AyMD ThreadRipper

>>60734321
Enjoy being a corelet.

>>60734321
>r9 1998x 16 core/32 thread 3.9ghz @ 155W
>Inlel Skyisfalling-X 7920 12core/24 thread @ 160W

INTEL HAS RETURNED TO THE ORIGINAL HOUSEFIRE MEME

>>60734321
The annouced tdp is around 120 watt no ? It's not that high

>>60734346
No it's 155

>>60734352
Yeah i saw that
These cpu must need a pretty big rad

>Intel
>165 "TDP" (since everyone knows Intel lies like shit about TDP)
>actually 250W
Better get life insurance for your neighborhood while you're at it.

>>60734327
Enjoy waking out of a comatose with third degree burns and no house.

>>60734381
AMD is probably lying too though, there's no way 155 is enough.

File: DA-Iz0kUAAADyE7.jpg (66KB, 1024x576px) Image search: [Google]

66KB, 1024x576px

>>60734321
>>60734352
>>60734376
>>60734384
Threadripper has a massive contact plate which means better heat dissipation, and the dies are soldered unlike Intel's toothpaste TIM. Cooling should be no problem.
>>60734394
Threadripper is a higher stepping (B2 vs. B1 on Ryzen). And they're cherrypicking dies because their yields on the 8 cores are 80%.

>>60734321

>32 threads

>Faggot devs continue to consider their application multithreaded if they do nothing but dump audio and background asset loading to another thread on another core, leaving 30 threads idle most of the time

hue

>>60734321
>massive die size
>fire hazard
The larger it gets, the better the heat dissapation, the cooler it should run.

Wouldn't this cause massive latency?

Those are some big threads

>>60734321
>die the size of an orange
It's an MCM, there are two 8-core dies connected via HyperTransport.

>>60735146

Pretty much this.
I'm still on an i3 3320/gtx660 and I see no reason to upgrade. Most of the things I run are single threaded or use 2/4 threads at best. I don't play games, don't do editing, my gpu can already render complicated scenes in a few minutes. Even with my cpu/gpu set to conserve energy they still perform fine.
Even after they die down I'll probably buy something similar in performance.

>>60734321
>Die the size of an orange

*Package* is the size of an orange.
There are 2 dies and they're both about fingernail-sized.

>>60734321
>I hope you guys have good home insurance.

americans and their """houses""" made of plywood

>>60734321
>die
dies

>>60735948
an electron travels at over 600 mph, so no

>>60736823
>There are 2 dies and they're both about fingernail-sized.
pic of delidded thread-ripper?

File: C2Q.jpg (86KB, 998x759px) Image search: [Google]

86KB, 998x759px

>>60737248
It'll look something like this but with smaller dies

File: AMD-EPYC-Four-Die-with-Infinity-Fabric.jpg (36KB, 957x539px) Image search: [Google]

36KB, 957x539px

>>60737248
we don't have threadripper but have epyc. think half of it.

File: Epyc%20Processor_1495111710.jpg (62KB, 500x353px) Image search: [Google]

62KB, 500x353px

>>60737397

>>60736504
Congratulations on not being the target audience, I guess? Are your low requirements a point of pride? You could probably scrape by with a Core 2 Duo or Phenom.

>>60737397
Does that mean AMD can go even one step further and create an Epyc Threadripper since the dies are so small?

>>60737037
And? Electron's speed doesn't increase with distance so the bigger the distance the longer the time, he's concern is valid but misplaced - internal cpu latency is dealt with by hw engineers, not end user. What they write on the box is what you get.

>>60737473
they use the same socket. so, yes.

>>60737473
Yes, the sockets are compatible, as >>60737511 said, but keep in mind that the threadripper motherboards have fewer RAM channels and PCIe lanes compared to the server line (EPYC), so there may be incompatibilies.

File: 1485886098705.png (52KB, 318x308px) Image search: [Google]

52KB, 318x308px

>Intel getting blown the fuck out by AMD
>can't compete on cost to performance

Intel are getting AMDicked

>>60735948

>>Wouldn't this cause massive latency?
>electrons moving at 2,200 KM/s
>"MASSIVE latency"

What a fucking retard you are. Just neck yourself before you make more brainlets to pollute the earth with.

>>60734321
>I hope you guys have good home insurance..

4x as much as how much the place is worth.
I'm ready for make -j33

>>60734321
Enjoy being a dielet

File: 14884s859s94763.png (366KB, 1144x888px) Image search: [Google]

366KB, 1144x888px

>>60734321
AYYYYYYYYMD strikes again

>>60737838
>>60737037

Actually the speed of electrons is measured in cm/s inside conductors, when current is applied.

>>60738007
Do you actually think signals are transfered via electron drift?

>>60734406
What other coolers are avalavle besides noctua (dont like brown or the noctua jew tactics)

>>60737248
You mean two die.

>>60738082
>noctua jew tactics

elaborate pls

>>60738648
Probably referring to the fact that if you want black Noctuas you need to get the more expensive industrial PPC fans.

>>60738724
>le industrialPPC

You dont even know what PPC stands for you fucking homosexual shit eater

Is there a lower priced single user PC version or is this all for workstations and low end servers.

>>60734321
if that thing is priced at 1000$ or lower, there will be ZERO incentive for anyone to buy intel anymore. fucking i7 6900k is 1000$ TO THIS DATE

>>60738999
Threadripper is basically 4 Ryzen core complexes (CCXs) stuck together. If you want desktop Ryzens they've already been released for months.

>>60736504
Thank you for being a normal guy, not a consumerfag like kids in /g/

>>60738724
They sell grey colored fans for the exact same price as the poo colored ones. I have a bunch of them in my PC right now.

>>60739275
Rumors are that it will cost $850 for the non-X 16 core.

File: AMD_CEO_Lisa_Su_20130415_cropped.jpg (169KB, 662x781px) Image search: [Google]

169KB, 662x781px

>>60739569
JESUS FUCKING CHRIST HOW DO THEY KEEP DOING IT?

>>60739658
80% yields for fully functional 8 cores helps a little

>>60734321
It needs to be WIIIIIIIIDER

>>60739658
It costs $110-$120 for a fully working Threadripper CPU. And they have 80% yields.

File: wideripper.png (349KB, 800x627px) Image search: [Google]

349KB, 800x627px

>>60739707
Yuno like wideripper!

