SOON?

SOON?

Other urls found in this thread:

wccftech.com/amd-cpu-roadmap-leak-7-nm-starship-14nm-naples-snowy-owl-zen-core/
computerbase.de/2018-02/amd-epyc-embedded-3000-ryzen-v1000/
twitter.com/CPCHardware/status/925348852147908608
pcper.com/news/Processors/AMD-EPYC-Rome-Rumors-7nm-64-Cores-2-Designs
twitter.com/NSFWRedditImage

>not waiting for 7nm Zen2

>Not building an htpc in april and waiting for 2019 sweet Zen2 and lowered ram/gpu prices

Joke is on YOU. I'm getting Zen+ for my miniserver then Zen 2 for my desktops.

>Not waiting till 2020 for the best CPUs

so, 2600 is 30% faster than 1600

[citation needeed]

>amd
hahahahah

2500k here should I zen+ or zen2 QQ will i get fucked in ram

>not waiting until 3020 to upload your consciousness to the cloud

7nm or bust

ram isn't an issue, ram speed is
you can overclock ram by fucking off on the timesings fairly easily, and anything short of 5ghz is not pushing speeds that demands cas latency to be low to get more gains.

look at your motherboard,
look at their ram list
get the cheapest 3200 you can get

personally, I would tell you to hold out because amd seems to be doing something interesting with zen2 that i don't have confidence will fit in a current motherboard socket, unless they make 2 cpus, one for home, one for servers.

just for refrence

im on a taichi with 32gb of 3200 ram that was clocked at 2400 till the 1.0.0.5/6 bios revision, then I can get 2966 or 2933, i forget which on a 1700 at stock

My shit is likely faster then yours, but not enough to justify a new rig at ram prices as they are today.

I think geek bench put single core at 15% higher and multi at 30% then the former amd cpu, its all people have to go on, but the numbers do kind of line up with a clock rate increase which this process is thought to be all about.

What are you doing posting here, are you done with the laundry yet?

shit, that's alot

>unless they make one for home one for servers
??

>not waiting for 5nl Zen3

Pffhh, It's like you're trying to do something in current year.

Would it be Intel™ 15% faster or actually 15% faster?

I'd say we should have more information if not a soft launch by the end of March. The fact though that we basically have 0 information on it except one questionable leaked bench leaves me apprehensive. They are playing their cards close to the chest with this one. . . Part of me thinks it might just be a 5% boost in general which is OK, but I honestly expect 10% jump at least. They've been working on Zen 2 in earnest since the beginning of last year, so I would think this refresh would have some of those improvements coming with Zen 2.
At a 5% boost it's like, eh why bother? At 10%+ it's more like "Intel BTFO all hail Lisa Su".

>lowered ram/gpu prices
doubt that will happen in our life time

Will they make an updated Threadripper with Zen+ too?

Breadtripper 2nd gen was in the slides where they announced everything so yes. I wouldn't be surprised if there were a few new epyc skus as well.

yes. q3-q4.

the next version of epyc will have 5 dies and 64 cores, and the only way everyone thinks this will work is if they have 16 core modules, and have them connect to a brainbox for all the other shit.

if amd doesn't make a server version and a home version, this would mean that home cpus are going to be mcm, and I don't know if the companies will support that on current boards even if it would technically be possible.

its geekbench faster, so fuck if I know

what I do know is if its just a base clock to base clock, that would be a new base of 4.1-4.2
the jump in multicore also suggests they did something with infinity fabric.

amd has a lot of areas that aren't too hard to fix or bring up, one of which is clock speed which 12nm is thought to be a process made specifically to address that.

nvidia is going ahead with a headless miner board, so priced right that will be what miners buy if supply holds out, especially if the means to get gpus get clamped down on to 1-2 per household.

they only aret doing a epyc on zen+

>wanting to become literal part of botnet

if they can at least reach 4.5ghz+ overclocks, then intel is fucked

There' s alot of low hanging fruit, so I'd expect at least 10% if AMD keep their trend of underpromising and overdelivering

Unless the rumored headless Mining/Cuda focused cards will deliver way better performance in relation to price, people would still flock to ordinary cards.

My guess: gaming focused lineup will significantly increase in price, with the high end card priced similiar to an Titan XP. Huge push for cloud computing
The headless, castrated equivalent at half the price.

Waiting for TSCM 7nm Ryzen 2.

