RYZEN TEST ITS HAPPENING! INTEL IS GONE

>>59047899
DOA
O
A

File: Mysterious Merchant.gif (4KB, 452x523px) Image search: [Google]

4KB, 452x523px

Oy! amd is shit. Buy intel.

File: 1487480822932.jpg (18KB, 384x384px) Image search: [Google]

18KB, 384x384px

goodbye intel welcome amd to the throne

>>59047899
How is the 8core performance 10x better than single core?

>>59048017
SMT n shiet.

>>59048017
>what is SMT (16 threads)

File: Untitled.jpg (39KB, 400x435px) Image search: [Google]

39KB, 400x435px

It's truly over. Kiketel on suicide watch. Can't wait to see 1800x benchmarks.
>boosts to 4.5GHz because of XFR
>mfw

>>59047899
Pretty impressive all around. Lines up perfectly with the performance CanardPC showed in their ES preview.
Set it to 4ghz and give it a good RAM kit.

>>59047899
Why the fuck are people doing all these tests with DDR4 2133?

Makes me think it doesn't scale for shit with better memory. I've seen so many leaks that are all using 2133.

You'd think, being limited to dual channel, it'd need faster memory really badly.

>>59047899
Old CPU/mobo prices gonna drop like a stone after this. I'm hoping to get something good from the firesale.

File: Bench.jpg (542KB, 1645x947px) Image search: [Google]

542KB, 1645x947px

Here's my 5820k at 4.2Ghz as comparison.
Can't wait to see what the top model brings to the table.
I seriously hope that AMD manages to obliterate the fuck out of Intel with this new lineup, screw them and their Jewish prices.
It's about time we saw some competition in the field again.
Thank you based Keller.

>>59048085
Unlikely, most every board on the market is rated for 2667 or 2933mhz.

>>59048107
What's the point of DDR4 then if we can't have 3200+ MHz memory?

>>59048122
You realize that the rating on the board isn't a maximum of what you can hit by manually overclocking it, right?

>>59048122
What's the point of anything? Just stick with your Core 2 Duo w/ DDR2.

>>59048136
Some of the boards are rated at "3200+"

But I want to see how well it actually scales with better performing memory.

>>59048122
2667 and 2933 DDR4 are already considered overclock even on intel.
Source: http://ark.intel.com/products/97129/Intel-Core-i7-7700K-Processor-8M-Cache-up-to-4_50-GHz

>>59048136
>>59048192
>>59048253
>>59048272
I understand that, but why do Intel CPUs only "support" 2133 or 2400MHz (haswell only supported 1600MHz)? I heard you get instability when you have 3000+MHz RAM or that it limits cpu overclocking.

>>59048301
2400mhz with quad channel is more bandwidth than the CPU can use.

>>59048313
you got some facts to back that up jack?

154 cb Single core

how good is this compare to intel

>>59048401
Very, considering it's 1/3rd of Intel's price.

>>59048344
Not him, but it doesn't really matter since the theoretical performance gain of high clocked RAM would come at the expense of stability.
Even if RAM are certified to run at a let's say 3200 MHz the memory controller would still have to work out of its specifications to drive them at such speed.
Comparison are to be made with all components at their stock frequency.

AMD products being cheaper and better than Intel's, it's just like the good old days during the Athlon

JEWTEL BTFO

>>59048417
Isn't that score around the 5950X or 6950X which is $1700?
And this is the $389 CPU not the $500 one. So not even 1/4th its price.

They were asking about the single core score, though.

>>59047899
About time for Intel to get some competition, maybe if ryzen delivers the next intel gen cpus will be an actual upgrade

>>59048401
4ghz Skylake i7 6700K scores 182.
Ryzen at 4ghz would be extremely close.

File: 5409914+_fec7170fb9610920af2ec3fd588e9cc4.png (395KB, 800x1388px) Image search: [Google]

395KB, 800x1388px

Would single-core performance be about the same for the top end 6 core varients, at the same clockspeeds?

>>59048301
Its not about stability, its about the minimum JEDEC specifications, and what the manufacturer wants to spend time validating. Intel guarantees 2133mhz and 2400mhz DDR4 because they've done extensive internal testing with those, as they are the very low end and minimum JEDEC specifications for the standard.
The memory controller can work just fine with higher frequencies, but intel doesn't guarantee it because they haven't done the internal testing.

>>59048485
why do people care if Intel improves? You people say that dumb shit every thread and every comments section on sites.

If Intel actually gets a real 15% increase in performance and adds more cores instead of the 3% increase in performance per generation, they'd just use that to justify charging twice the money because they have 20% better performance.

I don't give a shit what Intel does. I'll just buy AMD who prices things fairly even when they perform better.
When they were wrecking Intel's shit with the Thunderbird, Athlon 64, and Athlon 64 X2 they still charged very fair prices.

Any idea what the ghz it run on. the xfr make it harder to judge the performance

File: tXI2dAB.jpg (298KB, 840x740px) Image search: [Google]

298KB, 840x740px

see you in 3/2

>>59048540
>why do people care if there's actual competition in the market
geez I dont know peter

>>59048512
Yes, it's the same Zen cores.

>>59048580
They're the same but there is different binning and less cores can usually overclock better as they don't spread as much heat to one another.

File: you've got mail.jpg (78KB, 509x720px) Image search: [Google]

78KB, 509x720px

>mfw I purchased 1600mhz DDR3 because I fell for the 'it makes no difference to game performance' meme.
>mfw I can only overclock it to 2100Mhz

>>59048440

With a light overclock 6950x easily gets over 2000 in this test, while it scores something like 140-150 in the single core, which is the same for most of these Intel +4 core models.
That thing is near 2000$ though
5960x at 4Ghz gets around ~1600 in Cinebench, stock is about 1300
So yes, this is a very good result and basically matches the 5960x/6900k while being 1/2 of the price.
If it does well in real world tests, it makes zero sense to buy any Intel enthusiast level CPUs in the future.

