting data on where the IPC improvements are at in the b1, b2 steppings, any other stepping coming with adapting to 14nm+, and then to 7nm. That is potentially three architecture refinements from b1 Ryzen, 2 steppings for TR and Epyc to 7nm. If Intel can get 5-10% per refinement (less now, or at least Kaby), there isn't a reason to think AMD couldn't get IPC to skylake with those revisions. That plus what comes from going to 7nm, I'm betting even to edge to AMD at that node against cannonlake.

Sent from my SM-G900P using Tapatalk

Edit: also, with the boned boards because Intel lied about power specs, good luck with that!

 

Clock for clock the Zen B1 and B2 stepping are behind Intel's current offerings on single threads. This is fine though as it is the nature of Zen and multi thread and brute force is where AMD is shining and at least clock for clock running with Intel.

The main thing being is for apps today it has been proven, even going back to Gen 3 of the icores, that the IPC is fast enough. It is other things like SSD that have sped up the desktop. For everyday use you will most likely not notice a difference between an I9 or Ryzen system. So having either should work fine, now it is just a matter for OEM's to get the box's out.

Be it individual or business consumers the majority of purchases will be OEM system. Once those come out there will probably be a war on as the Zen, for the money, is much more of a work horse.

 

Overpriced Disappointment - Intel X-Series i7-7820X Review
(New i9 7820x vs Ryzen 1700)

 

One thing he does not mention is the price premium of the x299, nor the fact of the overheating VRM issue.

 

The VRM issue is pretty new, and if you aren't looking for it specifically, the nicely shrouded area on the motherboard make's it look aok. Most people wouldn't realize it until too late, poof!!.

Is the price higher than x270? Intel boards are always pretty high priced. Maybe he thought that went without saying...

I would think the range of boards would offer cheaper options with Ryzen, but when you start loading on features and options, the Ryzen boards can get pricey too, but maybe not as pricey as X270 and X299...

 

https://www.newegg.com/Product/Prod...3144053&cm_re=msi_x299-_-13-144-053-_-Product

https://www.newegg.com/Product/Prod...3&cm_re=msi_pro_carbon-_-13-130-973-_-Product

And AMD Version;
https://www.newegg.com/Product/Prod...7&cm_re=msi_pro_carbon-_-13-144-017-_-Product

 

You selected the highest priced MSI x299 and the lowest priced X270

Picking the highest from MSI motherboards, the X299 is $100 more, but it is a new release that hasn't had time to go into full production (thank goodness), while the x270's have been out for a while and already had price drops.

Here are the most expensive x270's:
https://www.newegg.com/Product/ProductList.aspx?Submit=ENE&IsNodeId=1&N=100007627 601292371

and x299's on newegg:
https://www.newegg.com/Product/ProductList.aspx?Submit=ENE&IsNodeId=1&N=100007627 601298415

sorted by highest price.

So, given the much more mature and wider range of boards for x270, they have by far and away the highest cost top motherboards...water jacket cooled

The average x299 price seems higher...staying above $200.

The AM4 motherboards are much more reasonably priced:
https://www.newegg.com/Product/Prod...6&IsNodeId=1&bop=And&Order=PRICED&PageSize=60

 

Actually, he was making the point that the higher priced X370 boards with the R7 1700 still are less than the processor alone. If you add in the cost of ram, you now have a system going for the price of the MB and the CPU of an Intel rig. With that extra, you could even step up the videocard game a bit to the next level, say 1070 to a 1080 or a 1080 to 1080 Ti. Just saying, it's the better value.

If you want, I could price you out two killer systems, although because of the ram costs currently and return on benefit, I'd be looking at 3600MHz ram for the average build. Personally, even if I couldn't hit the higher speed, I'd keep my 4133 and downclock like I currently do to 3733, then run it tight at CL14 (7.56ns latency,
0.54ns cycle clock time, approximately, it's kinda nice!). But, point is, aside from that, cost wise, it still makes sense to get the 1700, then either upgrade the cooling or the graphics card. That may change when ram gets flush again and prices drop (or DDR5 comes on the scene and we get 4200 as normal).