File: wideripper.png (370KB, 800x627px) Image search: [Google]

370KB, 800x627px

>>60739707

File: 1469880875535.png (185KB, 1900x1400px) Image search: [Google]

185KB, 1900x1400px

>>60739719
>>60739724

>>60734343
someone update the breddy gud meme about pentium 4

File: 1492122290568.png (992KB, 801x1500px) Image search: [Google]

992KB, 801x1500px

>>60739773

>>60737412
Good question, how will the die disposition on TR?
2 upper or 2 left/right? From a thermal radiance perspective, the better would be to lay down both dies diagonally

File: dies.png (6KB, 960x540px) Image search: [Google]

6KB, 960x540px

>>60739824
take your pick

>>60734406
>Have to buy completely new CPU coolers (applies to air coolers, aio coolers and custom loop water blocks) just because of the massive fucking IHS

Jesus christ AMD. It's not even something simple like needing a new bracket. You actually need to buy an entirely new cooler because of the size.

>>60739838
My guess is the left for better thermals.

>>60739868
if we look at the socket composition, right seems more likely. even though it would be worse for thermals.

That`s a big CPU

>>60739906
If it was the right, I'd expect it to be taller, more inline with its width for the resistors. Then again on ryzen they're very close to the IHS.

>>60739980
Four rack-units

>having a 32 core CPU at home
Isn't this type of thing for shit like vm servers?

>>60734321
>4 ghz
Really?

File: 1494052174535.png (2MB, 973x6115px) Image search: [Google]

2MB, 973x6115px

>>60740010

Even 16 threads is beastly for anything you are likely to do at home. 32 threads is well into serious business territory.

>>60740010
Threaripper is ""only"" 16 cores, the package looks the same as their server CPUs but the internals are different

>>60740018
two separate dies with higher steppings than the r7s, no reason not to, given proper cooling

>>60739658
I guess Jim Keller really is the GOAT.

>>60739842
Considering the cost of the CPU, motherboard, and RAM, a new cooler is basically nothing.
Even better if you have watercooling since you only need a new water block.

File: 1478907044435.png (394KB, 493x639px) Image search: [Google]

394KB, 493x639px

>>60734321
Oi goy

Why can't they just make a 1m by 1m CPU that has like 1000 cores and 16GB of built in cache?

Sure you'd need like a case the size of a fridge and a cooling system that sends the hot air directly outside the house to stop it melting but damn would it be good for compressing memes.

>>60740133
Because that's retarded and clusters/supercomputers exist.

>>60740133
Because the only area in which this kind of computing power is required is scientific High Performance Computing where other and more easily implemented architectures are present

>>60734321
Is this actually how big it is? I still have no idea because I've seen /g/ Photoshop it to any size from a credit card to a license plate.

>>60740133
It takes around a month to spin forge a 40cm silicon ingot. I'm sure you can extrapolate from there

File: TB2f4vIvrJmpuFjSZFwXXaE4VXa_!!2341477237.jpg (335KB, 1060x1047px) Image search: [Google]

335KB, 1060x1047px

This is the reality of Intel's 18 core.

>>60739906
No, it's the one to the left, because the GMI links that connect the dies are on the wider side of the die.

>>60740202
those are the most pathetic cinebench scores I have ever seen

even if threadripper has a shitty scaling in relation with a single 1800X it still performs better than anything from intel at similar core counts

File: ryzen_cinebench_multi_threaded-100709909-orig.jpg (87KB, 1160x697px) Image search: [Google]

87KB, 1160x697px

>>60740202
And the Coup de grâce

>>60740202
>18c/36T
>below 3000
Intel is fucked.

>>60740202
>>60740244
>slower than two 6900k
>2 more cores
>newer arch with higher memory support

You can thank the Intel Patented RingBus(TM) for this

>>60740244
>>60740202
That explains why it's delayed to next year.

>>60740257
yeah, intel's scaling is even more shit at quad sockets, where you barely get 2.5 worth of performance

>>60740274
*2.5 processors

>>60740202
HCC dies for consumer were a mistake.

The more you increase the bandwidth per node on the ringbus, the slower it gets, it works somewhat okay when cores aren't high clocked, but not when you need high clocks at HEDT

I hope their XCC dies are not 3 buses, that would be fucking terrible

>>60740202
>165W with those shitty scores
kek, do these things at least have a good amount of pci lanes or it's just 44?

>>60740305
44 only.

>>60739842
>buys 155w TDP $1,000< CPU
>keeps old cooler
?????????

>>60740305
>>165W

Intel's TDP is rated at their stock clocks, not turbo clocks.

>>60740175

It is four CCX's bolted together so that is how big the package is with the IHS. It really is just double everything of a R7.

>>60740310
>>60740332
what is the advantage of paying twice the price for barely similar performance and a lot less features? has intel gone mad and are trying to destroy themselves on purpose?

>>60740348
You could fit 8 CCX under the same heatspreader, AMD Is just keeping compatibility with the inevitable EPYC consumer release (either 3 die[doubtful], or 4 die with 24 cores.

>>60740360
>has intel gone mad
they got caught their pants down. they didn't expect AMD could deliver with ZEN even though Jim Keller was back and they know what he did back then.
so they didn't do anything these past 5 years.
and now, because of the pressure from shareholders, they can not cut prices, thus they are chimping out.

>>60740360

Chances are they simply didn't expect AMD to get so close with ryzen and now their plans that have been set in stone for a while are hastily being changed where they can to have a competing product. It is why the X299 cpus feature wildly different support for I/O as Intel is trying to get something out within (in the context of the scale these things work on) 30 seconds.

Zen scaling like it does has blindsided Intel.

>>60740383

Assuming that is true and there are no issues with tracing that would be one hell of a dense chip which would make cooling impractical in most deployments. Now if 7nm pans out like AMD hope your idea might be a reality (remember a good 50% of AM4 pins are unused).

>>60740408
7nm will probably bump CCX to 6 cores and keep the rest of the structure similar to what we have today

>>60740408
>hell of a dense chip which would make cooling impractical in most deployments.
The heatspreader is the same, its still 4 dies spaced out at lower clocks.
Cooling them would be piss easy.

File: reality.png (1MB, 1920x1080px) Image search: [Google]

1MB, 1920x1080px

>>60740202
>>60740244

File: [HorribleSubs] Long Riders! - 01 [720p].mkv_snapshot_18.48_[2016.10.29_02.06.23].jpg (48KB, 436x402px) Image search: [Google]

48KB, 436x402px

>>60740487
>Kara no Shoujo
Is that a big spoiler? Hopefully I would have forgotten this by the time I play it.

