Are 32 Epyc cores at 2.7GHZ really going to use around the same

File: 15101108142l.jpg (228KB, 752x752px) Image search: [Google]

228KB, 752x752px

>>61059266

>>61059266
why not ?

Yes, that's how voltage works.

File: bang4buckwatt.png (49KB, 1500x719px) Image search: [Google]

49KB, 1500x719px

>>61059266
i3-7350K is comically inefficient and overpriced. There is no reason for it to exist, at all.

Yeah, AMD can have 600 cores, they're still slower in games than 2 Intel cores

>>61059266
>>61059277
TDP = heat watts generated by the cpu not power draw

>>61059403
>Thread about server CPUs
>MUH GAYYYYYYYYMES

Kill yourself back to /v/

>>61059418
You think gaming wouldn't be listed there if AMD was able to be listed? Memes like "VDI, dense, DBMS, analystics" are there to misdirect you from its lack of gaming performance

reminder to ignore and report trolls

>>61059405
heat and power are related anon.
can you into physics?

>>61059523
>The thermal design power (TDP), sometimes called thermal design point, is the maximum amount of heat generated by a computer chip or component (often the CPU or GPU) that the cooling system in a computer is designed to dissipate in typical operation. Rather than specifying CPU's real power dissipation, TDP serves as the nominal value for designing CPU cooling systems.
http://www.silentpcreview.com/article169-page3.html

>>61059405
Guess where that heat comes from...

I wonder which gayme needs 5GHz singles thread performance.

>>61059629
Dorf Dortress.

File: tom-false.png (71KB, 390x390px) Image search: [Google]

71KB, 390x390px

fake news

>>61059266
>why 32 cores cpu use more power than 2 cores cpu?

>>61059615
Out of your ass

>>61059690
Exactly

>>61059664
This is a shitty meme.

File: tom-mostlytrue.png (76KB, 390x390px) Image search: [Google]

76KB, 390x390px

>>61059702

>>61059610
>it's just coincidence that under typical load the chip draws as much power in watts as the manufacturer's specified required heat dissipation.

You do realize that the rated tdp does not matter to overclocks right? So long as cooling is sufficient to keep the chip from throttling or shutting down it'll run at full speed. Electricity and heat are both energy. You apply energy to something it gets hot. The heat generated is directly linked to the electricity drawn. Whether or not that's how they define tdp doesn't change this fact.

File: 220.jpg (27KB, 350x392px) Image search: [Google]

27KB, 350x392px

>>61059610
energy in>resistance occurs>heat out

That's how a cpu works anon. More power = more heat = higher "TDP"

Back to school you go

>>61059610
CPU isn't providing power to other compentns in any meaningful way.
The power draw is directly going to heat.

>>61059266
>seems silly
Zen is retardedly efficient at lower clocks/voltages, anon.

File: 1473141498708.jpg (217KB, 1536x1152px) Image search: [Google]

217KB, 1536x1152px

>>61059864
>>61059905
A while back I had an argument with someone going to school for semiconductor engineering over this. I'm glad I'm not the only person not retarded on this planet. I don't have a suitable image so have this.

>>61059418
>Thread about server cpus
>i3 is server cpu now, my dude

>>61059864
that's not how it work

>>61059266
Zen is stupid efficient. Their 8 cores sometimes draw less power than Intel's quad cores, that's just insane.

I mean you need to pour satanic voltages into 1800x for it to reach 6900k's power consumption.

>>61060043
>thread about epyc power consumption is actually about kaby lake i3 it was compared to

>>61059664
>>61059740
Politifact literally is fake news though. I live in the city of the fucking newspaper that owns it.

File: snapshot-2017-06-24-09:51:02.png (67KB, 973x205px) Image search: [Google]

67KB, 973x205px

>>61059266
We're starting to see some Intel price cuts in the HPC sector -- mostly likely in anticipation of the upcoming Epyc release.

I think this is the first Intel price cut of the year -- somebody please correct me if I'm wrong about this.

>>61059740
lmao this one is good

>>61060161
Yes, but a THIRTY TWO core using the same power as a bit overclocked 2 core is more insane.

