Another, another Zen 2 Leak

[cj09beira]

3 hours ago, BuckGup said:

For what reason? 10nm IO and then 7nm cores for their stacked chipper design? 14nm Ian perfectly fine for IO and is easy and cheap to make and going to 10 wouldn’t yield that much

intel has fuck you money, so why would they limit themselves to just one research team working on a node 

1 hour ago, fasauceome said:

Sounds like the same hype surrounding Zen, I would temper your expectations. "Decimating the i9" is a pretty huge leap, I definitely wouldn't expect it.

 

1 hour ago, imreloadin said:

Who else is ready for the inevitable derailment of this hype train? I'd be really happy for a 10% improvement over the current 2x00 series...

guys the differences on the node alone are huge, just look at vega 20's frequency gains, then add to that more cores, good avx and whatever else they done, its going to be really good now we should keep our frequency expectations in check, i say around 4.8 max turbo, with some extreme editions maybe going into the 5ghz range

5.8 i wish but not realistic

[CarlBar]

2 hours ago, fasauceome said:

The hope for Ryzen was the improved single threaded, more multithreaded wasn't the priority for the consumer. They want to show that Ryzen is great for gamers, and they wouldn't do that with a double core count cinebench run.

 

And R15 is known to favour AMD CPU's and they still showed off a 16 thread part so they really didn't do that either.

[mr moose]

This threads just a mess, I don't know what came first, the AMD hype or the cries of Intel fanboys...

 

I can't see ANY domestic CPU's having a 5 Ghz base in the next few years.   And any that do will likely only have it on paper because as soon as you load up you'll either loose all the other cores or the whole thing will bog down.

[cj09beira]

1 minute ago, mr moose said:

This threads just a mess, I don't know what came first, the AMD hype or the cries of Intel fanboys...

 

I can't see ANY domestic CPU's having a 5 Ghz base in the next few years.   And any that do will likely only have it on paper because as soon as you load up you'll either loose all the other cores or the whole thing will bog down.

why though intel could at any moment release such a cpu, not with high volume, but they certainly could do it, it would have to have 250w tdp but who cares 

[mr moose]

3 minutes ago, cj09beira said:

why though intel could at any moment release such a cpu, not with high volume, but they certainly could do it, it would have to have 250w tdp but who cares 

Could? won't? might? can't?  Hasn't and isn't likely to based on trends.

 

I don't see it.   Look at the trends, AMD and Intel aren't exactly sitting on proven designs sandbagging the market.

[DeScruff]

But wait... 5.8Ghz... that sounds familiar...
Isn't that Der8auer's record for a Ryzen CPU?

[cj09beira]

9 minutes ago, mr moose said:

Could? won't? might? can't?  Hasn't and isn't likely to based on trends.

 

I don't see it.   Look at the trends, AMD and Intel aren't exactly sitting on proven designs sandbagging the market.

it is amd's 50th anniversary in a few months so they might want to do a special edition to commemorate that achievement, what better way than releasing a 5.0ghz cpu, 

[mr moose]

2 minutes ago, cj09beira said:

it is amd's 50th anniversary in a few months so they might want to do a special edition to commemorate that achievement, what better way than releasing a 5.0ghz cpu, 

This thread is a mess.

[NovusBal]

 

Starting to run out of salt for this thread!

 

But in seriousness - those clocks seem really off, unless if they are simply LN2, or extreme overclock results rather than stable daily use then i'd believe it.

 

Look in terms of CPU power, if they can hit the same SC as a 9900k with the increased core count that'd be a massive improvement in my eyes.

[ravenshrike]

8 hours ago, Taf the Ghost said:

5.8 Ghz is only happening on a Chiller, at minimum. That's going to be more LN2 range.

If they can hit 5.8 on a water chiller, that's still pretty damn good and you're probably looking at 4.6-4.8 base clock. Which puts them more or less at clock parity with Intel's top end.

3 hours ago, fasauceome said:

If this chip could hit that speed, no way Lisa Su would be debuting a mere tie with the i9 9900k as CES.

Remember that they undersold the gains on Zen 1 by 12% until its actual release. If their plan is to give as little time as possible for Intel to react to their progress while not outright lying about the chip's capabilities it makes sense.

[Stefan Payne]

2 hours ago, MeatFeastMan said:

Radeon VII provided 25% more performance at the same power, and it was basically just Vega '60' put on 7nm. There was no changes to the actual chip really other than adding an extra 8gb of vram.

Not quite, they added another +2048bit Memory Interface, increased 64bit to 1:2 from whatever the VEGA10 had and some other slight changes.

So its not just a shrink...