>>60740202
Kikeripper indeed.

>>60740536
Very.

>>60740536
Study shows spoilers actually increase the enjoyability of a work

>>60737838
When shit is operating at 3.9GHz any latency can be massive.

If it's no problem, then why aren't all CPUs huge? Why would companies put all this effort into making them smaller?

>>60734321
>>Die the size of an orange
That's not the die you retard. That is the heatspreader.

The juuuge socket is so hjuuuuudg because it has 64 pci-e lanes and thus requires 4096.

AMD is winning
bigly

>>60740952
that's not at all the reason for all those pins

even on AM4 50% of the pins are unused, this is clearly so amd can do serious upgrades and keep the same socket

>>60741048
We already know this socket supports 32c/64t.

>>60741063
yes, but implying that *all* the pins are needed as of now, is wrong

>>60741048
>even on AM4 50% of the pins are unused
What? No. Grounds are not unused.

>>60741063
The single largest thing accounting for IO needs is the memory. Supporting 8 channels of DDR4 per socket requires an awful lot of pins.

>>60741106
I/O and memory channels use the majority of pins

266 pins per DDR4 channel = 2128 for 8 channel memory
3 pins per I/O lane = 384 for 128 lanes
That's where the large majority of the pin budget goes, pin per core is unknown even to this day, but it's miniscule compared to the these two.

tl;dr, you've got a shitload of pins left on both AM4 and SP3

>>60734343
kek, u fell for the AMD TDP meme
Intel published the worst case scenario TDP for all of their CPUs, which means that they will rarely ever cross their rated TDP in normal circumstances.
AMD publishes their AVERAGE TDP rating, so that means that the TDP can climb much higher than what they claim on their marketing materials.

If Intel says that their CPU's TDP is 160W, that means that it's worst-case scenario has their CPUs emitting that much heat at their peak.
When AMD says that their CPU is 155W, that means that their CPUs can go anywhere from idle TDP (40W) to their peak (200+W), but average out somewhere in the middle for most of the time.
Intel still has the lower TDP, you morons.

AMDrones are so stupid that it's cute

File: aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9FL0MvNjY5MjUyL29yaWdpbmFsL2ltYWdlMDAxLnBuZw==.png (74KB, 1112x833px) Image search: [Google]

74KB, 1112x833px

>>60741174
You can go to bed now, Brian.

>>60741185
>not posting the 6950X
why?

>>60741200
Because they didn't test it, I guess?

>>60741185
Everything below that 1600X exceeds their TDP, hilariously enough only not the 9590.

>>60741185
>posting a benchmark that Ryzen specifically throttles itself for
Try posting the other power consumption charts from that same review, you fucking Poojeet shill

>>60741235

Fun fact: because of how thermodynamics work every cpu in existence that has some sort of turbo feature is allowed to exceed its TDP for short periods.

File: dN3UmtS[1].png (107KB, 1920x984px) Image search: [Google]

107KB, 1920x984px

>>60741174
>AVERAGE
>V
>E
>R
>A
>G
>E

KIKETEL FAGS ARE LITERALLY MAKING THIS SHIT UP NOW, THIS IS FUCKING HILARIOUS


>J-J-JUST WAIT! INTEL WILL WIN NEXT YEAR! I-I-SWEAR!

File: The lake, the lake, the lake is on fire.png (1MB, 1110x5110px) Image search: [Google]

1MB, 1110x5110px

>>60741185

>>60741283
>Inb4 Kiketel fags try calling TomsHardWare on shilling for AMD

I think this would be the point of no return.

File: Screen Shot 2017-06-04 at 18.04.38-fullpage.png (1MB, 1359x4713px) Image search: [Google]

1MB, 1359x4713px

>>60741253
Which one? Either way Intel embarrasses themself.

File: same.gif (625KB, 200x178px) Image search: [Google]

625KB, 200x178px

>>60741268
>>60741283
>>60741310

>>60741310
>>60741283
>>60741185
>these are tested from the rails
>Ryzen has a integrated southbridge, Intel does not
>extra few watts off the board and to the processor

Kek Intel and it still get demolished

>>60741340
AMD hasn't even enabled the FIVR on Ryzen yet.

>>60741333
Calling samefag on giving evidence, like it makes the truth invalid?
You're all sorts of retarded.

File: prime95lol.png (38KB, 1116x835px) Image search: [Google]

38KB, 1116x835px

>>60741185
This one is even better.

>>60741504
>downclocked corelet

Kek

>>60741504
>8 cores consumer as much power as DOWNCLOCKED 4 core
Mother of fucking God somebody please remove Intel from this planet.

>>60734321
LOW YIELDS!!! LOW YIELDS!!! LOW YIELDS!!!

File: 1276182361461.jpg (149KB, 1000x1400px) Image search: [Google]

149KB, 1000x1400px

>>60741504

>>60741584
this is not an intel architecture tho

>https://www.guru3d.com/news-story/amd-ryzen-14nm-wafer-yields-pass-80-threadripper-cpus-on-track.html

File: AMD-EPYC-CPU-740x416.jpg (26KB, 740x416px) Image search: [Google]

26KB, 740x416px

>>60741584
Don't you mean really fucking high yields? It's not one giant die under there, dude. It's two seperate dies with 80% yields on each. Pic related is Epyc, same thing but only two dies.

>>60741623
>AMD ten years ago: "We made the first REAL quad core on a single die. Only losers put two dies in a single substrate"
>AMD today: "There is nothing wrong with putting more dies in a single substrate"
No wonder that no one takes AMD seriously anymore.

>>60741682
They clearly realized their mistake. Intel, on the other hand, seems to have gone to AMD's previous position. lol

>>60741682
>implying anyone ever took intel seriously

>>60739790
You forgot to edit the bentium 4 logo at the bottom right.

>>60737327
Why intel isnt doing this again? Q6600 was amazing and so was almost every other c2q.

You're a big Die.

>>60739838
Its still not correct because you are sizing the dies the size of 2 dies, when there are space for 4 dies under IHS

So there'll be cpus in this socket range that aren't 16 cores (10-12 I heard) but can get up to 4ghz like 1700-1800 and have all those pci lanes still. At least that's what I read.

In that case, wouldn't it be better to get those over a ryzen, depending on the price?