>>59048650
Still makes no difference for games.

>>59048665
It makes zero sense to but any Intel at all then. Cheaper mobos, no socket a year, no goytax for unlocked CPUs, insane price/performance ratio.

>>59047899
Wish I could even have a passing interest in Ryzen, but I must be an Intel fan. Compatibility on AMD is a lot worse than Intel. I have an AMD laptop that refuses to run a lot of older software that Intel-based computers can run fine. I don't think Ryzen would change that, or have I missed something?

>>59048665
The 6950X is 3.0ghz/3.5ghz turbo stock.

But supposedly it does like an 1850 stock. So okay, that is better.

Both being 3.5Ghz is a pretty even test. 1537 on 8 cores vs 1850 on 10 cores for the 6950X is favorable to the Ryzen.
If the Ryzen had 10 cores @ the same 3.5ghz you'd expect a 1930.

I'd like to see the test on good memory, too. Even memory for each one.

>>59048420
Wait, so if I buy RAM clocked to DDR4 3600, what negative effects might I see versus having it at 2400? I always heard it was harmless beyond cost.

>>59048650

It really doesn't make any real difference though.

Maybe 2-3 FPS at most

File: BlackOps.png (33KB, 1299x994px) Image search: [Google]

33KB, 1299x994px

>>59048784


That meme needs to die, it sure as fuck can make a difference and no, it's not just Fallout 4 that benefits from faster RAM.
Going from 1600Mhz RAM to 3200Mhz in some games would be on par with upgrading from Gtx 1070 to 1080, possibly even a bigger impact.

>>59048737
There's basically no difference between AMD and Intel they share instruction set extensions. I have a hard time believing you.

>>59048794

Only kids and manchildren would want to have 156 FPS instead 136

>black cocks 3
>1.87 ghz increase

>>59048737


this : >>59048797

I've never experienced any issue with an AyyMD with incel based programs


the only time it happened was with Novidya PhysX and an ATI card

>>59048665
The benchmark apparently ran at a boost of 3.9GHz, since the firemark score paired with the cinebench score showed a boost clock of 3.9. So I'd assume that the bench was run at 3.9GHz, which is still pretty impressive for an octocore.

>>59048797
I am telling what I have seen from experience. Have a 4790k, can play Thief: The Dark Project. It also works on an i3 550 (it isn't low end that is the problem). I have a Compaq Presario with an Athalon II. While the other two "just werk", I still have yet to figure out how to get this computer to work fine with a game from 1998. All on Windows 7.

>>59048771
Speed doesn't matter much.

Effectively latency does. Which I think the formula for that is 2000/mhz*CL.
Correct me if I got that wrong. But assuming that's right, you want a result that's less than 9. So DDR4 2400 is still very very fast if the Cas Latency is only 10, CL10 DDR4 is expensive as fuck.

>>59048856
That might be the problem. Intel integrated graphics and Nvidia "just werk", but my laptop uses whatever integrated graphics AMD uses. I guess if I had a Nvidia card it would work fine?

>>59048797
Maybe the stuff he's running was compiled on intel's compiler which is known to gimp programs that run on non-intel cpus.

>>59048771
The negative effect is that to run those DDR4 at 3600 MHz you are actually overclocking the memory controller to hell.
Side effect is that like every other kind of overclock you could get some stability issues.

>>59048899
Its cycle time x CAS timing to determine latency in nanoseconds.
Cycle time is derived from the data rate of the DIMM.

>>59048899
Wait. Why does latency matter more than speed? I thought it was speed first, then latency?

>>59048923
It depends entirely on the workload.

>>59048904

Could be it.

>>59048794
the only thing that picture shows me is that DDR4-2133 is already over twice as good as it needs to be

>>59048923
Because CPUs pretty much never use that absurd amount of bandwidth in DDR4.

It's all down to how quick they can retrieve and store to memory.

>>59048918
k?

>>59048871

Yeah I didn't really think it was running at the base clock anyways, but like you said, it's a good result.
And it's not a bad single core result either.
AMD has basically gotten to a level playing field with Intel here, while selling their products at 50% less price.
It's going to be interesting seeing what Intel's answer is.

Funny thing is, that Intel has essentially fucked themselves over big time by pricing their 10 core flagship near 2000$
Had they kept to the tradition of maxing out their flagships at 1000$, they'd be competitive even in this situation.
But nope, they had to go and slap an extra grand to the price and now they can't compete, due to their exponential price increases after the 5820k/6800k models.
Those are the only reasonably priced high end Intel CPUs on the market and they can't go toe to toe with Ryzen.

>>59048916
That is what I don't get. I thought you could overclock RAM further, but the speed you purchase is the tested stable frequency. How likely are these stability issues to appear?

I get a score of 173 single core and 840 multi core on my 4770K @ 4.2Ghz with my old MSI mobo and DDR3 1600 at 2100Mhz.

Color me unimpressed.

>>59048879
Well there were definitely AMD CPUs available at the time thief was released and I'm sure we would have heard of the game having an incompatibility with AMD CPUs. I'm pretty sure this is a software issue.

Update your directx 9.0c stuff using the web installer.
https://www.microsoft.com/en-us/download/details.aspx?id=35

>>59048971
@4.2Ghz
So effectively a 144 at 3.5Ghz
So worse.
And you have half the cores.
And it probably uses more wattage to get nearly half the score.

And your 4770k cost more MSRP new.

Why aren't you impressed? Are you just trying to be contrarian hoping someone will buy into your dumb shit, or are you retarded?

If this makes you unimpressed, I can only imagine your reaction to the i7-6900k, i7-6800k, and so on.

File: yep.jpg (11KB, 173x211px) Image search: [Google]

11KB, 173x211px

>>59048092

>>59048915
ICC just doesn't optimize for AMD CPU, it doesn't cause applications to crash or hang on them. Thief is such an old game that even unoptimized builds should run absurdly fast on anything.