 

Yup, that was his main point, for the price of the 7820x alone you can get a 1700 + nice motherboard, with some cash left over

The x299 boards are a no go until they get better power and cooling for power anyway, so waiting for the higher core count CPU's to release and bring those x299 redo's sounds like a good idea - maybe even wait another couple of months past that release in "October", if it happens then.

Still waiting for ThreadRipper CPU's, and motherboards... gonna be a good time to build a few months after that shakes out as well.

 

Exactly. You have the 12 core (my value pick on Intel) and TR, then all the board fixes. Something tells me they won't do TR like Ryzen and that we won't get the same rough start Intel had (anything similar on the design of the VRM should be addressed for the X399 boards on these months also, meaning that shouldn't pop up with them, hopefully that is true on supporting the energy needed for the chips, two 8 pins sounds good)...

Sent from my SM-G900P using Tapatalk

 

for an enthusiast (i guess there are many levels) i'd call myself that i belong to the medium pack where i'd spend quite a bit of money for what i want, understand things to a point where it is not TOO time consuming and using them in a more practical way. unlike the top people who loves custom loop water cooling, the extra hassle of learning it, obtaining the material and self diy and time spend just isnt worth it, of course LN cooling as well simply not practical.

i thought a few here would be enthusiasts but then again im probably wrong because theres enthusiasts of cpu/display/ graphics/storage etc. me being in the CPU/storage area can't afford to not go intel, assuming priced correctly which imo they are, coming from intel.

the storage performance from AMD is simply so bad either because they have got no chipset driver or these NVMe drivers in windows just not optimized for ryzen, or there maybe other issues involved.

from my stand point it is very clear, if you want the best storage performance other than paying huge amounts for ramdisk or NVDIMM thats actually consumer practical, the next choice is sub 500GB optane SSDs. putting optane SSD on a intel board vs on amd board, 4k random performance pretty much drop by 30-40%,

 

@ole!!! Most even enthusiasts do not need the storage system speeds you are obtaining here. as far as if that is your requirement then standard Optane SSD's will work on neither Intel or AMD. It is using Optane as a caching for HDD's/SSD's with the dimms that will not be supported.

Now 4k speeds might be a driver issue, but where do you see these results?

 

At this point in time, there is simply NO reason what so ever, at all, to upgrade and/or purchase anything in relation to CPU or GPU.

Too much on the horizon within the few 1-2 month. The price drops are going to be insane, especially once mining Ethereum isn't beneficial with the current GPU's.

 

Do you mean desktop-wise?

I'm planning on buying the ASUS ROG all AMD laptop with Ryzen 1700 and RX 580.

Though, I was contemplating on getting an ALL AMD desktop now and then spend extra cash on an all AMD laptop with Vega.

Dunno to be honest.
A desktop for me would be like going back, but its also a few hundred £ cheaper than a same laptop.

 

Come back to the dark side... We have cookies!

Sent from my SM-G900P using Tapatalk

 

I mean everything. With Coffee around the corner, Vega not completely released, all hardware purchases should be put on a pause.

 

Bah!!, Coffeelake only has 6 cores, we already have 8 core Ryzen coming to laptops first

Vega may require new cryogenic tanks be built into laptops, refilled weekly at the local Cryogenic Emporium, or we can get the "cooling juice" piped into our offices direct to our desks for constant refill's.

We'll need the "cooling juice" for everything soon enough, better start planning on it now.

Maybe we can switch to liquid Hydrogen and liquid Oxygen, and fuel them at the same time?

Hey, speaking of Hydrogen Fuel, why don't Hydrogen Fuel Cells use liquid Oxygen instead of using loud heavy noisy air compressors to get Oxygen into the Fuel Cell as the "oxidizing agent"? Seems kinda bass-ackwards.