>>60741682
You mean.
>AMD ten years ago: "We made the first REAL quad core on a single die. Only losers put two dies cut off from defective units in a single substrate"
>AMD today: "Instead of using unsold cut downs and slapping them together into small core count components and selling it at a premium, we've decided to put multiple full length 8core dies on a single substrate, then discount the current line so that we can deliver 16-32 core chips in a near perfect cost to performance increment."

>>60740328
Some AIO or water blocks under loops are way beyond 150W.
Since the ihs is this huge, i suspect even older AIOs with adequate mounting can do the job as the IHS will spread the heat

>>60741788
what? the dies are exactly the same for all ryzen models regarding size, and threadripper's package size is exactly the same as Epyc's, and both have 4094 pins

>>60740881
remember how good C2Q line was? That was 2 c2d glued together. Worked fine? Yup.

>>60741819
and the cores couldn't talk to each other inter die, and used north bridge.

>>60741798
I doubt most people would be able to use all 8 cores of a Ryzen 7, let alone all of Threadripper's. Its motherboards may also be more expensive and you'll certainly need much better cooling.

>>60741682
If massive dies allowed a high % of dies, no one would make separated dies. The only reason to have multiple dies, that have it own problems like interconecction under substrate, is that you cant really have such massive die without a petty aweful rate.

The way amd positioned their production, they have almost 0% waste, since even half failed CCX is binned to down products.

>>60741817
the pic shows 2 dies on the package, its wrong because it shows both with almost size of the ihs.
This was my original question, how tr dies will be mounted, since there are "2 lines 2 columns" under the IHS. The mounting is directly related to how its better to use a heatsink

>>60741835
really shows how out of touch intel became, literally jewing there customers for maximum shekels. 5y old cpu is literally slower 15-20% than newest 7700k. Literally only reason its even those 15% faster is because its smaller nm and is able to clock slighty higher.

Home users have second market for intel they price there products at incredible prices for home users, so they can price at ludicrous prices for server/professional market.

>>60741897
It's 2 different packages, each with a different die configuration inside, what are you on about?

>>60741850
all the cores are already getting used in some newest games. They are utilized at like 40%, but you can expect higher utilization once we get new get gfx cards.

>>60741917
money makes you more greedy and ignorant. they literally thought they are untouchable.

>>60740881
>Why would companies put all this effort into making them smaller?
The same reason phones became nearly invisible and now we've got 7" phablets.
Miniaturization leads to bigger products down the line again because now we can just cram a TON of shit into the same space

>>60741850
If you don't know if you need all the cores/lanes a HEDT platform provides, you don't need HEDT. Period.

>>60741850
True.

I'm thinking ryzen for streaming vidya and stuff. Not sure if all the extra pci lanes of threadripper would help with that.

Might mess around with 3d modeling too. Just for fun.

>>60738007
no shit it was a joke.
I don't thinkthe bottleneck is the fking speed of light over 6 cm vs 3 cm

>>60734384
your argument is shit and not based by actual chance, idiot

I do believe you eat uncooked food at your house then?

>>60734321
I've been wondering where they're putting all that cache. How large is the L1 cache anyway?

>>60742288
32kb L1d, 64kb L1i afaik.

>>60739842
If you're buying a processor that's over a thousand dollars you might as well invest in a good cooler to go with it.

>>60744519
The top end TR might end up costing less than a thousand.

>>60744519
His point is that you can't use good coolers you already have.

>>60744546
There's probably a bracket. But if you plan on OCing that shit, you're going to want a full sized contact plate.

>>60744566

There being a bracket isn't the issue - the issue is the socket is massive and the baseplate of your D15 physically doesn't cover the IHS.

>>60740202
>>60740244
Holy fucking shit, that's just embarrassing. Intel got way too complacent.

>>60741682
Intel 10 years ago: By using MCMs for core2quad we can improve yeilds and reduce costs enabling us to deliver what would be server only CPUs to the consumer market
>Intel now: ADD MORE RINGBUDSSS :-DDDDDD WHAT YOU MEAN 1.7% IS A LOW YEILD WATCH THIS JHH YOU AINT SEEN SHIT

>>60741504
My fucking sides.

>>60744879
>Intel 10 years ago: By using MCMs for core2quad we can improve yeilds and reduce costs enabling us to deliver what would be server only CPUs to the consumer market
thats amd now

>>60740881
Yields are inversely proportional to the square of the diameter

>>60734406
I know TR is a MCM but does that literally mean they're taking Ryzen candidates and gluing them together?

>>60735146
>muh vidya
these clearly aren't gaymer chips

R5 and R7 are plenty enough for that

>>60745206

In practical terms every zen based chip in existence (including those that are announced but not released) are MCM - there are no single CCX ryzen chips yet. Chances are raven ridge will change that but that is pure speculation.

>>60745243
>(including those that are announced but not released) are MCM - there are no single CCX ryzen chips yet
AMD announced the mobile chips at Computex, single CCX by the looks. Hell, it's right there in OP's image.

>>60741174
that's why you need a watercooling solution "minimum" for the Skylake-X /Kabylake-X processors, because 160W are absolutely not doable on air, right?

>>60739842
Eh, it's just a new waterblock for my custom loop. More than the purchase itself I'm bothered by the PITA of having to open up my loop, drain it and then bleed it again. Last time I had to do that it took an entire afternoon with proper bleeding.

>>60745225
These are the only CPUs AMD makes which are adequate for high-end multi-GPU systems intended for 4K gaming. Mainstream Ryzen doesn't have enough PCIe lanes and those can matter specifically at very high resolution. It's completely niche, sure, but I bet there will be some interest in the "low end" 10C Threadrippers for 4K gaming, especially since the cheapest Skylake-X with sufficient lanes is a whopping $1000.

>>60745265
we can expect APUs with a CCX and a small Vega GPU in the future.

>>60745462
>Ryzen doesn't have enough PCIe lanes and those can matter specifically at very high resolution
Nigga what in the actual fuck are you spouting?

Atleast they are soldered :^)

>>60735957
for you

>>60745834
It doesn't have 32 PCIe lanes available for 2 graphics cards, how is it hard to understand?

>>60746038
You're not fully saturating PCIe 3.0 x16 with a GPU anyway.

>>60746059
You haven't ever used SLI or CF at native 4K on x8/x8, right? Just to be clear.

>>60746070
Why would I? I don't fall for shitty memes.

>>60746126
Thanks for your honesty, that's all I needed to know to completely discard your opinion, as you clearly have no idea about the subject you're trying to discuss.