>>59048975
Sadly I am not around my AMD laptop at the moment. I do believe I tried installing the web installer, but maybe it warrants a retry. I don't have much of a reason to be loyal to Intel besides the concern of my software potentially breaking with AMD, so it would be nice to allay that concern. I have just seen info in the past saying "if you have an AMD CPU, x will not work". I forget what software has said it, but I do recall seeing messages warning about AMD.

>>59048696
10-20% difference in some. Fallout 4 specifically sees a significant improvement (up to 30-40%).

>>59049004
Not that big a difference considering how old the CPU is even at 3.5Ghz. Also factor in the 840 score doubled if it had 8 cores instead of 4 it would probably be around the same MP score. Maybe a bit less but not by much I would guess. What does Kabylake score BTW?

>>59048967
That's some grey area shit, RAM are guaranteed to run at the rated frequency, but the memory controller is not.
During the testing phase RAM manufacturers use high binned CPUs and most likely pump some insane voltage in the memory controller.
That's how they get them to run stable at such high frequency.

>>59048975
>>59049036
Also make sure thief is fully patched

https://www.fileplanet.com/53360/50000/fileinfo/Thief-v1.33-Patch

That seems to be the latest patch.


There's also apparently this
https://www.youtube.com/watch?v=HHeLYADi6VM

I don't know if it's true that thief does not properly run on multiple cores but that is also a possibility.

A batch file containing:

start /affinity 1 thief.exe

would automatically set it to use one core.

There are no AMD motherboards at the moment, any news on what they'll look like?

>>59049085
back to /v/ with you

File: 1-o7CuH1Ub7TIa-xy3Kvw6Cw.png (35KB, 212x255px) Image search: [Google]

35KB, 212x255px

>>59047899
>Next Gen Intel
>i3: 4 cores no Hyperthread
>i5: 4/6 cores + Hyperthread
>i7: 8 cores + Hyperthread

I think Intel always had the technology to push this further, but jewed out every generation.

>>59049048

>if it had 8 cores instead of 4
Let me stop you right there.
If it had 8 cores instead of 4, you'd be looking at a 4-5x higher price.
The biggest thing about Ryzen is that it offers Intel's high end performance for 50% the price.
That's insane and basically puts Intel completely out of the game.

>>59049107
They've had the tech for years now, but they can't price it anywhere near where they need to.
Because if they do, we're looking at an unprecedented sudden price drop and Intel simply doesn't lower their prices.

>>59049107
They need a separate APU line then. Congrats, you just killed Intel's lineup.

>>59049088

Mostly ugly.

Asus makes the only decent looking one.

And even then there is no mother board premium design like what went in to the 2011-3 socket boards.

>>59049107
They have Xeons with 10-15 cores so yeah they were always capable of bringing up the lower end they just never had incentive.

>>59049085
Thief is indeed limited to one core. When I use it on my 4790k, I either use a fan mission loader that basically runs the batch for me as well as reinstalls a old codec known to break. Neither prevents the game from starting, but it freezes as soon as the main game starts. AMD's problem is that I couldn't even get to the main menu.

I did try 1.33, I have three different versions of Thief which I test out to ensure full compatibility. 1.14, 1.33, and Gold. I will retry this with the BAT file and the web installer. I agree that it shouldn't have issues on AMD.

>>59049139
This I don't doubt. I am impressed with the cost to performance ratio. However it's taken AMD what? 10 years to get this far? It only shows how Intel have been fucking lazy during that time and Jewing the fuck out of everyone because they had no reason to change. I welcome the competition. I'm just saying that my 'old' CPU was not too bad just overpriced and having shitty TIM. But it was the best I could afford at the time.

>>59049101
/v/ doesn't seem to talk about Thief a lot.

>>59049193
AMD had shitty management before Su. Now Intel and AMD swapped, with Krzanich being an absolute fucking retard.

>>59049199
Probably a better fit on /vr/ anyway.

>>59049199
/v/ never talks about good games unless it's by IPL, some kind of RTS or its any Ace Combat.

>>59049107
Coffee Lake is apparently going to be bringing 6core intels as "mainstream". Which I guess means a bit higher clocks and under $400 instead of $400+ for only 3.4Ghz

Afforadble 8core i7s isn't happening, though. Intel is too afraid of people using i7s in servers instead of Xenons.

>>59049237
This will change my friend if AMD gains large momentum with Zen

Do you think their $1000-1500 CPU cost just a BIT less to make. I wouldn't be surprised if the i7-6950x produced for ~$500 or less

Ryzen+Vega when?

my 2500k and r9 290 are getting old

>>59049284
Vega is releasing Q2.

i feel kinda bad for people who bought intel in the last year.

File: ASUSROG3200+.png (558KB, 1269x895px) Image search: [Google]

558KB, 1269x895px

>>59048107
>>59048122
Asus to the rescue with 3200MHz+ support

>>59049302
It's okay, the chose their destiny.

>>59049258
>IIRC the I7 4790 manufacturing price is around 50-80 (Max $100)

Buuut Intel and AMD still needs to research and create a whole new generation which cost a lot usually

>>59049296
>>59049284
VEGA will be introduced on 28th February. It will be around the market around April.

>>59049193
>This I don't doubt. I am impressed with the cost to performance ratio. However it's taken AMD what? 10 years to get this far
3 years.
They spent like 5 years shitting on Jim Keller's legacy trying to make Bulldozer work while waiting for him to lead what is effectively was is Phenom IV.

In another 2-3 years there's going to be another significant performance increase with Zen+.

>>59049258
The reason Ryzen is so cheap is because the entire like 12-17 SKU list of models is ONE chip.
They make ONE chip, lots of them, over and over, so many of them on a waffer. So many waffers.
That allows them to scale.

Whereas for Intel, the 6 core is its own special design. So is the 8. And the 10 is based on a Xenon.
Intel has too many manufacturing processes, and for no reason.