 

I do use ramdisk on occasion (benchmarking), but I'm considering going to 64GB instead of 32GB and start looking at ram caching. But storage is not where my primary attention is at the moment. It really is much less fun the CPU and ram overclocking. Graphics, that's my newer area for OC. But I'd rather spend the extra on a water chilled system than be boxed into intel.

I'm looking at keeping my P770ZM with 4790K until 7nm/10nm arrives. The price premium on performance laptops in large, where I can build something better less mobile. This is due to the flux they are doing in graphics cards form factor, as well as having no need until them (great performance jump).

Also, liquid oxygen is fairly expensive (if using the amount needed) and very combustible. By using the pump, it costs less and has the atmospheric nitrogen, making it somewhat safer. Or that is my guess...

 

Don't forget that the shining star of the Ryzen lineup is the 1600 best bang for the buck bro. I know the 1700 is good too, but the 1600 is close enough, and significantly cheaper.

Coffee seems reasonable though, you can NEVER disqualify a good amount of competition! It breeds innovation, better prices, and supreme choices.

 

Are you insane?

Why go through all that trouble?
I think it would be easier to just make passively cooled system made with a combination of carbon nanotubes, graphene and synthetic diamond heatsinks.

If they were able to bring down Polaris from 150W to 65W with a simple underclock that was 'mild' in comparison to stock desktop 480 and undervolt, I would imagine they will be able to do pretty much the same with Vega (bring it down from 300W to say 150W).

The Fury line cards were consuming a lot of power due to much higher voltages. And when Fury X was undervolted, power consumption dropped down to Nvidia levels easily.
If AMD is using the same approach for Vega to keep the yields up, undervolting will likely be doable, as well as bringing it to laptops with LOWER TDP than what OEM's needed to do with Nvidia.

As I see it, Nvidia right now has desktop grade GPU's with same TDP.
Asus was able to bring down desktop RX 580 TDP by 2.5 times sacrificing maybe 5% of total peformance and undervolting - so on the up and up, AMD looks to have BETTER suited hardware for laptops.

Although this is all just speculation since we don't know how different to Polaris Vega actually is.

 

Either we evolve past the thermal problems, or we succumb to them and become mired down in them.

It should be fun

Here is my inspiration for the Hydrogen Fuel Cell thoughts and using Cryogenic Fuels for laptop cooling

Spoiler

Is the Chevy Colorado ZH2 the new Humvee? We Go For a Ride!
"Compressor noise" first shows about 1:10

The Chevrolet Colorado ZH2 fuel cell time-lapse build

Chevy Colorado ZH2: Is this Hybrid Hydrogen-Powered 4X4 Truck the Humvee of the Future?

AUSA 2016: Hydrogen Fuel Cell powered ZH2

 

Considering how high Intel usually prices their hardware... the overall difference between current lineups and Ryzen when comparing performance is nowhere near high enough to justify going with Intel's more expensive options.
Especially when you take into account that software is barely starting to include support for Ryzen (and when it does, it EASILY eliminates the gap and even overtakes Intel).
I'd rather support AMD at this stage.

 

Yup, should be fun to see how they compete, BGA low TDP vs socketed medium TDP, 6 vs 6 cores and 6 vs 8 cores, with Ryzen 1/2 the cost, AMD FTW!!

 

Are they vegan cookies?
If so, I'll consider it.

And which dark side are you referring to?
Desktops? Or laptops?
We ARE on a laptop forum.

 

Ah, the joy's of "meat" cookies, gotta make some of those for the 4th.

Some call them Hamburgers

 

I'm sure some could be made, vegan is pretty evil afterall... (j/k)

And desktops are the darkside in this context... We may be in a laptop forum, but we are also discussing desktop chips.

 

Yeah, but regarding laptops... Will the best from AMD give us cheaper laptops due cheaper silicon? Would you buy laptops if it was cheaper due AMD or would you look on performance? For me is best possible performance still the main choice - reason if I buying new hardware.