>>60746070
8 lanes each is more than enough to run any title at 2160p, even with 10-bit color output at 4:2:2
That's just about 7.6GB/s of bandwidth to each card, but just the raw, uncompressed bandwidth of the output itself is a mere 1.5-4GB/s at 2160p 60-100FPS with 8-bit color. Even if you add the overhead, 7.6GB/s is still plenty for 4K output.

If anything, the bottleneck would be the PCIe bus inside the CPU, not the bandwidth going to that bus.

>>60746145
Last I checked the only things that came close to saturating PCIe 3.0 x16 were capture cards, sorry I don't keep up to date on all the latest memes.

Anyone have a clue as to prices? What's the 16c/32t could going to go for? X399 motherboard prices?

I have a 1700X/X370 setup as is. I honestly don't NEED more cores. It's just I want dem pcie lanes. I wanna run 3 nvme SSDs and 2 gpus all at Max bandwidth.

>>60746303
Rumors have Threadripper starting at $850.
Motherboards, who knows.
I'd guess $400 for a decent one.

>>60736504
Hey that's pretty cool, what do you do with all your free time? Pls don't diddle kids

>>60746315
>starting at $850
>starting
this is the rumored price of the non-X 16 core, so there are at the very least 2 SKUs below that price range

>>60740202
Under 3k and I'm pretty sure I remember seeing a cinebench score from the 16 core Threadripper that was 3050. If both of those scores are actually, then Intel is fucking finished, especially if the prices are correct.

>>60740360
They have the mentality of vista era Microsoft. They think hey are invincible and not subject to the laws of the market.

>>60734321
>bigger plate
>house fire

pick 1

File: ringbusbusbusbusbusbusbusbusbusbusbusbusbusbusbusbusbus.png (762KB, 10000x706px) Image search: [Google]

762KB, 10000x706px

>>60744879
>Intel now: ADD MORE RINGBUDSSS :-DDDDDD
You got it, chief.

File: shit wrecker with certification.png (400KB, 816x1020px) Image search: [Google]

400KB, 816x1020px

>>60734321

THANK YOU BASED SHITWREKER

BIGGEST DICK BIGGEST HEATSINKS

>>60741504
intel runs prime95 way more than twice as fast

File: 1b70c8cd-73d1-4cab-bc44-903e46d0351a.png (105KB, 255x345px) Image search: [Google]

105KB, 255x345px

>>60740360
There is no advantage. There is no redemption. What happened was Jim "The Shitwrecker" Keller.

>>60748842
Source: My ass.

>>60748855
keller didn't even touch the ryzen design

File: shitwrecker.jpg (430KB, 1209x1050px) Image search: [Google]

430KB, 1209x1050px

>>60748866
He was project manager. He commanded the plebians to create an architecture worthy of Keller. And they delivered it unto him. And the Keller look upon it, and it was good.

File: lel.png (24KB, 893x210px) Image search: [Google]

24KB, 893x210px

>>60748861

>>60748861
if ryzen is 5x slower, why does it still use 1/2 the power lol

File: 1496368165831.jpg (53KB, 627x619px) Image search: [Google]

53KB, 627x619px

>>60748522
M- Muh latency

>>60741174
You keep spouting this blatant lie despite being corrected on it repeatedly.

Intel TDP ratings are for base blocks not "absolute worse case". AMD goes by "average usage" which is much closer to reality.
Intel's is like the old EPA gas mileage ratings, while AMDs is like the new ones.

>>60748855
>>60748920
>Jim Keller
>Jem Killer
>Jew Killer

File: aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9JL1IvNjU2NDUxL29yaWdpbmFsLzAxLVdhdHRhZ2UtQmFycy5wbmc=.png (28KB, 711x533px) Image search: [Google]

28KB, 711x533px

>>60748946
If 1800X is so much slower, why is it's TDP use right in line with where you'd expect it to be? Just sitting there wasting power for no reason? Suddenly became 10 times as power inefficient? 8 cores < 4 cores in prime95 for no reason?

>>60749180
>Just sitting there wasting power for no reason?

must be

>>60749231
Yes, thank you for your expert analysis Mr. (looks at namecard) hughfeghot.

>>60749252

just like an amd cuck to ignore any evidence that his favorite company is a failure

>>60749293
You noticed a discrepancy, but have no explanation for it other than "AMD must suck horsedick huhuhuh". Would like to see some logical explanation for this shit.

>>60748946
Also, in this image you posted, it shows the i7-6850K losing to the i7-6700K, despite the 6850K having more cores. So there's clearly something more going on here than just "AMD sux huhuhuhuh".

>>60749432

yeah, it's a serial workload genius

>>60749480
Uh, you know Tomshardware is using the stress test with small FFTs, not the actual prime finder thing? It uses 100% of all cores.

File: LRM_EXPORT_20170521_103222.jpg (2MB, 1110x5110px) Image search: [Google]

2MB, 1110x5110px

>>60741253
Oh look, another knowing and deliberate liar making anti-AMD shitposts. I'm shocked. Really.

File: 8283E83E-551C-4B02-8ED2-FA08634EB67D-1995-000003133F8351FC.png (2MB, 1920x1200px) Image search: [Google]

2MB, 1920x1200px

>>60734321
>3ghz
>18 cores
>36 threads
>only 2 boost to 4.0ghz
>created by taping two xeons together and then making a toothpaste sandwich with it
I hope you have a good insurance bundle...

>>60734321
The dies are taking a fraction of total space faggit.
Oh and it is soldered.
Oh and dies are running cooler if they are distanced from each other.

>>60750669
>on a huge PCB
I like it, but it's touching my autism.

>>60741333
the posts are replying to different posts, that's not how samefagging works newfag

File: 1327280893169.jpg (26KB, 384x288px) Image search: [Google]

26KB, 384x288px

>>60750599
>Tomshardware

>>60749550
still 1/4 the speed or less

File: AYY.jpg (273KB, 1032x774px) Image search: [Google]

273KB, 1032x774px

>>60750772
(You)

File: 1490702465219.png (68KB, 500x500px) Image search: [Google]

68KB, 500x500px

>>60750772
The place renown for shilling intel is also shilling AMD?
See
http://www.tomshardware.com/reviews/best-cpus,3986.html

File: [Asenshi] Little Witch Academia - 10 [2761C95B].mkv_snapshot_13.30_[2017.03.13_08.34.12].png (275KB, 484x588px) Image search: [Google]

275KB, 484x588px

>>60750781
>Intelshills are this desperate to disprove their CPUs are housefires

>>60750870
>amd fags satisfied with sub par serial performance

>>60741174
>it's
What's your first language?