Now if they instead just made the i7-6850k essentially in higher volume, and binned the ones where a core or 2 doesn't work as an i7-8700k, then yeah. Then they could sell both their 6core and high end 4core for cheaper. That's also without an iGPU.
But I think they'd still be reluctant to sell them for nearly as cheap as Ryzen because intel is intel. They'd rather spend billions on marketing and Feminist Frequency donations than to sell things cheap.

>>59049284
Probably 4-6 months from now.

>>59049311
who the fuck uses that many USB slots anyway

waste of shekels

>>59047987
>>59047998
>>59048013
>>59048017
>>59048022
>>59048045

Cant see from the picture its a fake. /v/>>>>>

>>59049311

What the fuck is that 3D printing mount?

>>59049311
>3d printing mount
wat dat

>>59049330
6,8, and 10 core Broadwell-E are all the same die.
6 and 8 core Haswell-E are the same die.

>>59049284
Oh sorry you just meant Vega. I thought you were talking about Ryzen+Vega APUs. Yeah Vega will be in the early part of Q2.

>>59049352

Get your own question.

>>59049330
Vega is releasing in may or before (probably April)

>>59049296
>>59049315
>>59049330
>>59049371
so summer build then

I know it's all speculation right now, but what are the odds that there's a Vega gpu that can compete with a 1070/1080 for less than $400/500?

>>59049330
>Whereas for Intel, the 6 core is its own special design. So is the 8. And the 10 is based on a Xenon.
>Intel has too many manufacturing processes, and for no reason.
Basically none of this is true.

All of Intel's main line is one die. All of intel's Extreme and Xeon line are one die. The only difference is the binning.

>>59049351
>>59049352
>>59049374

3d printed cake decoration

https://rog.asus.com/forum/showthread.php?85505-Make-it-your-own-The-ASUS-3D-Printing-Project

>>59049351
>>59049352
Something you can use to mount your 3D printed cooling mount, cover, etc...
Just marketing BS tbqh

File: haswell family.jpg (2MB, 3951x1814px) Image search: [Google]

2MB, 3951x1814px

>>59049413
Intel does produce a lot of die configurations for various desktop, OEM, and mobile SKUs. That anon is just terribly confused.

>>59049330
>the 6 core is its own special design. So is the 8. And the 10 is based on a Xenon.
the 6, 8, and 10c chips are based off the 10c die IIRC, and are bottom of the barrel for that die.

Intel high-end server chips have 3 basic dies: The high core count die (which has up to 24c), the mid-core count die (which i think goes up to 16c), and the low core count die, which goes up to 10c.
Its been this way since Ivy Bridge-E (15, 10, 6 core dies for the server chips).

If anything, its their goddamn consumer dies that have too many variants. Haswell chips alone had something like 7-8 different dies for the consumer chips.

And then there's their continuing boner with making XBAWKS HUEG super-high-end dies. Starting with Nehalem-EX, they've made at least 1 ridiculously massive die every generation, with one pushing 680mm^2+

>>59049367
Oh wtf you're right. It's a 10 core die. Single CCX.

I remember there was a 6core intel die in the past, wasn't it? They were all in a row. Maybe I'm confusing it with a Bulldozer die.

Holy shit. There is no excuse for it costing so much. Only that they price it so high that they don't make them in enough volume to begin with.

>>59049402
>I know it's all speculation right now, but what are the odds that there's a Vega gpu that can compete with a 1070/1080 for less than $400/500?
The enthusiast Vega GPU was about 30% faster than the 1080 on Doom Vulkan, but that was months ago before it was really optimized.
I'd expect a $450-$600 price.
$450 is really pretty likely, but if it's that low vendors are probably going to jew due to high demand and they'll really be over $500 still.

>>59049440
I wonder how big will 32cores Skylake-EX be.

>>59049371
Dear god we will get this bullshit all over again

File: tmp_28999-85150818125414.png (53KB, 650x450px) Image search: [Google]

53KB, 650x450px

>>59049048
>What does Kabylake score BTW?

So does this affect anything else but muh gamus?

File: IvyBridge-E Compressed.jpg (3MB, 2814x2481px) Image search: [Google]

3MB, 2814x2481px

>>59049460
>I remember there was a 6core intel die in the past
Ivy Bridge E, its what the LGA 2011 Xeons of that generation were made off of

>>59049485
Probably pushing 600mm^2 again. Meanwhile AMD just slaps 4 of their 8c dies on the same package and sells that. Doing it as an MCM does have the not inconsiderable advantage of better yields that using smaller dies brings, which is cheaper. I'm really looking forward to what the 32c Naples CPU can do.

>>59049439
They could have just done 3 instead of 8 there for mostly the same resulting models binned down.

I guess part of this works because they have many different fabs each running a design.
They got away with this, and still had high profit margins, simply because they were very overpriced. But because of these things, they can't compete with AMD simply making 2 godly designs that are binned down into many models.

But yeah, thanks for backing up what I said.

>>59049489
AMD APUs were amazing for ultra-budget builds. And the new ones may actually work for a legitimate midrange.

>>59049506
why did they only o/c the top 2? Id like to see the 7700k vs 4790k when both at standard then both clocked to their max.

>>59049539
>Probably pushing 600mm^2 again. Meanwhile AMD just slaps 4 of their 8c dies on the same package
AMD are 4 cores per CCX.
The 32core Ryzen Nables is going to be 8 CCX's. So it'll likely be under 100mm^2.

Ryzen is 8 cores and a lot more cache that's slightly smaller than the i7-7700k. 44mm^2 49mm^2 to

>>59047899
Ho much is the 1700k????

>>59049581
389 bucks

File: 6th_Gen_Intel_Core_die_flat.jpg (3MB, 2052x3000px) Image search: [Google]

3MB, 2052x3000px

>>59049485
Skylake is something like 122mm2, and over half the die is IGP or uncore.
4 cores in Skylake along with their L2 and L3 caches take up 49mm2. So in core logic and cache alone 32 cores would be 392mm2. Layout that in an actual sensible way, add uncore, quad channel memory, and its north of 450mm2 easy. 500mm2-550mm2~ is a conservative estimate for the size of such a die, but it could easily top 600mm2 depending on what other accelerators are included.