>>61060718
That's KNL and it has nothing to do with mainline CPUs.

>>61060718
Probably because nobody has found a use for Xeon Phis yet. Aren't those the shitty Atom cores or something?

>>61059403
This thread is about CPUs that do actual work, friend.

>>61059496
>Memes like "VDI, dense, DBMS, analystics" are there to misdirect you from its lack of gaming performance

Yeah server grade chips having bad gaymen performance will be the death of AMD

They know it too which is why they list useless features instead of average frame times in AAA games

>>61059403
>they're still slower in games than 2 Intel cores
lol, even the similarly priced 1500X/1600 can beat that shit in basically everything.

>>61060743
What part of
>STUPID efficient
did you not understand?

>>61060755
>That's KNL and it has nothing to do with mainline CPUs.

Yeah, but there has to be some kind of reason for this **extremely** **rare** Intel price cut.

What would explain it better than fear that Epyc could cut into the lower end of the HPC space?

>>61060857
No, these are PCI-E coprocessors and no one simply buys them.

File: 1496325684368.jpg (71KB, 552x661px) Image search: [Google]

71KB, 552x661px

>>61060806
>what is the G4560

>1/4th the price
>1/2th the cores
>1/4th the threads
>110% the FPS

>>61060873

It's one of the very fastest-growing segments of the market. It's true that currently, HPC revenues are not a high percentage of overall revenues, but this market is a future play for Intel.

The timing of the price cuts is extremely suspicious. Notice that Intel released a new price list on the 18th, just 11 days after their previous price list -- just to make drastic cuts in only Xeon Phi. That smells of desperation, timed to coincide exactly with AMD making a big new reach for the low end of the HPC market.

How can anyone seriously claim that's a coincidence?

Maybe the price cuts are simply because Intel started feeling guilty about overcharging so much all these years?

>>61061082
G4560 gets rekt by even the lowest end R5 1400. And Raven Ridge APUs will compete with it directly.

File: 1498053534725.jpg (43KB, 630x630px) Image search: [Google]

43KB, 630x630px

>>61061082
>The last good thing from intel is the £50 g4560 that cannibalises intel entire <£150 line
is this what it's come down to?
lmao

>>61061096
Because motherfucking V100, anon. Not like NVIDIA will ever yield enough of this fuckers for at least few server rooms but come on.

>>61061119
Yeah, they pretty much had to obsolete the entire line of i3's to "win".

>>61059266
if you pay more that 120$ for a CPU, someone should cut off your hands

>>61061119
Ty for admitting
<150£=dominated by G4560
>150£=dominated by 7700k

Literally no reason for AYMD to exist in the consumer market atm.

>>61061172
Server CPUs aren't your typical toaster oven PC fare.

>>61061183
It's more like those are the only two intel CPUs still worth half a shit, and everything else is garbage compared to Ryzen.

>>61061186
we are not talking about server hardware

>>61061107
>4560 gets rekt by even the lowest end R5 1400.

If by "rekt" you mean being equal and superior in some gayms then yea

For 1/3rd the price of the shitty 1400 mind you.

>>61059496
intel doesn't advertize xeons for gaming either you mingbogglingly dumb motherfucker

>>61061214
Not really. https://www.youtube.com/watch?v=DbDpMWo7XTk

And like I said, the real competitor to the G4560 is the Raven Ridge APUs which aren't even out yet. It's doomed once those come out.

File: 1496423670269.png (234KB, 882x758px) Image search: [Google]

234KB, 882x758px

>>61061183
I mean if you want
>MUH FPS
then you should get back to your containment board >>>/v/

File: 1497219463659.jpg (11KB, 343x377px) Image search: [Google]

11KB, 343x377px

>>61059986
>onee-chan

>>61060756

A xeon phi is Intel learning the hard way that a gpu isn't just thousands of tiny cpu cores glued together.

File: dddd.jpg (52KB, 541x603px) Image search: [Google]

52KB, 541x603px

>>61061239
Didn't even do this on purpose
HOLY KEK

File: 1496874279794.jpg (3MB, 1312x3195px) Image search: [Google]

3MB, 1312x3195px

>>61061281
>rooting for intel
>the year of our lord
>2 shekels have been placed into you account

How GNU are you?