>>60750894
I don't know what you're talking about, I'm eating a big bowl of cereal right now. Yum.

>>60740202
And this is going to be what, $3000.00?

Oh my.

>>60734321

credit card sized chips when?

>>60751040
now

>>60734406
>their yields on the 8 cores are 80%
Jeez

>>60751010
$2000.

>>60751107
Ok that's much better. Are people claiming a $900 Threadripper is going to out perform this? Where is AMD's competitive chip coming in at? Even $2k seems a bit optimistic for that performance.

>>60751200
Take an 1800X cinebench score and multiply it by 2. That's about where it's going to land.

File: amd-zen-core-scaling.jpg (33KB, 597x430px) Image search: [Google]

33KB, 597x430px

>>60751200
Lest you think he >>60751215 is full of shit, here is AMD's own benchmark showing ~1.99 scaling from 8c/16t to 16c/32t

Literally 1.99*Ryzen 1800x score is pretty close to what you should expect.

File: intel-core-i9-x-series-2-1200x633.jpg (84KB, 960x506px) Image search: [Google]

84KB, 960x506px

has the threadripper price been confirmed? there was a lot of wishful thinking about the ryzens with 1800x etc.

>>60750912
>being this retarded

>>60751291
No confirmation but it's expected to be $1000 or less for the top end 16 core.

>>60751215
>>60751252
You can't be serious. That means a Cinebench score of like 3300 or something close to that. I do quite a bit of video processing and it's not uncommon with my 4770K to have to set it up and find something to do for up to an hour. Hell I built another PC to deal with the time I spend waiting on video to finish. I'm just a nobody and it makes financial sense to find room in my budget for a $900 CPU if these figures are realistic.

How is Intel going to sell $2000 cards to mid/large businesses if they are being beat in workloads by 12% by a CPU that costs less than half their new offering? I'm going to have to wait and see how this shakes out. Forgive me if I'm sceptical. It just doesn't add up logically that Intel gets caught sleeping this badly.

File: 1486510190271.gif (155KB, 640x360px) Image search: [Google]

155KB, 640x360px

>>60751409
>caring about serial ports that haven't been on PCs since the mid 2000's
lol.

>>60751438
>anime posting and being a retard

might as well end it now

>>60734321
>16 FUCKING CORES
intel cucks BTFO

>>60751424
Zen scales near linearly with core count. Intel's old, crappy ringbus architecture can't scale anywhere near as well.

You can see Threadripper demonstrated at Computex. https://youtu.be/NeuamYdfU4w?t=16m

You can download the blender test they used here. http://www.pcworld.com/article/3151464/hardware/you-can-find-out-how-your-cpu-compares-to-amds-ryzen-for-free.html

>>60751424
*Sell $2000 chips.

>>60751424
>It just doesn't add up logically that Intel gets caught sleeping this badly.
It happened. Intel sat on their ass for 5 years when they should have been making a new architecture. Now they have nothing and their new architecture isn't out until 2020 at the earliest.

>>60751424
I'd say ~3,100-3,200 is probably more realistic. The Zen architecture scales to 14+ cores massively better than Intel's three ringbus circus, and it's way cheaper to make.

>16c
>32t
>respectable TDP
>impressive clocks for a high core count part
>sane price
>actual platform for a prosumer/workstation system
>perfect incremental upgrade between consumer parts and enterprise parts

AMD is pretty good these days

File: 1493952761093.jpg (336KB, 784x847px) Image search: [Google]

336KB, 784x847px

>>60751291

16/32 tentative price is $849

X399 boards cost as well / glorious TR4 design

FINALLY quality board offerings, dual nics get

File: jewshown.png (138KB, 354x504px) Image search: [Google]

138KB, 354x504px

>>60751504
>16/32 tentative price is $849
THAT'S TOO CHEAP GOY

BUY MY OVERPRICED 12 CORE INSTEAD!

File: cdf[1].jpg (56KB, 648x841px) Image search: [Google]

56KB, 648x841px

>>60751453
>Gets assblasted when someone posts an anime picture on a japanese inspired futaba (imageboard).

Someday you'll grow up. But for today, I love blasting that tiny little ass of yours, leaving you a nice goopy animay creampie.

File: OwO what's this.jpg (17KB, 540x255px) Image search: [Google]

17KB, 540x255px

>>60751520

>>60751504
>mobo needs 2x8 power connectors
wud da fugg

>>60751504
where the fuck are the m.2 ports?

>>60751581
Four full length 8 core dies. What the fuck did you expect?
Think of the V droop if it was pulling from only two.

>>60751593
>extreme gaymer motherboard
>casual normie ports
pls

>>60751605
from one*

>>60751465
Pretty blown away. The AMD team who pulled this off deserves a raise.

>>60751485
Guess that makes sense since I've had a hard time finding much benefit in upgrading my 4770 these past several years. One year the new Intel analog to my CPU performed the same but just used 10w less power. That was a headscratcher. I thought "if this is all they came up with after two years what are they paying their engineers to do?"

>>60751487
Even a bit lower score wise how can Intel put a $2000 price tag on their new chip and send their sales guys in to presentations at companies? You have to buy a new socket so that means new hardware to lease. What happens when AMD comes in after them and says "clean out half your racks because you won't need them and also you're getting a performance boost."

>>60751605
b-but muh mini atx builds, youd need a psu the size of a truck

>>60751629
>Even a bit lower score wise how can Intel put a $2000 price tag on their new chip and send their sales guys in to presentations at companies?
"We have more cores than the other guy".

>>60751424
>Forgive me if I'm sceptical. It just doesn't add up logically that Intel gets caught sleeping this badly.

Their entire business strategy has been laser-locked on riding out post-Sandy success and Bulldozer's failure. Everyone should have figured out by now that Intel can't be bothered making decent products unless they know they have to. They expected Zen to suck like Bulldozer and have been completely blindsided.

do you think there will be a 2 x 1700 threadripper? so still 16c/32t but ~130tdp?

>>60751682
This is what the TR lineup looks like. https://www.techpowerup.com/233945/amd-readies-nine-ryzen-threadripper-models

>>60751652
you mean mini itx? i kinda doubt they'll make mini itx boards for x399

>>60751715
There's nothing mini about Threadripper. That socket isn't going anywhere except on a full sized board.

>>60751652
>mini atx
>for x399
why the hell you want that ?

