Possible Kitguru leak - Ryzen 2 Feb launch, 12 cores and 4.6Ghz?

[N4CR]

Saw this slide on Reddigg and surprised to not see it discussed here. Frequency range is close to what I expected, however think 5Ghz is BS without serious OC, single core perhaps? Doubt it. Speeds approaching 4.5Ghz should be no problem with a small process change, moving away from their successful LP process. Not sure on the source of the slide though beyond KitGuru. 12 cores has been mentioned in past by AMD (thought this was for Zen2/7nm though), plus the speeds of some of the CPUs are not out of feasibility for a non low power 12nm process. That said, the move up from 8 core is highly unusual for a respin/shrink, especially given the interval between them.

Source of slide is here https://www.kitguru.net/tech-news/zardon/leo-says-ep4-intel-rant-week/ at 08:15
Anyone else have further info? 5.1GHz smells of fake to me but other parts check out within expectations, if any of this is partially accurate, the IDF could be in for some very long shifts..

edited a few mistakes while tiredposting earlier.

 

[Araxie]

Its not being discussed anywhere because no one believe it could be true and are taking big mountains of salt with that info...

I think that kind of leaks just hurt AMD as potential people looking to buy ryzen or even threadripper will just hang on and "wait" until cheap 12 core high clocked Zen.. so.. lot of potential sales just wasted..

Don't take me wrong if its really true I will be the first one on buy at least 6 of those 12 core ryzens.. lol.

 

[N4CR]

Its not being discussed anywhere because no one believe it could be true and are taking big mountains of salt with that info...

I think that kind of leaks just hurt AMD as potential people looking to buy ryzen or even threadripper will just hang on and "wait" until cheap 12 core high clocked Zen.. so.. lot of potential sales just wasted..

Don't take me wrong if its really true I will be the first one on buy at least 6 of those 12 core ryzens.. lol.

Fair point but AMD has already talked about Zen+ refresh early next year, so logically, part of it could at least be accurate. Current process is LP, if they get ~10% from a non LP process, that's ~4.4GHz. Even if it's 8 cores, (which I expect) this isn't so bad. Agreed though, I'd even pull the trigger if they can manage mid to high 4GHz and improve the few teething Zen issues.

 

[Gideon]

Perhaps possible but I think highly unlikely it will be hitting that high of a clock speed.

 

[IdiotInCharge]

Fair point but AMD has already talked about Zen+ refresh early next year, so logically, part of it could at least be accurate. Current process is LP, if they get ~10% from a non LP process, that's ~4.4GHz. Even if it's 8 cores, (which I expect) this isn't so bad. Agreed though, I'd even pull the trigger if they can manage mid to high 4GHz and improve the few teething Zen issues.

It should be possible for AMD to raise their clock ceiling, and if they've been able to address their RAM troubles and interconnect latency issues with a 'Zen+' stepping/respin, they should be able to track a lot closer in performance across the board.

 

[N4CR]

It should be possible for AMD to raise their clock ceiling, and if they've been able to address their RAM troubles and interconnect latency issues with a 'Zen+' stepping/respin, they should be able to track a lot closer in performance across the board.

This is what I'm hoping/expecting in a more realistic manner and it would be enough to make them pretty close to Intel in gaming, thus enough for me to pull the trigger. All round use is my problem and I'm not playing the dead platform game with the 8xxx series. 12 cores this soon though I think is BS. Otherwise I have to consider Intels' bigger sockets which last longer typically.

 

[IdiotInCharge]

This is what I'm hoping/expecting in a more realistic manner and it would be enough to make them pretty close to Intel in gaming, thus enough for me to pull the trigger. All round use is my problem and I'm not playing the dead platform game with the 8xxx series. 12 cores this soon though I think is BS. Otherwise I have to consider Intels' bigger sockets which last longer typically.

Well, if they can get the clocks up for the desktop, that'll certainly help; the hard part for desktop users has been configuration issues, which seem to have gotten better over time with UEFI updates. They'll still have the base IPC deficit, and power consumption is an unknown and could be a limiting factor for some, but they'll be closer.

Going beyond the consumer sockets, though, is the need to get their package latencies down. This limits some (not all!) compute workloads.

They should be able to deal with both, hopefully, with the next spin.

 

[Pieter3dnow]

I would be very cautious with any information regarding performance from the 12nm finfet process at Global foundries as fact. The next Zen supposedly is Zen+ not Zen 2 that one is slated for 2019 on 7nm.
AdoredTV had has some speculation on the new process from GF both 12nm and 7nm.

