French frog Zen ES benchmarks

Why is none of you fucks talking about this?

Seems to be running at 3.15GHz with a 3.3GHz turbo, so definitely not the up-to-date Ryzen, but close.

>not the up-to-date Ryzen
That's why

>3.15/3.3GHz vs 3.4GHz

>caring about what french people say

ou vì, l'baguette au frèquence 3.15GHz, fellow french frog.

Same french canadian who published benchmarks for Athlon 64 eight months before release.
What this means?

it means we are 8 months away from release

Brilliant!

Translated French article:

"The all core turbo is probably even lower at 3.2-3.3 Ghz. 6900k runs at 3.5 Ghz in FP workloads like Blender which are more power hungry than INT workloads. In most games the 6900k is running at 3.7 Ghz as very few games scale to 16 threads (maybe BF1) and that gives it a 8.8% clock speed advantage at minimum. The 6700k is running with a turbo clock speed advantage of 23.5%. 4200/3400 = 1.2352 . We still need to wait for final product to come to meaningful conclusion. One thing seems to be clear - this chip is definitely Broadwell class for IPC."


Looks pretty good

BASED

>barely edges out a 6 core Broadwell-E at multi-threaded programs
>can't even match a $190 i5-6600 in gaymen
>uses more power than Haslel and Skykek
AyyMD is DOAyy

>168 registers
O_O

>6600
It's running at a slower frequency than that cpu nigger :^)

Still
>$500+ AyyMD CPooU is slower than a $200 Intel Masterrace CPU
It's easy to see which one is the better buy.

The 4c/8t version will be 150 usd and clocked higher with the multiplier unlocked :^)

AYYYYMD IS FINISHED AND BANKRUPT

They tested an engineering sample running considerably lower than final silicon.
Zen has near Broadwell IPC. They explicitly stated so themselves.

You tech illiterate retarded shill can't spin this.

Seems like IPC is at or very close to 6900k. I'm really interested in how overclockable the final zen chip will be. I expect the more expensive model to be 3.6-3.8 ghz base, I hope it will clock to 4 ghz, this would make it at least 1.2x faster.

I'm expecting 3.4Ghz base, just like Dr. Su said, however the boost I think will be completely dependent on how much cooling you can throw at it.


I'm excited for the era of exotic and new cooling designs to take advantage of AMD's dynamic turbo boost.

The turbo scheme isn't new, its still based on the thermal margin system they've been using for years. You could argue that its already present in Carrizo and Bristol Ridge. The only exception being that AMD explicitly list max turbo speed rather than having it be up in the air. The chips clock to what TDP can sustain, and if you have a bigger heat sink you can sustain higher clocks for longer.

The "XFR" is likely a few extra boost states on top of the stock turbo. They'll likely still be power limited, and they likely won't be very big. Its a simple principle that transistor leak more the hotter they are. If you can keep a chip cooler you can extract more frequency for the same power draw. It may only be 50-100mhz but its essentially free if you keep it within an applicable temp range.

IE
3.4ghz base, high granularity in pstates up to 3.7ghz all core boost, peak single core boost of 3.9ghz, all core XFR max of 3.8ghz, single core XFR max of 3.95ghz.

Skylake has 180 int, and 168 fp.

Bulldozer 2.0

>58142436
They stated granularity of 25Mhz and it doesn't seem like it'll be just another boost state above the stock turbo. Precision Boost is supposedly that, and then XFR goes above Precision Boost if there's still headroom.

>Skykek
it's Skylel you idiot

>i5-6600
>60 Watts

What?
That can be load consumption.

I can already see it.

AMD Ryzen will offer good multitasking performance but has a bad price-performance ratio regarding anything but heavily multitasking stuff, for instance gaming.

>AMD still can't into a proper SpeedShift competitor
One of the many small details that will cause AMD to lose out on marketshare yet again.

>SpeedShift
Literally what?

Ryzen has been confirmed at Broadwell/Haswell levels for Single core performance, getting that with good multitasking Is a good trade against Skylake.

I hope decent AM4 boards won't be as pricey as Z170/Z270 boards.

They probably will be but end up offering a better featureset.