File: 1491891327069.jpg (178KB, 904x542px) Image search: [Google]

178KB, 904x542px

>>60751593
>>60751613

3!! m2 slots are on board, not sure where and icing is a 10GB lan card included

really like this design, Asus got their game back, previous gundam didn't cut it but it's refined, awesome nevertheless

>buying gaymen boards

want to see signature deluxe or equivalent

>>60751736
>>60751715
>>60751742
you guys just dont understand the dankness of toasters

File: ripandtear.png (42KB, 215x191px) Image search: [Google]

42KB, 215x191px

>>60751753
I wouldn't want to put something called Threadripper into a toaster. It just doesn't feel right.

>>60738595
no

File: ASRock X299E-ITX.jpg (2MB, 1510x1132px) Image search: [Google]

2MB, 1510x1132px

>>60751753

>>60751736
MicroATX is the smallest I see it going to, and even then that would result in the waste of alot of PCI-E lanes.

>>60751801
So basically, you aren't going to see it.

File: pic_03.jpg (3MB, 3300x2550px) Image search: [Google]

3MB, 3300x2550px

>>60751752

I am done with the gamer aesthetic.

Give me function, serviceability, and ease of use. Pic related.

>>60751848

much agreed and that's one of the nicest cases I'ver ever seen

just noticed 802.11ad

>>60751593
See the extra DIMM slot on the right side?
That's not for ram, it's for a little riser card thing that has m.2 slots on it.

>>60740310
UP TO 44.

>>60751915
Thats pretty neat

>>60751848
wait. what that black thing and its function ?

>>60751753
Ryzen 1800x on a mini-ITX is enough cores and performance to sustain a single-GPU Mini-ITX build.

I don't think Threadripper adds anything significant that would merit a Mini-ITX board, especially considering the socket size.

mobo makers plz don't go into out of stock tactics just like X370

price point sure, just don't understimate demand

>>60740360
>>60740394
AMD is pretty stupid if the price rumors are true though. HEDT is niche, the supply/demand doesn't add up, not enough people are buying them to justify slashing the price, and the people who do buy them (professionals/prosumers) aren't as sensitive to the price as ordinary consumers.

>>60751959
>I don't think Threadripper adds anything significant that would merit a Mini-ITX board
An additional 16 threads.

>>60751959
yeah it would be more like a fetish thing, a niche within a niche, to get the most performance in the smallest build disregarding the practicality of it

>>60751753
Not those anons, but having a very small cute case sounds very appealing compared to the mid tower behemoth I have on the floor. My problem is that I horde too much storage and haven't moved it all to a dedicated NAS.

>>60751989
AMD needs mindshare almost as much (if not more) than they need money right now and the HEDT line as a whole can be considered a halo product for all other consumer lines. Having a competitive product for a low cost there will help engender a sense that their other products also possess the same value.

Also it makes Intel look bad, which is a less noble but still effective strategy.

File: sunshine.png (771KB, 1136x640px) Image search: [Google]

771KB, 1136x640px

>>60739724

>>60751989
Power usage per performance is incredibly important in that market. (Because heat generation needs aircon which is expensive)
Of which AMD is winning.

It's so much winning companies are going to wonder if there's something wrong:
"So it performs better, uses less power.. for half the price? There's gotta be a catch"

>>60752090
>which is a less noble
How so? They're not directly sledging Intel, they've genuinely released a decent product that's caught out Intel with their pants down.

>>60751848
Same goes with motherboards. Im sick of all the fucking designer heatsinks on motherboards. I just want fucking square heatsinks god dammit. And no fucking graphics over my motherboard either. I was just briefly shopping for AM4 boards and every single one of them was ricer garbage. Are there any AM4 server motherboards so I dont have to deal with that dumb shit?

>>60752042
do what i did, i had a full tower full of hdds, when i last upgraded to a new socket i needed a new cpu, mobo, ram so i just grabbed a new case and psu while i was at it and left all the hdds in the old case.. just pull it out of the closet when I need a file which is almost never, way quieter, i pay my own electricity so cheaper too

>>60752136
the non-gaymur republic asus boards are usually pretty good

>>60751796
what's this? a raspberry?

>>60752232
X299 mini itx motherboard

>>60752192
For some reason Asus AM4 boards have a low rating on NewEgg.

>>60751989
They're still reaping in fairly respectable margins, and they cleared most of their short- and medium-term debt off their books a while back.

AMD is in it for the long haul. And that starts with fighting for market share.

>>60751989
>HEDT is niche
Tell that to anyone that does more than basic youtube entertainment media with their PC. Especially for profit.

>>60751989
It's niche because Intel's pricing was/is insane.

>>60752255

The crosshair's early bios had a bug that would brick boards. It has been fixed for months but people still regurgitate that it bricks. Right now on the most current official bios it is basically the best AM4 board (though one might argue the taichi is the best).

>>60752090
>mindshare
Fuck off with that word. It only applies to the autistic.
What they need is advertising. Most normalfags don't even know what a CPU is let alone someone other than intel makes them.[

>>60752005
You're not getting it. Mini-ITX is setup to be ideal for a mobile and no-frills miniature PC for gaming; it (typically) has 2 dimm slots (enough memory to feed a CPU) but not 4. It has 1 PCI-E x16 lane, and not the usual 7 PCI-E slots an ATX board has (at various sizes/speeds), and has less SATA headers than an ATX board (as Mini-ITX cases don't supply much space at all for extra storage. Likewise, heat management is a challenge in volumetrically small chassis, where things such as the DAN A4-SFX case can't mount anything much taller than the stock Intel heatsink, and you're not cooling a double Ryzen die (Threadripper) with that kind of cooling restriction.

You're essentially asking for an incredibly unbalanced workstation for which there is no use for on the market.

Games (the one big niche that Mini-ITX is nice for) aren't going to benefit from those extra 16 threads over an 1800x. Anything that _would_ benefit from the extra 16 threads would be kneecapped by no more than one PCI-E slot, no more than 2 dimm slots, no more than 4 SATA storage ports, etc.

If you need a CPU-intensive workload done, then you're going to need to go with the appropriate form-factor.

>>60752349
The Crosshairs look exactly like the ricer shit Im tring to avoid; are the Primes ok?

>>60752406
I get the point of mini-ITX, that doesn't mean nobody would have a use for 32 threads in such a form-factor. I could certainly use such a CPU in my mini-ITX build.

>>60752416
>are the Primes ok?
ye

>>60752416

At face value none of the x370 boards are worth it unless you need the pcie lanes - they are so overbuilt for overclocking you actually can't push ryzen enough to stress the boards.