File: Zen_Summit_Ridge_First.jpg (73KB, 369x676px) Image search: [Google]

73KB, 369x676px

>>59049569
CCX is just the name of the structure AMD gave to the arrangement of Zen cores. The die is called Zeppelin.
Naples uses 4 Zeppelin dies. It is not comprised of 8 individual chips.

>>59049539
>>59049569
The current estimates for Zeppelin die sizes is 160-200 mm^2, with higher probability for the larger end of the range.

Naples could really end up in 700-800 mm^2 aggregate territory (plus interposer and assembly costs), but it will have 8 DDR4 channels with integrated AES-128 encryption, 128 lanes of PCIe 3.0, 64 MB of L3, etc.

>>59049589
are these concrete prices? I mean have they been confirmed?

>>59049632
And it will still be miles cheaper then Intel's behemoths.

>>59049641
see>>59048559

>>59049653
Sorry missed that so thanks

>>59049646
Maybe. The top end Naples Opterons could be priced at $5,000-$8,000. They'd still be cheaper than intel's offerings.
Intel is asking $8,898 for their 24c/48t E7 8894v4.
The slightly lower clocked 8890v4 is $7,174.
Only a 200mhz base clock difference between the two. 2.2ghz vs 2.4ghz. Skylake Xeons will probably have an even higher pricetag.

>>59049646
that remains to be seen. until 32c Skylake-EX (or -EP?) comes out, AMD may be able to charge pretty high premiums

>>59049631
Erm you can see right there, it's 2 sets of 4.

They're using the same exact arrangement to do the 4 core for the APUs.

And the 32 core nables will be the same thing except 8 sets of those 4.

>>59049632
>Naples could really end up in 700-800 mm^2 aggregate territory
How?
Each core is only 4.9mm^2. A octochannel memory controller and the extra PCI lanes isn't going to add that much.
Even if it was 64MB of L3 per CCX, it still wouldn't be that large.

>>59049730
No, top-end Naples is 4 Zeppelin die MCM.

>>59049641
Multiple retailers all around the world have roughly given the same prices (adjusted for the varying taxes etc).

So yeah, it's basically confirmed more than anything else is. Cheap as fuck.

>>59049344
>ryzen getting btfo
>amazing... doa...
>ryzen btfoing
>FAKE

>>59049730
The Zeppelin die features 2 CCXs, but its one singular die, anon. They are not fabbed independently and put together.
Naples is made from 4 dies.

And Zen cores are 5.5mm2, not 4.9mm2. Each Zeppelin die has dual channel memory, the combined MCM has 8 channels. There isn't a single unified memory controller anywhere.

Athlon days are back!

>>59048432
>tfw Athlon 2200+ was my first CPU

AMD NEEDS TO CAPTURE HALF OF THE MARKET NOW

File: Haswell-EX.jpg (496KB, 1408x920px) Image search: [Google]

496KB, 1408x920px

>>59049730
>Erm you can see right there, it's 2 sets of 4.
>
>They're using the same exact arrangement to do the 4 core for the APUs.
>
>And the 32 core nables will be the same thing except 8 sets of those 4.
Which will use 4 of those dies, instead of a single monolithic die like what intel does (pic related, Haswell-EX). Its what AMD has been doing ever since socket G34 released, using multiple CPUs of a base die design to form higher-end hardware.

The only other Zen die we're likely to see is the APU die. AMD's ENTIRE product stack for this generation will be derived from either the Zepplin 8c die, or the 4c APU die.

2 CPU dies for AMD vs ~10 CPU dies for intel.

>>59049730
Each core cluster is confirmed 44 mm^2 including L3.

If you look at
>>59049631
you can see by simple pixel counting that Zeppelin is ~4.5x the size of a single CCX, or just shy of 200 mm^2.

>>59049754
Actually it was around in January if you read an earlier thread it was shown to be a fake.
I call the picture fake and you instantly go on the backfoot assuming im saying the cpu's are shit, whereas im actually not.

You really are stupid.

>>59049784
>2 dies for an entire product line
That efficiency. Really tickles my autism.

>>59049784
>>59049805
MCMs could be a boon for AMD, but remember that we have seen exactly zero metrics on Infinity Fabric/GMI, which is how Naples will live or die.

>>59049836
i'm expecting it to be faster than HT3.1, at least.

How much faster remains to be seen.

>>59049836
Considering they want to make gigantic MCM APUs, I'd say they are pretty confident in it.

>>59049779
They almost did during the original Athlon (XP/64) days. They were dangerously close to becoming the leader of x86 but it suddenly stopped and then reversed (for some reason...)

>>59049894
Yup, I remember that, I bought one and an amd gpu.
Now im intel and nvidia so must be a kike and shill.
Its not like I buy what I think are the best available at the time

What is the possibility of AMD releasing their response to Broadwell-D and soon Skylake-D?
What about Avoton/Rangeley and its successor C3300 SoCs? I wish AMD gave Puma another chance at the embedded server market.

>>59049920
Its pretty likely considering Zen scales down in power so well.
They have a BGA Opteron socket for single and dual die parts. No reason for them to not target the full TDP gamut.

>>59049784
Jesus fuck so much was wasted on igpus

File: aH70PIR.png (882KB, 906x638px) Image search: [Google]

882KB, 906x638px

>>59050051

>>59050066
This triggers me
could've fit a fuckton of cache there

>>59049760
I was not saying they're fabbed independently and put together. I'm just saying that's how the die would be composed, as 8 of those groups of 4 with a shared cache between them on a die.
>>59049784
Yeah I know. Bulldozer and K10 both did the same thing.

>The only other Zen die we're likely to see is the APU die. AMD's ENTIRE product stack for this generation will be derived from either the Zepplin 8c die, or the 4c APU die.
Yep, that's what I was saying. That's why they can do this so cheaply. Their entire consumer product line is 2 dies that they make a shitload of and bin accordingly into different models.