It seems like AMD's going to be slotting themselves in right between the super high end Extreme chips and the HEDT chips Intel already has out, here's hoping it works out for them.

AM4 boards also have to serve cheap current Bristol Ridge APUs, and the coming Zen based mainstream Raven Ridge APUs.
There may be a few special high end boards, but I doubt the majority will be expensive at all.

Anyways, here in Germoney, an i5-6600K costs 250€.
Any other non-K CPU is shit because no overclocking options.

An Mini-ITX board costs 160€.

>SpeedShift
What?
Do you mean EIST?
If so, EIST is terrible.

Hes referring to pic related.
Ignore him, hes retarded.

Then you should be happy that all Am4 cpus are multiplier unlocked, even peasants can afford overclocking now.

For the low end there's a section for motherboards without southbridges, those should be obscenely cheap, I'd guess no more than 40-50$ range.

wtf is this shit

zen has integrated vrms on die while sky/kabylake don't so zen will obviously have better scaling

>typical AMDrone spouting nonsense he doesn't even understand

>wtf is this shit
Pretty self explanatory

>zen has integrated vrms on die
Does it?
Point to them.

AVFS/DFVS is not VRM.

>implying I can point to them

[citation]
this amdrone slide

You can see them.
The structures take up a significant amount of die area. Intel's parts originally had to use an MCM package to fit a separate die with the IVR on it.

>You can see them.
Are you trying to say you don't see any on the zen die??!?

So all of the rumors about its performance are true, and it's an equivalent to Skylake/Broadwell? If this is true, then a Zen 6c/12t CPU overclocked to 4.4ghz is going to be a god-tier purchase.

>shit gaiming performance

No thanks

You finally put 2 and 2 together. Good for you.
There is no on die VRM as you tried to claim, nor is there any on the Summit Ridge/Ryzen package.

Not all voltage regulation is the same. You fallaciously conflated things.

You talk like a real bummer faggot...

>gaming performance about on par with 6600k
This is pretty much all I wanted out of it. If it's a similar price and they offer mini itx boards at reasonable prices I'd probably end up getting it.
Still a bit early to make full fledged assumptions considering it's not the final product, is very likely on a dev board, and the sample data is pretty limited.
Looking forward to proper reviews when they come.

The ES they tested had a 3.15ghz base clock, and it wasn't even reaching its listed turbo frequency. Performance of retail chips will be even higher.

Final clocks would be higher and with turbo it would be near i7 performance. And you don't need expensive 8c/16t high end models for gaming, 4c/8t or 6c/12t models will have plenty performance

Question: do AM4 mobos support ECC RAM?

...

dude nobody that does anything serious uses fucking AMD.

Some of them probably will. ECC support is baked into the die since they're only using one design for enterprise and consumer parts.
Just like with prior Vishera parts.

ECC RAM will likely become very cheap over time.

Measly AM1 Kabinis have ECC support. They are popular with NAS builds

This is seriously all I care about at this point.

Of course, this is pretty much what I'm saying. I just hope the 4c/8t option comes sooner than later, as it would be pretty hard not to suggest it as the gaming cpu of choice, especially if the option to overclock doesn't come at an unjustified premium like it does with Intel.
I also hope they don't try to price it above the 6600k with the justification of "well it's got 8 threads like the i7!" Since that doesn't matter to me at all.
As far as I'm concerned the 6600k is the processor to beat if they want to appeal to the general consumer looking to build a gaming PC

Pretty disappointing DESU. Great IPC, shit frequencies.

>but muh engineering sample
They are going to struggle hitting 4 GHz w/o liquid nitrogen. Intel will slaughter on single-core, and probably their 6-cores will beat 8-core Zen in MT workloads.

This is probably GloFo's fault more than AMD's.

At least they'll compete in price, but it sucks they won't be getting good money from this.

> shit frequencies.
AMD already explicitly stated that consumer chips will have base clocks of 3.4ghz and higher. For an 8c/16t chip that isn't bad at all. In fact its exactly in line with intel's own Haswell-E and Broadwell-E chips.

>They are going to struggle hitting 4 GHz w/o liquid nitrogen
Bullshit.