File: cherry-picking.jpg (99KB, 500x371px) Image search: [Google]

99KB, 500x371px

>>61061317

>>61061378
>gets btfo
>resorts to memes

Yup, this is 4chan

>>61061708
https://en.wikipedia.org/wiki/Cherry_picking

>>61059905
>The power draw is directly going to heat.

No, it's going into computation and heat.

Learn thermodynamics. Work is being done, a CPU is not a perfect resistor.

>thread about server cpus
>/v/ kids turn it into another "MUH FPSSS" episode
Fuck off already

>>61061755
>converting energy to "computation"
Lol, retard. You are the one who needs to leas thermodynamics.

>>61060718
these cpus specs looks more like gpus specs, but really shityy once. And hugely overpriced.

>>61061822
>learn
FTFM

>>61061748
Cherry picking is a meme here newfriend

>>61059405
But this anon is right.

The power in != heat produced.

Yes, they're related, but the heat is a function of power in and the resistance in the circuits.

A chip could draw the same power as another chip but have less resistance (i.e. fewer transistors) and therefore produce less heat.

>>61061708
retard

>>61061748
>>61061378
correct

But it doesn't really matter anyway. In the very near future everything will place a high premium on multi-threaded rendering.

Vulkan, DX12, and Metal are all extremely focused on enabling multi-threaded rendering. UE4, Unity, Source 2, and id tech all support Vulkan already. Every other engine is doing the same. On every platform.

Multi-core is the future. Period.

le epic win XDDDD

>I live in the city of the fucking newspaper that owns it.
And how is that relevant at all?

>>61060756
>>61061291
https://www.xcelerit.com/computing-benchmarks/processors/intel-xeon-phi-vs-nvidia-tesla-gpu/#
The Xeon Phi costs literally 1/10 the price of the Tesla it's being compared to, and it's not 1/10 the speed.

>>61062757
Nobody cares about price when it comes to accelerators.

>>61059266
>>61059277
>>61059386
No. Do you know what's silly? Using way too much voltage on your 7350k to achieve 5ghz. Anything above 1.35 is just asking to grenade it, even with liquid cooling you shouldn't go past 1.36. I'm sitting comfortably at 4.6ghz with 1.335V without even using 50W during CPU benchmarks. AMD shills will do anything to defend their 180W stock clocklets.

>>61064096
>m-muh DEGRADATION

Literally reduces the lifespan of your chip from 15 years to 11, who fucking cares

>>61064096
Clocklets, ha. That's a new one. Still, I'd rather be down 500-600mhz on the approach to 5ghz than have a quarter of the threads. Corelets are truly the greater evil.

>>61064116
I completely burnt an AMD CPU within 6 months of purchase by running it at 1.5V to the point where it wouldn't even stay stable underclcoked, it matters a lot more than you think.

>>61064157
Because there's a difference between near suicide voltages and relatively safe voltages.
And I've seen people run Phenoms, Sandybridges and Bulldozers for years clocked to the fuck wall with nothing happening.

>>61064157

>I completely burnt an AMD CPU within 6 months of purchase by running it at 1.5V

I take it wasn't a bulldozer based chip then. That arch weas designed to handle way, way beyond what anyone can actuall cool in the home environment (to wit if you go for motherboard block as well the likes of a 9590 can handle nearly 1.6v). In fact the highest single core OC ever still belongs to a bulldozer based chip.

File: 1498249432135.jpg (600KB, 1671x1320px) Image search: [Google]

600KB, 1671x1320px

>>61064096

what's with the obsession for 5Ghz, the voltage level is insane anyway

just get a 4790k and 5ghz is easily reachable under 1.4v with a top mobo and psu

but yeah below 1.3v is the normie safe zone, over 1.3v 24/7 is the lifespan zone although the maker themselves spec the chips at 1.45vish easily since heat degrades much more than voltage but as he said 1.3-1.35

>>61064178
Again, most 5.0ghz 7350ks are not running at sane voltages. The one I saw on a review site had it running at around 1.75. Those power draws are an unrealistic representation of Kaby Lake overclocks at sane voltages.
>>61064259
this

>>61064259

>what's with the obsession for 5Ghz

5ghz was irrelevant when AMD got there. Now that Intel can finally hit such clocks reasonably reliably actual megahertz (in b4 megahertz myth) is all that matters.