>>60752458
I just want a board that doesnt look stupid. I want square heat-sinks or heatsinks whose shape serves a purpose. And no excessive graphics on the board. Im in complete shock that apparently that riced out shit is what most people want

>>60752416
The b350 Primes have decent VRMs but are pretty chinkshit in other areas.
The x370 has really good VRMs and is great value for money feature-wise, but there are apparently some issues, like the chipset heating up to over 60°C during idle for some people.

>>60752435
Consider the amount of Mini-ITX boards for Intel Mainstream vs Intel HEDT processors.

There aren't many Mini-ITX boards because the people opting for HEDT want all the features that come with HEDT (extra PCI-E lanes, extra virtualization features like IOMMU for GPU pass through, etc.).

HEDT isn't just a I WANT MOAR COREZ(tm) market. You're also paying for enhanced functionality, which you're kneecapping yourself when you limit yourself to the Mini-ITX platform.

AsRock, for example, made an X299 MiniITX motherboard, and seem to be the only distributor with interest to do so. I doubt it'll be utilized often just because of the way the form factor sacrifices many of the features X299 is supposed to give.

The same thing might happen to Threadripper. Can a Mini-ITX board be made? Maybe, but distributors know it won't sell well enough to recoup the R&D costs.

>>60752500
I never said it was practical, I was just pointing out that having double the threads is a reason for a mini-ITX board.

>>60751520
>grow up
>posts photos of cartoons

huh

Is it time to come back to AMD?

>>60752512
How the fuck would you.manage to cram SP3 socket into ITX board?

>>60752885

Threadripper will not be up your alley, but Ryzen put on a strong as fuck showing. Intel got caught with their pants down already so Threadripper made that worse by just kicking them straight in the dick. The X299 and i9 panic sale by them is telling of a company that views AMD as a very serious threat now.

If Ryzen 2 is even just a solid new archetype, I'm going back all in on AMD on everything except servers, at least until vSphere is able to support live migrations across Intel/AMD architectures. And AMD improves their virtualization which admittedly, probably not their immediate focus when launching the new architecture. It needs to mature a bit.

>>60736214
infinity fabric

>>60752931
Data fabric is based on CHT+.

>>60752920
I'm not a motherboard manufacturer.

Whats AMD doing with the new architecture for servers? Are there AM4 server boards currently or in the future? Or will they have a different socket for servers?

>>60752964
Servers use the same socket Thredripper uses.

>>60746412
>what do you do with all your free time?

I mostly make fun of imbeciles spending wads of cash on useless shit they don't need just so they can fill their empty pathetic lives.

>>60742097
Yes, it is somewhat. It's a non-negligible fraction of the wavelength at 3 or so GHz. The effect is best described with stray capacitance. There is a reason CPU dies are small.

>>60752512
There is nothing you can do with 32 threads when you're limited to 64 GB of RAM, a single GPU, etc.

You didn't read what that person said.
The reason you want such a powerful CPU is because you also have tons of storage drives and PCIe for the CPU to drive and communicate with. A big problem with servers in the past has not been that they don't have powerful enough CPUs, it's that each CPU can't drive enough components, which is why people have been going nuts over Naples and all its PCIe and lack of needing special bridges much much much more so than it being 32c/64t.
You can't run a bunch of VMs because it doesn't support enough NVMe drives and the lack of GPUs can also be a problem, for example.

Now Ryzen, 8c/16t, you can get as mITX and it makes sense. It's already limited to 64GB of RAM anyway. It already has a smaller number of PCIe lanes.

>>60753366
>There is nothing you can do with 32 threads when you're limited to 64 GB of RAM, a single GPU, etc.
But that's wrong

>>60751200
>$900 Threadripper is going to out perform this
$850 actually
Just to destroy jewtel

>>60741185
>4 core intel is pulling same wattage as 8 core AMD

JESUS CHRIST HOW HORRIFYING!!

>Zen's baseline is the Core arch's peak

how is Intel not BTFO?

>>60751520
Hot girl

So Threadripper is the way to go to build a workstation for simulation and modeling tasks?

>>60752926
Its comparable to first gen I7 and The bugs in this brand new architecture and its socket are fewer and less severe than the shit that went on with nethalhem.

The only thing intel has is single thread performance and some better multithread due exlusively to slightly higher IPC and clockspeed. In price to performance and even power efficiency they have gotten their nuts chopped off by AMD

Second gen just has to boost IPC a bit and up clockspeeds to the 4.4ghz range.

>>60753603
Yes.

>>60753603
oh fuck yes. Its going to have way more power at lower pricepoints and the Socket has 64 PCIe lanes for all of your peripherals and storage.

The intel equivalent is 44 lanes and they have their x299 boards tiered to make you pay extra money for RAID functions......FUCKING RAID COSTS EXTRA!!! Jewtel isnt just a meme....

>>60753655
Buy Xeons, goyim.

>>60753667
Never again schlomo!

>>60753721
Oy vey go- I mean friend, buy our Intel Xeons process for any kind of NVMe raids.

>>60752926
>vSphere is able to support live migrations across Intel/AMD architectures
Interesting, what's stopping this from happening? I would have assumed it would work fine seeing they're the same instruction set, the only problem I could foresee is if you had NUMA enabled.

>>60753655
>pay extra money for RAID functions
Just remove Intel already, this is not funny anymore.

>>60752920
>How the fuck would you.manage to cram SP3 socket into ITX board?
3d PCB with the socket turned sideways and only 2 DIMMs, possibly SO-DIMM.
It is technically possible.

>>60754230
>3D PCB
Or just use a daughterboard for the processor.

>>60754230
t. AsRock engineer

>>60754230
ASUS has built mITX boards in the past that had the power phase on a vertical daughter board. If you didn't care about using RAM with the tightest possible timings you can also have DRAM on a daughter board as well.

>>60751424
>It just doesn't add up logically that Intel gets caught sleeping this badly.
nah, this is basically history repeating itself as AMD fucked them up good back in the late 90s for a few years (until faildozer came out and got rekt by the p3 rebranded as "core")

>>60752255
asus' am4 offerings sucked dick at launch, every review seemed to have one (because everyone still thinks they're the best) but they were so obviously half-arsed it wasn't funny

>>60752920
put everything on risers like the x299 board does

>>60753366
>There is nothing you can do with 32 threads when you're limited to 64 GB of RAM, a single GPU, etc.
install gentoo