Zen+ should be able to hit 4.4GHz. That's what the globalfoundries process is rated up to.

the prior process was rated at 3.0 for its target, if this one is targeting 4.4 that's not even considering overclocks.

nah, what they would do would be put a chip on board that locks current mining math, thats what I would do.

this way you get to keep your core demographic that keeps your ass afloat, the gamers, and you can sell a chipless version, either the headless one they are making, or a chipless gpu for prices miners are willing to pay.

If nvidia sell the headless directly, with 0 warranty or the lest they are legally allowed to give, then it's all profit to them, so they could sell a 1070 powered gpu for 1070 prices, and a 1070 powered miner, would likely bring significantly more profits to them

gaming cares wont go up in price, they are already at the very edge of what gamers are willing to pay, and if they don't feed them, they are leaving a billion quarterly on the table.

>not waiting forever because things will keep getting better and better

I already have a Summit Ridge Ryzen so I don't think a move to Pinnacle Ridge is worth it.
Ryzen 2 will be worth the upgrade, though.

Any kind of lock, even hardware sided ones, would only be as strong as the last algorithm it was designed against. Modern GPUs are basically very fast, matrix -crunching devices. It doesn't matter what kind of numbers you throw at it.

>Zen in Ryzen nominally draws 10w per core at 3.6ghz
>about 6w at 3.3ghz
>7nm Zen2 likely to break 4ghz at 10w
>huge implications for bringing up clocks on the low end of the scale
>45w SKUs could handedly out perform current Ryzen 1800X
>1700 could be replaced by a sub 35w part


It'll be amazing to see the power curve for Zen2 parts. 7nm Threadripper and Epyc will definitely be big wins for them.

GF said 12nm had 15% better performance than 14LPP. ('Performance' in this context is clockspeed) which means a target of 3450mhz.
Considering the current OC wall of ~4ghz on 14LPP, we should see an OC wall of ~4.4-4.6ghz on 12nm Zen+

>not wanting to not have to want anymore

>Ryzen 2 will be worth the upgrade, though.
If the epyc lineup is telling of the next generation of Ryzen, they'll be adding more cores. The lowest part will likely be a 6-core.

If they managed to actually raise fmax that high it would be a huge feat. Zen's current handicap is the voltage scaling.

That'd be pretty neat, 6c per CCX would be pretty mean for VMs. I've got 6c assigned to a Windows VM now, but the last 2c are on another CCX.

How about the APUs?

We'll see when it's released, I'm only parroting the GF slides.
Actual performance remains to be seen.

They've just released Raven Ridge, probably next year.

>Not waiting until 2030 for 500pm AMD APUs that have 512GB HBM7 and 512 Arcturus CUs.

It's like you don't even want technology to improve.

>Shintel Nigger

If I was a betting man I'd put my money on lower peak frequency, like 4.2 or 4.3ghz with XFR.
Base clocks and sustained turbos are where we're likely to see substantial uplift.

If they could actually manage 4.4ghz at reasonable power levels, meaning the requisite voltage massively dropped, then we'd see a much higher benefit lower on the power curve as well. Power consumption always grows exponentially at the right side of the curve.
If you can reach 4.5ghz at 15w per core, then you can probably reach 4ghz at 10w, 3.8ghz at 8w, etc.

Like I said, it would check it against known math for crypto, this would likely be updated quarterly just to keep gpus for gamers and whatever unchipped/headless ones they make for miners where nvidia can dictate a higher price.

nvidia made a significant amount off miners, however they did not make a higher margin, they made the 'look at all the gamers who bought our shit' prices. what nvidia wants is 'look ar all the gamers who bought our shit' and 'look at these fuckin miners man' and get both groups of people buying their shiy not one sacrificeing the other entirely.

The next EPYC will still have 4 dies.
The next Snowy Owl will still have 2 dies, too.

It will still have 4 dies, they're adding more cores per die. The expected top-end part will be 48-core, not 64.

The expected too part is 64C though.

No, it isn't.
wccftech.com/amd-cpu-roadmap-leak-7-nm-starship-14nm-naples-snowy-owl-zen-core/

Boiz ya heard Jim
>Zen 2xxx could be manufactured @tsmc

If they ever do that it will bottleneck like hell, don't think IF can handle that much cores

>epyc
they presented the embedded CPUs' with one or two dies, which means that they could do the same for desktops or servers.

Still waiting for a threadripper with no dummy dies and 16/32 GB of HBM2 on the same package.

computerbase.de/2018-02/amd-epyc-embedded-3000-ryzen-v1000/

>Still waiting for a threadripper with no dummy dies
Wouldn't that basically be Epyc processor?