That is why I believe the prices so easily.

>>59050022
>Its pretty likely considering Zen scales down in power so well.
Is there any source for that claim? I'm kind of offput by the fact that Ryzen won't have any low or very low TDP versions at launch, while Intel has a bunch ready with their mid-to-high TDP desktop CPUs.
And what's going to happen with the Puma+ platform? I know Zen and K12 is supposed to take over for that market, but how soon? I've heard nothing from K12 since 2016 and nothing about a sub-15W TDP Zen version.

>>59050076
Cache is expensive, anon. Intel would just remove the iGPU and sell it for $200 more; claiming it overclocks better without the extra heat.

>>59049873
I'm sure some workloads won't notice at all, but the gains in yields come at necessary increases in coherency traffic latency. A 2S * 32c Naples config will have 8 if not 16 separate NUMA islands and can't realistically provide a fully connected mesh between all of them, meaning more of the same kinds of hops that plague 8S systems.

Honestly though it's probably smart for AMD to not try to be everything to everybody and focus on providing the best value in only certain large segments, leaving HPC and whatnot to Intel.

>>59050100
>cache is expensive
I feel like people would pay a premium for having 25+ mb of cache in an i7

>>59050094
What the fuck are you talking about? Ryzen is launching with 95w and 65w SKUs, including 8 core parts that are 65w.
It isn't hard to figure out what the power per core would be at a given base clock.

>>59050076
Or.. it could have been 8 cores with just as much shared cache between the other 4.

The iGPU size is pretty much exactly the size of the shared cache and CPUs.

Such a pathetic amount of cache, though, still.

>>59050100
Look at this goy we could almost hire him.

>>59050094
People are blindly believing TDP at this point. Yeah, lets wait and see. I bet it's only about 5-10% more power efficient than Broadwell-E.

>>59050094
Jim Keller I believe said that Zen scales across all power levels. AMD does not have a small core micro-architecture for low power platforms, except for K12 which is like Zen but ARM based. K12 has been put on the backburner though

>>59050135
>People are blindly believing TDP at this point
Power consumption numbers were visible on screen during AMD's Handbrake demo.
CanardPC published power consumption figures in their review. 95w TDP chips stay under 95w draw.

>>59050123
>It isn't hard to figure out what the power per core would be at a given base clock.
That doesn't mean that the architecture scales well beyond a certain voltage or frequency envelope. Ryzen can either be as smooth as Sandy Bridge with higher clocks, or an absolute wreck of a chip like Broadwell. Same would apply to lower voltages at low frequencies. No one's talking about how efficient Ryzen is clock-to-clock at lower frequencies with Intel's T/TE/L SKUs

>>59050135
>People are blindly believing TDP at this point
I'm talking about it as more of a market segment than actual thermal output, family. There's a sub-15W-25W TDP market for embedded devices/mobile, a 35W-45W TDP for "low-power" CPUs (aka deliberately gimped), and a 54-65W "normal desktop" CPU variety.

>>59050094
If we can have 8c/16t at 65W, logic dictates we can have 2c/4t, pretty standard on all but highest end of laptops, at 16.25W. Lower the clocks for lower TDP, the usual tricks, can get them down without any work to 10W. Optimized versions further.

>>59050261
Raven Ridge scales down to 4w, so its safe to say they can target low TDPs without a problem.

>>59050261
But what about 8c/16t and up to 16c/32t? Intel could scale Broadwell-D down to 45W for those.

>>59050076
Nah, it would be much better served by duping the block of cores. Cache very quickly hits a point where diminishing returns decides to power-bitchslap you in the face.

>>59050089
>I was not saying they're fabbed independently and put together.
Except they will be. Thats how the Naples CPUs will work: AMD will take 4 Zepplin dies, each containing their own pair of CCXs, cache, and memory controllers, then mate then together on the package. AMD doesn't do big dies, they cant afford it.

>>59048559
looks to me like zero reasons for anyone to buy the 1800x since it's $100 more for the same processor with a 200mhz OC

File: ap_israel_jewish_intel_sabbath_210_14Nov09.jpg (35KB, 201x210px) Image search: [Google]

35KB, 201x210px

>>59047987
>>59047899
>>59047998
>>59048013
>>59048045
>>59048055
Be silent filth

>>59050283
It seems deliberate, since only a very small fraction of the dies made will ever meet those voltage and frequency goals. People with money to spend will buy the much lower-available chips at a premium. The vast majority will buy the 1700X.

>>59050100
1/3rd of the die as cache isn't really more expensive than 1/3rd of the die as iGPU

>>59048794
Niche game cases are really few and far between though, sure you have VRAM heavy games like FO4 and CoD that were optimized for the consoles and shittily ported, but not many games abuse the framebuffer so hard that better RAM > PCIe > VRAM access makes any real difference.

The fact that these RAM kits affect performance noticably beyond margin of error tells you something about the game itself, not the PC platform as a whole.

I have an old 2013 xeon, should I upgrade to ryzen? I'm also planning to hop in on either 1080ti or vega.

>>59047899
ADVANCED
MARKETING
DECEPTION

>>59050278
Its power per core, anon.
The Ryzen 1700 is a 8c/16t 65w part with a 3ghz base clock.
The highest clocked Broadwell Xeon-D is 2.4ghz. Their higher core count parts have clocks that range from 1.5ghz to 2.1ghz.

If 8 Zen cores hit 3ghz a 65w they have absolutely no issue scaling down further. Going from 6w per core to 4-5w per core is nothing.

>>59050376
AMD made that same claim with Bulldozer, but then they had to go out and produce the Cat cores. Until AMD can prove they can do that with Zen, I'll be cautious about those claims.

>>59050333

Depends on how well does your xeon compare to something that's potentially a bit above the 6900k.

>>59050403
>AMD made that same claim with Bulldozer,
No they didn't. Why do you fucking retards just make asspulled statements like this?
AMD never claimed Bulldozer would scale down to 5w per module as competitive frequencies.