 

2 hours ago, MeatFeastMan said:

This thing is going to be so much better than people think it is. 

Yeah, AMD Sandbagging on CES was weird.

They never did that before.

We know they will have 16 Core CPUs.

They showed an 8 Core CPU just slightly beating the 9900K.

That means that AMD Optimized that system to be on par with the 9900K for whatever reason.

We also know the CPU might actually consume half of the 9900K CPU (+Rest of the System)

They didn't mention CPU Clockrates wich either means very high Clockrates or rather low.

The Rumormill rumors that the CPU might actually be clocked at only 4.3-4,5GHz vs 4.7GHz Intel...

And then there is Rome. Even with very low clockrates, a single Rome beats 2 Intel Chips...

 

So yeah, I agree that that looks like carnage...

[DrMacintosh]

Has to be fake. There is now way for AMD to have jumped clocks that much. 

[cj09beira]

41 minutes ago, DrMacintosh said:

Has to be fake. There is now way for AMD to have jumped clocks that much. 

5.8 is outside of reality, the top end is probably the 4.8-5ghz mark

[Stefan Payne]

43 minutes ago, DrMacintosh said:

Has to be fake. There is now way for AMD to have jumped clocks that much. 

depends what AMD has done.

Its not entirely impossible that they optimized the Pipeline a bit more for higher clockrates.

 

The Chip is based on Zen but mostly new, so its entirely possible that they modified the Pipeline.

 

Intel did the same THing from Northwood to Prescott and increased the Pipeline Lenth...

And its not uncommon to do that if you do a complete redesign...

[Belgarathian]

9 hours ago, S w a t s o n said:

Cmon guys, 16 cores at 5.8GHz in the AM4 package, what could possibly go wrong.

 

*CLC explodes due to steam pressure*

Still cooler than an i9-9900k with +100mhz OC

[imreloadin]

4 hours ago, cj09beira said:

guys the differences on the node alone are huge, just look at vega 20's frequency gains, then add to that more cores, good avx and whatever else they done, its going to be really good

Tell me again how GPU architectures are the same as CPU architectures? You can barely even compare two different CPU architectures on the same node let alone trying to extrapolate GPU node performance gains to CPU node performance gains...

[PianoPlayer88Key]

I don't expect to see 5.8 GHz or even close.  4.8 OTOH, I'd almost expect it.

 

What I'd really like to see is a significant jump in per-clockspeed performance.  At minimum, I want Zen 2 to do 500+ Cinebench R20 or 220+ Cinebench R5 at < 3.6-4 GHz single-threaded.  Better yet, Zen 2 should be a bigger jump (per Hz) over Ice Lake / Ocean Cove than Kaby Lake was over Piledriver / Bulldozer.

 

I don't want the improvements to stop or slow down there, either.  I'm actually not buying Zen 2 (unless my current system dies and even then only maybe), but am waiting for AM5, DDR5, PCI Express 5.0, etc.

 

When I do upgrade, I'd like my new CPU'l single-threaded to more than double my 4790K's multi-threaded performance (like it did to the Core 2 Duo T7250 or Athlon 64 X2 4000+ before it).  Multi-threaded, I'd like either 286-12 to 486DX4-120 size leap (in relative performance - I think it was upwards of 70-80x - times, not percent), or be able to encode to HEVC 4K max settings (placebo, q=0, all keyframes, etc) in Handbrake at >60 fps with other tasks running (4790K if I'm lucky does <0.5 fps with nothing else running), whichever is the bigger jump.

 

[CarlBar]

1 hour ago, huilun02 said:

I want the new lineup to be great, but at the same time it wouldn't make much sense for AMD to slash prices of the previous gen if the new ones are superior enough to justify higher prices

 

Yes it does. It's called clearing out the old stock. Standard procedure in many industries.

[CarlBar]

1 hour ago, PianoPlayer88Key said:

I don't expect to see 5.8 GHz or even close.  4.8 OTOH, I'd almost expect it.

 

What I'd really like to see is a significant jump in per-clockspeed performance.  At minimum, I want Zen 2 to do 500+ Cinebench R20 or 220+ Cinebench R5 at < 3.6-4 GHz single-threaded.  Better yet, Zen 2 should be a bigger jump (per Hz) over Ice Lake / Ocean Cove than Kaby Lake was over Piledriver / Bulldozer.

 

I don't want the improvements to stop or slow down there, either.  I'm actually not buying Zen 2 (unless my current system dies and even then only maybe), but am waiting for AM5, DDR5, PCI Express 5.0, etc.