>ancient roadmaps
Do you even understand what IF actually is?
Zeppelin has no HBM PHY you dumbfuck.

>ancient roadmaps
Post your source then.

Threadripper can run stable at 4.2

Yeah, they're the same thing with different sockets. I don't see AMD releasing a full Epyc in Threadripper form because it would affect the Epyc lineup.

>Do you even understand what IF actually is?

Yes. Do you?

twitter.com/CPCHardware/status/925348852147908608
It won't affect EPYC in the slightest, TR has no RDIMM support.
>Yes
That I doubt.

>>Do you even understand what IF actually is?
Instruction Fetch?

I sold my cpu, mobo and RAM last month. Was planning to get a r5 1600 but now... should I wait?

Yes only few months from now

>It won't affect EPYC in the slightest, TR has no RDIMM support.
With Threadripper being cheaper and running at higher clocks, it would absolutely threaten Eypc's market. HPC builders would be going for TR over Epyc.

Infinity Fabric. Why are you even here?

>Why are you even here?
>Instruction Fetch has nothing to do with CPUs

They won't.
RDIMM support is crucial for any non-childish tasks.

>twitter.com/CPCHardware/status/925348852147908608
Damn, check'd
pcper.com/news/Processors/AMD-EPYC-Rome-Rumors-7nm-64-Cores-2-Designs

>>Instruction Fetch has nothing to do with CPUs

I never said that. Try harder, Intel shill.

>Reddit spacing
>People I don't like are intel shills
Go suck a dick

>hurr durr Reddit
>everyone I don't like sucks dick

Having fond memories from when you sucked your dad dry, user?

>Telling you to suck a dick is the same as assuming you're always sucking dick
Why are you being so defensive, user?

are they going to release an apu in the level of a ryzen 5 1600?
have that right now with a 1070, but haven't really played any games past few months, so would be fine downgrading the gpu, without giving up the cpu.

Not wnough die size dumb nigger

Is it though?
ECC UDIMM is good enough for a lot of tasks.
So what if you throw an error, unless you're doing some extremely long-running single task that can't at all be broken into some sort of work units, then maybe you NEED RDIMM.
For anything else? Why bother - if you throw an error rerun that work unit.

UDIMM ECC is cheaper, the machines that run it are cheaper too.
I guess you do lose out on quantity, with a current max of 16GiB per module, but the spec allows for 64GiB UDIMMs

>Always keeps his cool, performs better under pressure
>Tfw amd literally had a wide reputation of frying itself back in the day, and even if that's fixed nowadays, no one can remove its past.

Seriously people, why are you still mindlessly shilling companies for no reason?

Because fags like your should be gassed

*you

>nah, what they would do would be put a chip on board that locks current mining math, thats what I would do.
How would they even do that without completely crippling gaming performance?

Literally wishing to upload own brain on someone else's computer

Intel had a reputation for overheating and being LOUD too.
Also if we're going to go more than 15 years into the past, F00F.
Oh and i820, I didn't forget.

Dual-socket motherboard with 7nm Epyc 48 x2 for 96 coarz?

Late 2019.

>F00F
Ah, the good olden times

Because Sup Forums wouldn't exist otherwise

7nm will be a sandy bridge moment and it will be glorious. Even though I already have a 1950X, if the improvements are as great as they're saying in the interviews already, I'll be upgrading then.

Both were kinda shitty back then, but that doesn't mean one can use the overheating argument vs the other. As long as the technology remains the same, heating will forever be a problem for both sides

Are u retarded or something?
Remember pentium? Remember core 2 duo? Remember the up to 50 iops loss from microcode patch? Remember raid keys in 2017? Remember 2000$ cpus throtteling from jizz on DIE?
Remember gtx 4xx? Remember 1.5 and 3.5 Feature RAM? Remember nvidia drivers cooking there cards ded? Remember gtx 7xx vs gtx 9xx Performance drivers?
Nigger either ur a hard shill or just dumb as fuck
AMD manages to compete against two bribe masters with a pocket money of R&D
It shows where the real engineers are and who the apple (s) of semicon are
Also rember competition u brainded nigger

50% iops loss*
Also if there would be a Company that would manage to bring better products than amd and would stand everything amd stands for, i would also shit on amd just like on shintel and nivea

>not waiting until 3021 to become God

AMD incel confirmed

...

...

i hate it when they announce new shit a month after completing a new build.

Hey BK when's Cumlake getting released?