>This is probably GloFo's fault
Its no one's fault, because theres nothing wrong here to point at.

Hmmm... Do you think they'll support RDIMM's?

You do realize that any Broadwell-E, the worst clocking chips Intel has put out, can easily do 4.2 GHz on all cores on air, right?

Skylel-E / Kabylel-E will likely hit 4.5 (tougher to cool though).

AMD chooses the stock clocks much more aggressively. You have no idea how much I want Zen to rock, but… it's AMD. And GloFo.

So, when are they gonna release anything at all?

>waiting for AMD new videocard
>waiting for AMD new CPU system

Damn it, come one, give me a date and speed it up.

Time is money after all.

Aren't AMD better at video encoding and have better iGPU performance?

I just need to encode videos and HEVC 60fps 4k support for future proofing.

Here I was trying to be nice so you wouldn't have to keep making an ass of yourself, but you had to keep going.
We're comparing stock clocks, not what something can overclock to, not what your shitbrained speculation is. Ryzen has stock clocks in line with exactly what intel is offering in identical core count parts. For you to insinuate in any way that this is a problem is entirely basless.

Still trying to imply that there is a nonexistent issue because Global Foundries is involved really highlights how childish and flat out stupid you are. They're running Samsung's 14nm LPP process, not something they developed internally. Even if that weren't the case they're an entirely different company now since absorbing IBM's entire foundry business.

3.4ghz and higher inside of a 95w TDP.
That isn't bad. Not even remotely.

Take your childish FUD to Sup Forums and stay there where you clearly belong.

>AMD Ryzen Gaming Benchmarks – comparable performance to the Core i5-6400

>basless
Baseless would be ignoring the last 10 years.

>Samsung's 14nm LPP
Like Polaris? Not a great showing.

Funny you should mention this, too, because when Nvidia moved the low end Pascal parts (GP107+) from TSMC 16nm to Samsung 14nm, the clock frequencies went to shit. But yeah, that's that and this's this, right?

>3.4ghz and higher inside of a 95w TDP
That's not bad (not brilliant either), but it doesn't mean you can actually clock it higher.

>Baseless would be ignoring the last 10 years.
90nm and 45nm processes were fine.
You're talking out of your ass. You're blindly regurgitating things you know nothing about, and you can't get away with it for the single fact that I'm smarter and more informed than you are.

>Like Polaris? Not a great showing.
Actually it is. You're just proving that you're legitimately ignorant, bordering on violence.
It clocks higher than any prior GCN ASIC, and does so with less voltage. A considerable increase in fmax with lover drive current. That is impressive. Considering the ASIC itself only pulls 110w, and its default voltages are set higher than necessary to increase the number of dies that passed validation is very impressive. Simply going into Wattman and changing the power target setting can shave off 15w from the ASIC power draw without losing any performance.
That makes 14nm LPP look incredibly good. All you can say negative about the RX 480 is that AMD rushed them out haphazardly so volume would be high.
The fact that high binned versions are now in new Apple products should tell you something.

>but it doesn't mean you can actually clock it higher.
Su said in no uncertain terms that Ryzen CPUs would have base clocks of 3.4ghz *and higher*
They have turbo frequencies higher than base clocks, they additionally have XFR on top of this.
That does in fact mean they can clock higher than 3.4ghz.
You cannot spin this any other way, tech illiterate retard.

You are blatantly lying in an attempt to spread FUD like a butthurt little vidyagame company fanboy. Go back to Sup Forums where you belong and shitpost about the console wars or whatever you children are doing these days.

Signs point to all chips being unlocked, but possibly locking overclocking to higher end boards, probably B350 and X370

January 18th.

>I'm really interested in how overclockable the final zen chip will be.
This is what will make or break Zen in my eyes. IPC seems to be around Haswell or Broadwell, at this point I'd say there are too many leaks pointing to this range for it to be significantly above or below. This means it needs to hit say ~4.5GHz to be a worthwhile option as far as I'm concerned. If it can manage something like >4.7GHz on a custom water loop I'll be quite happy, even if it's just a 6C and not 8C model.