File: 1496344154345.jpg (231KB, 1000x1000px) Image search: [Google]

231KB, 1000x1000px

>>61059905
That's an extremely simplified way of stating it.

Right now, we don't really have a way of performing computations without generating heat (look up Landauer's principle). Thus, heat is a byproduct of computation; some form of "work" is being done to produce heat. Electrons are being forced around through transistors in meaningful ways, but since they're not physically doing any work, we technically consider all power to be converted directly to heat.

The problem is that it's much more complex than that. We know something happens because of the power draw. It's one of those things where we need to expand our terminology to explain it.

File: 1443908927995.gif (2MB, 448x252px) Image search: [Google]

2MB, 448x252px

>>61064378
Actually, there is electrical work being done.

>>61060806
>>61061082
>>61061107
>>61061119
>>61061183
>>61061317
Why are you faggots talking about consumer budget CPUs in a server CPU thread?

File: 1495960142424.png (2MB, 2560x1440px) Image search: [Google]

2MB, 2560x1440px

>>61061317
I bet you thought Fallout 4 was good

File: 1496804329875.webm (3MB, 768x480px) Image search: [Google]

3MB, 768x480px

>>61065407

File: 1471543216001.jpg (214KB, 471x979px) Image search: [Google]

214KB, 471x979px

>>61065442
Kek, but whats the sauce on the game assuming their is one?

>>61065613
I'd guess Higurashi.
t. don' play games, didn't watch the animu

>>61059386
there is a reason, intel binning
if it weren't for i3's to pad their bottom line i7s and i5s would cost far more

>>61064722
because it's all intel has to offer at this point. You'd have to be insanely high to ever consider using xeons or servers now.

>>61064378
It's called switching you retard. It takes energy to turn a FET on and off or hold it on because gates have resistance, capacitance, and leakage to the source and drain. The faster you switch, the more you lose to the gate resistance and capacitance because the reactance goes down as frequency goes up. This is basic EE, and all the terminology is well established.

File: fmax-vmin.png (55KB, 1509x905px) Image search: [Google]

55KB, 1509x905px

>>61059266
Power per core is something that everyone would do well to understand along with power increasing as a square of voltage.
Desktop Ryzen CPUs show a power per core of about 5.5w at 3ghz nominally. Some obviously better, some a little worse.

If you have 4 Zeppelin dies, each with 8 cores, and each core is pulling 4.5w~ at 2.7ghz you have a power consumption of 144w without any uncore. Factor in uncore of roughly 10w per die and you hit the 180w TDP mark.
Of course power consumption does not directly correlate to TDP, but the two can be the same if the safe temperature range and thermal conductivity of the die and package allow. In the case of AMD's 14nm LPP parts the TDP does closely fall in line with average power consumption.

In the case of intel's i3, just look at the voltage scaling of the other Skylake and Kaby Lake chips. A Kaby Lake core requires 12.5w to hit 3.6ghz. Reaching 4.2ghz requires 21.5w. A 600mhz clock increase requires almost a 75% increase in power drawn per core.
As clocks increase further, the voltage required increase, and power operating as a square of voltage really starts exploding.

>>61066342
AMD general goes by "average peak usage" for their TDP.
Which means either the (8?) 2 core max turbos run through some common programs and how much that uses, or the 2.7GHz all-core turbo for some programs.
So yeah, it'll be a bit less than the theoretical TDP of the all-core. Though if I'm not mistaken, they traditionally measured Opteron TDP differently since it's more common for HPC to run flat out a lot.

And for those that are wondering, Intel TDPs are based on their base clocks. That's why they're so hilariously off.

>>61060718
>50% price cut
god damn, the intel jews were really fucking you.

>>61059405
this