 

When I do upgrade, I'd like my new CPU'l single-threaded to more than double my 4790K's multi-threaded performance (like it did to the Core 2 Duo T7250 or Athlon 64 X2 4000+ before it).  Multi-threaded, I'd like either 286-12 to 486DX4-120 size leap (in relative performance - I think it was upwards of 70-80x - times, not percent), or be able to encode to HEVC 4K max settings (placebo, q=0, all keyframes, etc) in Handbrake at >60 fps with other tasks running (4790K if I'm lucky does <0.5 fps with nothing else running), whichever is the bigger jump.

 

 

Be prepared to wait 2 years then.

[PianoPlayer88Key]

48 minutes ago, CarlBar said:

 

Be prepared to wait 2 years then.

I was anticipating 2 and a half or 3 years for my upgrade or so.  like, either Q4 2021 or Q1/2 2022.

My dad bought his 286-10 in January 1989, and his 486DX4-120 in October 1995.  I think the 286 CPU would have been about $275-300 or so, based on ads for the 287-10 math chip, which I think uses the same silicon & would likely be about the same price.  Invoice said $940 but that was a bundle of several parts, not the CPU alone.  The invoice for the 486 says it was $102, so a litlte more than 1/3 the price or something like that, for about 70-80x the performance I'd guess.

 

I bought my 4790K in January 2015, so October 2021 would be about the same time interval.  (Also my Corsair AX760's warranty I think should expire in January 2022 or thereabouts.)

 

[alatron978]

21 hours ago, NunoLava1998 said:

I think he meant 4GHz and 4.8GHz. 6GHz is stupid as fuck lol

I know it's stupid haha but the listing does say 5.0 and 5.8 (probably fake) but if not RIP INTEL

[Mihle]

AMDs showcase at CES was not really to show performance, it was to show efficiency, how much power it used compared to Intel. If they wanted to show performance, they wouldn't have used a 8 core CPU.

 

This "leak" is total bullshit btw.

[Stefan Payne]

3 hours ago, Mihle said:

AMDs showcase at CES was not really to show performance, it was to show efficiency, how much power it used compared to Intel.

It was a bit of both.

It shows that AMD, with an unknown frequency, is able to keep up with Intel at the same Core Count.

Though it was weird that they showed a mid range chip that might not even ever exist in real life...

 

They absolutely didn't show their best!

3 hours ago, Mihle said:

If they wanted to show performance, they wouldn't have used a 8 core CPU.

True but they also kept the Frequency a secret.

The Rumor mill rumored that the Chip might be clocked at 4.3-4.5GHz.

 

3 hours ago, Mihle said:

This "leak" is total bullshit btw.

Lets just say its rather implausible but also totally possible even if the possibility of those high frequencys is not that high...

 

You have to forget everything you know about ryzen as it is a new chip based on the Ryzen Design, not just a shrink but a redesign.

 

People rumored that they gave up the CCX per die -> 1 CCX = 1 Chiplet.

 

Also AMD already has experience with higher frequency designes, though it was rushed. But if you're mean you would say that Zen has the Bulldozer Execution units but got a new frontend and L3/Memory Controller design...

 

With Zen1 they played it safe and optimized for efficiency and not clockrate. They needed a hit...

Now they might redesign the chip to increase the clockrate...

They have experience and know what to do. The only question is: Will they do it?

Though that might not be something we see in Desktop soon because of the high power consumption...

 

A 5-6GHz Chip, even if at 7nm, would be somewhere around 200-300W. 

At 3-4,5GHz we're at 100W or less.

[GoldenLag]

5 minutes ago, Stefan Payne said:

A 5-6GHz Chip, even if at 7nm, would be somewhere around 200-300W. 

At 3-4,5GHz we're at 100W or less.

the question is how high voltage we can put into 7nm and how the curve behaves. 

 

that will be the deciding factor in this. 

[CarlBar]

5 hours ago, PianoPlayer88Key said:

I was anticipating 2 and a half or 3 years for my upgrade or so.  like, either Q4 2021 or Q1/2 2022.

My dad bought his 286-10 in January 1989, and his 486DX4-120 in October 1995.  I think the 286 CPU would have been about $275-300 or so, based on ads for the 287-10 math chip, which I think uses the same silicon & would likely be about the same price.  Invoice said $940 but that was a bundle of several parts, not the CPU alone.  The invoice for the 486 says it was $102, so a litlte more than 1/3 the price or something like that, for about 70-80x the performance I'd guess.

 

I bought my 4790K in January 2015, so October 2021 would be about the same time interval.  (Also my Corsair AX760's warranty I think should expire in January 2022 or thereabouts.)

 

 

Fair enough i've just seen a lot of unrealistic expectations around stuff ;).