Those low end boards have barely enough power delivery, they are useless for overclocking either way

>At least they'll compete in price, but it sucks they won't be getting good money from this.
If servers and datacenters can break away from Intel's shit ecosystem with more cores/threads then they will.

They don't give a shit about overclocking just so long as they have some amount of performance, and the real money is in datacenters and mainframes anyways. If they can offer 32c/64t for the same or less money than intel with similar or better features then big companies will devour these chips.

I'm more worried GloFo won't be able to hit their fabrication targets for volume with these benches.

>It clocks higher than any prior GCN ASIC, and does so with less voltage
It better do, going from 28nm to 14nm! But they aren't competing against 28 nm, they are competing against Intel's 2nd 14nm process. Nvidia on TSMC 16nm still gets better perf/area and better perf/power.

>The fact that high binned versions are now in new Apple products should tell you something.
The only thing it tells me is that Apple goes for the lowest bidder, at a huge expense to its consumers, since Nvidia's mobile parts are superior.

Apple doesn't care about performance; they were shipping a fucking Cape Verde GCN 1.0 chip from 2012 last year on their "Pro" models. They go for the cheapest "acceptable" chip they can find.

Also, good job ignoring the only 1:1 process comparison we have (Nvidia's Pascal).

>and higher
How higher? That's the whole point. Intel can do 4.5 easy, in parts you can buy right now. The fact that AMD is tight lipped has never been a good omen.

I really disliked how they not only disabled turbo, but also downclocked Intel's chip 200 MHz on their first comparison. Couldn't they find an ES chip capable of sustaining 3.2 GHz for 60 seconds?

>The fact that AMD is tight lipped has never been a good omen.
>AMD making a big showing of their parts has never been a good omen
Jesus fuck pick one and stick to it you shill.

...

No doubt the 32c part could be absolutely great for datacenters if it performs well.

Are there any details about it? # of sockets? Memory channels?

True but the chipset isn't exactly what determines that. If the board has enough phases and utilizes quality vrm's and caps, it ought to be able to handle a decent boost in clockspeed.
I don't agree with locking even the option of overclocking behind a chipset that effectively doubles the price of a board.
The fx series boards were pretty fairly priced, I'm just hoping the am4 options follow suit.

Tech illiterate Sup Forums child, you really need to quit trying.
The Ellesmere die used in the RX 480 is not designed to be dense, it is designed to yield high. Trying to talk about performance per die area here shows you couldn't be any more clueless. It is the least dense GCN design they've ever made.

I see you utterly backtracked on your blatant lies about clockspeeds. Realized you were cornered so now you're trying to weasel your way into another irrelevant argument.
You're so glaringly stupid that its not even worth responding to your childish FUD bullshit.

Big showing?

A big showing would have been a comparison against a 6900K with default clocks and turbo enabled measuring performance and power usage using several different benchmarks and publishing the results.

Right now we don't even know the retail SKUs. When is this thing even coming out?

They did compare it directly against a bone stock i7 6900k. The i7 system did have turbo enabled, they were extremely clear about that.
They did show power consumption of each chip during the bench runs. The Head of PCPer even pointed it out on twitter.

Is someone paying you to be this stupid or do you do it voluntarily?

Next he'll pick on your part of the statement where they didn't do several different benchmarks in spite of them doing a benchmark which is in Intel's recommended set of benchmarks.

Notice how he is trying to deflect to them not having announced SKUs or a launch date, in spite of them saying several times both things would be announced at CES.

93% of 6900's hit 4.2
76% can hit 4.3
36% can hit 4.4

pretty much all 8 cores are like this, they do not hit the highs that 4 cores can,

such as the 6700 21% are able to hit 4.8ghz (it's the only number closer to the 36%)

but realistically at this point both chips are going to drive any game passed 144hz, so in a gaming sense they are equals, but load balancing is going to be so much better on the 8 core.

An ES with any number of bugs and peculiarities that's running a turbo clock lower than the absolute minimum base clock of the retail parts isn't close to the retail product.