 

Even with the desktop CPU, the laptops are not much cheaper. I'd consider a Ryzen desktop CPU if the firmware isn't borked. But the graphics card premium is a killer and the flux in upgradeable gpu parts is why I'm waiting. I want performance, but the recent bs from companies means I'm waiting for the dust to settle. Also, even though I want power in mobile, I'm going to stick to single card because it now makes sense to build a powerhouse desktop, and have a powerful laptop, but not at the level of what I would have done before the gpu crap. Then the Q-max crap and the Pascal flashing now...

Sent from my SM-G900P using Tapatalk

 

Well, with Asus Rog, we will get all AMD desktop level performance in a laptop with an RX 580 to boot.
No OC-ing of course... but that doesn't mean we can't try it anyway to see how the cooling might take it.

The only 'downside' here is that the GPU is mid-range.

You do end up paying more than usual for having an 8 core cpu in a laptop though... I think the prices should have been £200 lower.
£1200 for Ryzen 1600 laptop and £1400 for Ryzen with 1700 option.

Then again, the current pre-order prices are just that... pre-order. They might not be set in stone.

So, if I do get an Asus ROG laptop with 1700 inside... question is, will there be a need for me to get a summit Ridge laptop when it comes out?
Probably not.

The laptop would be no different than a desktop... though the desktop would have an upgrade path available for Ryzen 2 and 3... if the CPU's remain on the same socket/chipset.

But if the laptops won't be BGA, then we will likely have the option to upgrade to Ryzen 2 and 3 as well.

 

You may not be familiar with Asus laptops, but the firmware and laptop boards are NOT like the desktop boards. They also may include a whitelist, meaning no change in CPU without a modded firmware. With MSI you have SVL7, with Clevo you have Prema (Love his work, cannot state that enough). I don't know anyone doing custom Asus firmware. Without that and guarantees of an upgrade path, I have ZERO interest in it. If someone confirms the non-whitelist (by throwing an 1800X in it, for example or getting a statement from Asus directly of a Zen 2 &3 upgrade path), then I'll change my opinion. Also, if they solder the GPU, until I get the equipment to reball, and have a custom firmware to troubleshoot, no interest in the slightest.

 

Here's the thing... most laptops can't be upgraded anyway... and the CPU differences are marginal at best unless you are jumping from say an 4 core to 8 core cpu, or even 12 or 16 core one that has a similar TDP (which can be tolerated by the cooling system).

I am aware of usual laptop bios limitations in microcode listings.
However, it's supposed to be B350 motherboard... and in that sense, it might actually support future processors, provided the current ones aren't soldered and Asus doesn't blacklist them away.

See, we NEED the unlocked UEFI/BIOS that's as similar to the desktop as you can get.
With same microcode updates that provide fixes and performance enhancements (because it would be stupid for future microcode updates to come to desktops only and not to laptops).

Perhaps we can write to Asus and let them know that laptop upgrade paths would be desirable considering the price, sustainability, etc.

Actually, we could 'sell' the idea to Asus on sustainability. Why waste resources into creating a whole new laptop when they can offer proper upgrade paths for their hardware.

But this would likely mean inability to go outside TDP limits (given inherent limits to outdated cooling techniques we have to put up with) - although it should be more than enough to give us a sizeable upgrade path to Ryzen 2 for example on 7 nm that might even be able to shrink down some of the TR's to 65W.

 

The silicon is certainly cheaper... but the laptops themselves with Ryzen are not if the pre-order prices are anything to go by.

Ryzen 1700 for example is CHEAPER than 7700HQ. They are clocked comparably even and 1700 would blow 7700HQ completely in multithreading (and hold its ground in single threaded).
The motherboard in Asus ROG laptops will be based on B350 (a cheaper motherboard).
Keeping that in mind along with everything else, I wouldn't have priced the ROG with Ryzen 1700 more than £1300 or £1400 for example.