It wasn't until Excavator that they could hit 2.45ghz at 5w~ per module.

>>59050403
AMD had the Cat cores planned well ahead of Bulldozer. What they didn't anticipate was that Bulldozer would suck at its own niche and end up being better at scaling down than the Cat cores were.

>>59050428
>No they didn't
http://www.anandtech.com/show/2879/3
https://bits.blogs.nytimes.com/2011/11/07/amd-betting-on-power-consumption/
>Chuck Moore, chief technical officer of A.M.D.’s technology development group, said a new chip, code-named Bulldozer, “is designed from the bottom up to take advantage of low-power technologies.” Each chip has conjoined cores, the big management portions of the chip, which share some real estate and architecture. There are monitors on the chip that judge how large a computing load is current, and whether it requires a lot of power, or a little. “We’re plowing new ground here,” Mr. Moore said.
Now shut the fuck up

>>59050459
You just posted vague generalized statements that literally are not relevant at all.
Pathetic little kid.

>>59050459
http://www.pcgameshardware.de/CPU-Hardware-154106/News/In-2009-AMD-wants-to-challenge-Atom-653085/

Bobcat was officially named in 2007 at Computex but it wasn't until like 2011 that they finally released it. The two core designs were being worked on side-by-side for a while

The NDA lifts on the 28th?

>>59050511
Bobcat was never meant to be used in servers, idiot. Bulldozer's Opterons were, but then we all know what happened on that front...

>>59050568
Were micro-servers a thing when bobcat was in the planning phase? I think the market shifted a little and bobcat was able to pick up a niche. Bulldozer was just not competitive in servers at all they didn't lose anything to bobcat they just never had anything to begin with.

>>59050607
>Were micro-servers a thing when bobcat was in the planning phase
Yes. Intel and IBM had that market cornered, with AMD coming a distant third before 2009.

>>59049302
Well... they should have waited for ryzen. They were warned ;)

I feel bad for people buying new computers

are their old ones shit already? LMAO

>>59047899
I T
I S
O V E R

I N T E L
B T F O

>>59050459
"Low-power technologies" doesn't necessarily mean that it will compete as a sub-5W chip.
A huge 140W server chip can also implement those low power technologies to drive down power consumption, which would otherwise may be well in the 200s.
Being able to scale an architecture from laptop to server doesn't mean that it competes with absolutely low power embedded chips either, and as far as I know AMD never claimed that Bulldozer will be able to do so.
That's exactly the reason why the Cat cores were developed in the first place.

With Zen, though, AMD scrapped the Cat cores and SOMEWHAT hinted, that Zen will be able to scale down to Cat core territory.
We'll have to see how that works out for them.

>>59047899
Appears it was overclocked to 3.9GHZ. So it might just be 95% the IPC of Haswell here.

Still good, but not beating Broadwell-E.

>>59050816
it straight up says 3.5Ghz in the image...

File: AMD-Ryzen-7-1700X-Cinebench.jpg (312KB, 1674x840px) Image search: [Google]

312KB, 1674x840px

>>59047899
Jesus Christ, at least post the full resolution pictures.
This is disgusting to look at.

File: 1484178418948.png (106KB, 968x954px) Image search: [Google]

106KB, 968x954px

>>59050816

>>59050856
But another from the same source said 3.9 so eh.

It's also not impossible to fake.

I'm managing my expectations and I think you should to.

Embargo lifts in just 1 more week.

File: tmp_5036-fq6C4Np689511355.png (668KB, 1818x1040px) Image search: [Google]

668KB, 1818x1040px

>>59050816
95% of Haswell is still like 92% of Kaby Lake. Which is nice.

>>59050856
Chip has a boost clock of 3.9GHz, it was shown on the Firestrike benchmark

File: AMD-Ryzen-7-1700X-CPUZ.png (110KB, 756x838px) Image search: [Google]

110KB, 756x838px

>>59050283
Obviously, the 1800X is somewhat of a Halo product like the 6950X or the Nvidia Titan. Slightly better performance for an unreasonable price.
Also, we don't know how hard they're binning their chips yet. A 1700X might as well be already on its clock limits, while the 1800X still has decent breathing room.

>>59050883

>>59050816
>Appears it was overclocked to 3.9GHZ

You blind or what? It clearly says 3.40 Ghz in the Cinebench screen.

I think Ryzen is just on Skylake IPC.

Still wondering which x370 mobo to get...

Looking good. Can't wait to get anally shafted by UK retailers price gouging on these.

>>59050941
>I think Ryzen is just on Skylake IPC.
In some workloads it's definitely on par with Skylake, while on others it will fall down to around Sandy Bridge. You have to look at the average across all workloads, though.
On average though, I'd still put Zen somewhere around Broadwell-E.

>>59050883
>>59050907
>>59050922
I'm starting to wonder whether these reported voltages or frequencies are wrong, or if AMD/GloFo pulled some black magic bullshit out of their ass with their process.
It surely can't be able to stay under 0.7v at 3.9GHz.
Though, some slides at ISCC hinted at Ryzen operating at around 0.9V under load.

>>59050961
If you want the best possible power delivery, it will be either the Asus Crosshair, Assrock Fatal1ty or Assrock Taichi.

>>59051040

SenseMI.

It's the secret sauce to it's power consumption at idle.

>>59051051
Yeah I was considering the asus crosshair and the taichi

File: 1466370274330.jpg (51KB, 309x302px) Image search: [Google]

51KB, 309x302px

>>59047899
SHUT IT DOWN

>>59050907
Yes it's still nice.

It's just not "omg intel BTFLMAO 3 times as much money AND slower"
1/3rd the money and still 92% of the performance is fantastic but if the Cinebench is at 3.9Ghz and not 3.5Ghz it's a significant difference.

>>59050941
god you people are retarded. Shit can be faked. These are NDA breaking leaks on some chink forums. I'm just being cautious instead of acting like a 14 year old like you.