Most reliable speculation pits this thing around haswell/broadwell tier performance, and the recent benchmarks (ie the reliable ones) indicate it might well be significantly better with integer operations. it'll likely fall short of skylake but not by much. that's huge for a company that was considered essentially doomed a year ago. Anyone saying otherwise is memeing and needs to fuck off back to their console war threads on Sup Forums.

intels 8 core are in the single digit % range for ones that hit 4.5 under any condition.

What you are seeing is something simple.
If this is real, it gives us a new minimum.

Take a look at the performance again.

just below a 6900 in almost anything work related
sandwiched between a 6600 and 6500 in gaming, with average power consumption lower than intel's 8 core.

this is showing me that for gaming, it meets my minimum needs, and for heavy lifting applications, its exceeding my expectations by a fair amount, and this is on unfinished, slower then stock, possibly bug ridden die

also, do note, with gaming, the numbers are the important thing here.
the 6700 is at 118.2%
the zen es is at 97.3%

fuck the points and its 118 over 97, that means that amd's cpu is 21% slower then an i7

this means to hit 60fps, that i7 needs to hit about 72fps
and for amd to hit 144 the i7 needs to hit 174 fps.

these numbers are not bad

>possibly bug ridden die
Let's be real here, even if this is an ES and obviously unfinished, this is a sample AMD sent out for the press to test. It's not going to be a "bug ridden die" even if the retail chips will clock higher.

>this is a sample AMD sent out for the press to test.

No its not. Its a sample from a mobo manufacturer used to validate their boards.
AMD haven't sent out any press samples yet. The closest thing to final consumer products in anyone's hands is an OEM sample.

Not sure where you're getting that, but if it's supposed to be a chip to validate mobos it most definitely can't be fucked up and buggy, even if the clocks are lower.

> it most definitely can't be fucked up and buggy

They most definitely can, and tend to be. The manufacturer just needs to get the board to be able to post with the chip in place. Its extremely common for chips to have chicken bits flipped on that totally disable entire portions of the die so that they can't be sold and used to violate NDA. Its also the case that numerous massive hardware bugs have made it into final production silicon. It has no effect on being stable enough to post.

I'm more interested in the price, I almost bought a 6700 last week and decide to wait because of the Zen livestream.

Any prediction when they gonna announce the price first before the release date?

Four core is probably going to start at $350. Six core at least $400 and the eight core above $500

At CES probably

Not a single budget pricing line?

That can't be true, specially for AMD where their most expensive CPU was around $300 at launch 4 years ago.

Their budget CPUs are Bristol Ridge which are priced up to $150.

And for reference this is how we know its an engineering sample
2D3151A2M88E - thats the OPN code

The first digit, 2, indicates that its a prototype/ES chip.
D indicates that its for a desktop socket
Then comes the frequency
a bunch of other indicators for socket type, core count, revision, and etc.

Don't listen to the moron.
Over a year ago an OEM *hinted* that then titled Summit Ridge CPU would top out around $500, and the mainstream Raven Ridge APU would top out at $250.
Thats been the expected ballpark pricing, and subsequent leaks have all fallen in line with that.

They aren't pricing anything above $500, and the lowest binned parts sure as fuck aren't going to be $350.
Su stated that they wanted to shake off the perception of AMD being the budget alternative, and later said that ASP would be higher than any prior FX parts. Though they still don't have the brand strength to compete with intel 1:1 on pricing. They're trying to win back market share, and value is the best way to do it.

But both FX and their APU series was priced almost the same, why the hell it would be difference now?

Their shitty eight-core flagship is barely on par with the i7-6850K. You know they're going to price it like it in order to "compete", even though they're conning consumers into buying half-assed CPUs that have less features and capabilities than their Intel counterpart.
Hell, I don't even think Intel has to do any sort of price drop to respond to Zen, AMD is going to shoot their own foot with pricing or overhyped marketing biting their asses again. People still remember Bulldozer and the lies AMD's marketing team made to polish their turd.

Was meant for this

Do you have any self awareness at all? You sound like a 14 year old dumbfuck on Sup Forums.

Intel is making 6 core desktop chips mainstream with Coffee Lake, and this is a direct response to AMD's Ryzen CPU line.
Keep on shitposting, under age shill.