I'm planning on a Ryzen build regardless. I just don't blindly believe every leak.

>>59051058
Possible, yea.
Would be nice to see the actual voltage and wattage at load.
Zen also has a ton of different voltage domains. Will be interesting to see how this translates to real world power consumption.

>>59048650

> Running DDR3 above 1300 MHz

Are you mentally retarded?

>>59048794

No reason to buy higher that 2133 MHz I see.

>>59049045
>10-20% difference in some
>Fallout 4
Alright bro, you already listed 1 out 4-5 games that it makes a difference.

>>59051250
Kill yourself, poor filth.

>>59051265
>Kill yourself, poor filth.
Couldn't find that game on steam or anywhere else. Can you give me a link?

>>59051275
t. 1333MHz cuck

>>59051287
>t. 1333MHz cuck
Another obscure game I haven't heard about. Would you mind listing all of them in just one post and with links to where I can obtain them?

>>59047899
Oh look the SAME FUCKING WEBSITE AGAIN LOOKZ LEGIT GOYZ ZEEZ GUYZ UZE ZEEEEEEEEZZZ

File: MbaJfJA.png (19KB, 633x385px) Image search: [Google]

19KB, 633x385px

>up 6% today

so this is what it feels like to be a jew

>>59051312
wow this needs to get downvoted

>>59047899
Alpha
Male
Dicks

here is a question nobody asks, Intel got XMP will ryzen be able to OC RAM?
It's kind of important seeing how modern RAM is high CL and high mhz.

>>59051369
AMD mobos as old as the phenom days and probably older have supported XMP

>>59049343
>Not having 13 USBs on RAID
Pleb.

>>59051093
If Cinebench was running at 3.5 that would imply Sky/Kaby IPC which I simply don't see. That's why I don't feel like these leaks are all that fake. We shouldn't expect Intel tier performance but it should be fairly close. But at a lower pricetag. And this is what every leak is indicating. I'm taking a very middle of the road approach to this, nothing seems crazy in either direction outside of the prices. And that's only because we've been conditioned to accept $1000 8 core CPUs.

>>59049799
>read an earlier thread
the burden of proof is on you faggot

File: 1475317187832.jpg (29KB, 600x394px) Image search: [Google]

29KB, 600x394px

Round CPU when?

inb4 CPU bug brings Ryzen IPC down to Ivy Bridge level

Yall niggas got me hyped for the bulldozer when I first built my pc. I'm never going to trust you amd shills ever again.

>>59051757
Then wait for reviews and shit.
Only about a weeks time from now.

>>59049799
>Actually!
You've yet to provide any proof of it being fake, why should I believe you?
>You really are stupid.
Didn't meant to pop a vessel.

>>59049703
>that price difference for 200mhz

Fucking hell

>>59051636
Honestly they should be round.

The heatspreader is much longer than the small die beneath it.

The heat of the die beneath the heatspreader radiates out radially so a round heatspreader does make more sense.

>>59048823
>>Only kids and manchildren would want to have 156 FPS instead 136
If you have a 144hz monitor that difference is actually pretty important...

>>59051970
I thought >>59049703 was misleading... but no. All that anon did is state prices.

The thing is that those chips are used in servers where the licensing fees for the software on them is like $50k.

So spending another $2k each processor, $8k total, to get 10% more performance out of that software you spent $50k on is nothing. It makes A LOT of sense.

Anyway, I think Nables Zen will compete with the E7-4800 v4. So around $800 for a 16c/32t 2ghz/2.4ghz turbo or so cpu, up to $2000 for a 32c/64t at 1.4/1.8ghz turbo or so chip.
And the HBM APU Zen will displace the workstation, compute, high end workstation, and world domination segments leaving intel only with the "many core" space to compete.

>>59052120
If you have Freesync/GSync it isn't, really.

Its about fucking time AMD got their collective shit together. Intel have got lazy with nobody to challenge them. So it could be a revival of the Athlon x2 Levels of hype.

>Don't worry guys I'm sure Zen's shitty memory controler only supports 2133MHz that's why they only used that shit memory on it there's no way the memory controller can be wide as Intel's and Intel will crush it in memory performance

https://videocardz.com/66204/amd-ryzen-supports-ddr4-3600-mhz-memory

Those tears.
Is there anything else AMD hasn't demolished from Intelkikes? Your wallets?

>>59052387
>Biostar only tested what they had laying around
>random 3400mhz and a 3600mhz kit work

Neat. AMD didn't miss a step with Zen.

>>59052387
That website


Can't you guys find a reputable source yet?

File: AMD-Ryzen-7-1700X-Fire-Strike-Physics.png (587KB, 1818x1040px) Image search: [Google]

587KB, 1818x1040px

>>59047899

This is unbelievable! Do we have a time traveler between us?

>>59052628
There's a link to Biostar's site you fucking kike.

>>59052650
Anyone can time travel with Microsoft® Windows™.
Takes about 4 clicks

R3-1100 benchmarks when

File: intel.png (234KB, 800x612px) Image search: [Google]

234KB, 800x612px

>>59047987

Btfo

>>59051012
This.
UK prices are going to be insane, fucking jews always ripping us off.

https://archive.fo/2017.02.21-212855/https://www.amazon.com/dp/B06X6N8QGB
https://archive.fo/2017.02.21-212937/https://www.amazon.com/dp/B06W5Q7B38/
Thanks Reddit!

>>59057449
>AMD YD170XBCAEWOF Ryzen 7 1700X Processor & Corsair Hydro Series H100i v2 Extreme Performance Liquid CPU Cooler, Black Bundle

hmmm

Oy vey! Ryzen will be DOA! Don't wait buy Intel now, goyim!

File: AMD Vs Intel Updated for Ryzen.png (71KB, 1605x295px) Image search: [Google]

71KB, 1605x295px

>>59047899

>>59057327
You are now aware that your nanny state with its super high taxes is what fucks you.