 

[sirmonkey1985]

i could see 5.1 boost clock potentially happening if they're actually able to get the base all core clock speeds that high, i just don't know if you'll be able to get 5.1Ghz OC across all cores though. either way if the clocks on zen+ end up in those ranges and there's still room to overclock i'll definitely replace my R5 1600.

 

[Master_shake_]

with a base clock of 4.6 it will blow by anything intel has to offer at this point.

 

[N4CR]

I would be very cautious with any information regarding performance from the 12nm finfet process at Global foundries as fact. The next Zen supposedly is Zen+ not Zen 2 that one is slated for 2019 on 7nm.
AdoredTV had has some speculation on the new process from GF both 12nm and 7nm.

This is in line with what I had learned also (Zen+), so 'Ryzen 2' seems a bit weird. That said the lower of the clocks are definitely in line with expectation.
There would be no point if they release a respin with 4.2GHz or something like that, even if they fix teething issues.
Also noticed the next video from that had ASUS bios updates confirming 12nm updated CPUs, so timelines in that slide are probably not far off...

So we have mid 4GHz very likely. 12nm very likely, jan-feb very likely.
All we are not sure on is 12 core (I think unlikely until Zen2 - cores are not the issue, it's GHz/compatibility) and the upper frequency range possible.

10%+ is 4.4GHz with top bins.

i could see 5.1 boost clock potentially happening if they're actually able to get the base all core clock speeds that high, i just don't know if you'll be able to get 5.1Ghz OC across all cores though. either way if the clocks on zen+ end up in those ranges and there's still room to overclock i'll definitely replace my R5 1600.

5.1GHz on single core you mean? Perhaps, maybe.. just can't find enough information on their new process.

 

[gigaxtreme1]

They could call it Ryzen 2, It's just semantics for product differentiation and not Zen 2.

 

[sirmonkey1985]

with a base clock of 4.6 it will blow by anything intel has to offer at this point.

on the heavy multithreaded stuff definitely but on the lighter use stuff it's hard to say given how aggressive intel's boost clocks are between 1 and 4 threads. we'll see if they had to give up anything to get the higher clocks or if the process it's self fixed the problems. it'll also be interesting to see if AMD implements the boost intelligence feature into zen+ that is used in Raven Ridge to compete better with intel's multistep boosting depending on whether you're using 1-4 threads.

They could call it Ryzen 2, It's just semantics for product differentiation and not Zen 2.

zen is the name of the architecture while Ryzen is the brand name for the processor series.

 

[N4CR]

They could call it Ryzen 2, It's just semantics for product differentiation and not Zen 2.

Great point. AMD also has called 7nm Zen+ and Zen2, like Zen/Ryzen before launch we won't be sure until this 12nm release naming is official.

on the heavy multithreaded stuff definitely but on the lighter use stuff it's hard to say given how aggressive intel's boost clocks are. we'll see if they had to give up anything to get the higher clocks or if the process it's self fixed the problems.

Would certainly reduce the lead Intel currently has in single thread and some gaming loads, to the point where it would likely be ~10% here or there, then it becomes (for most) more a decision based on price and if you need multithread, e.g. all round workstation + gaming. Of course if Intel drops 8 core mainstream then this becomes a closer race. Either way we win and I'm excited to see what comes in 2018 and 2019.

 

[juanrga]

with a base clock of 4.6 it will blow by anything intel has to offer at this point.

That is pretty sure single-core turbo... even assuming the slide is legit.

 

[juanrga]

By eleven time...

Zen+ was renamed to Zen2. Zen2 is coming in 2019. There is no more Zen+.

Next year is Zen. Pinnacle Ridge uses Zen. Pinnacle Ridge uses a 14nm+ node, which has been relabeled as '12nm'. Last known real roadmap showed Pinnacle Ridge has up-to 8-cores.

1800X is already in the TDP limit of the AM4 socket. Those 95W are a marketing label, real TDP is 128W.

12LP is 14LPP+ (aka an improved 14LPP). I can expect slightly higher clocks for 12LP, maybe 10% higher clocks at isopower. But expecting 10% higher clocks plus 50% moar cores on the same power slot? Very difficult to believe.

 

[Nightfire]

Yes, we know 12 LP is just 14 or whatever. You must hunt down every thread that mentions 12 LP

Either way, 10% will be huge for gaming purposes.

 

[Master_shake_]

That is pretty sure single-core turbo... even assuming the slide is legit.

The chart if legit looks like it says 5.1 turbo.

 

[JustReason]

By eleven time...

Zen+ was renamed to Zen2. Zen2 is coming in 2019. There is no more Zen+.

Next year is Zen. Pinnacle Ridge uses Zen. Pinnacle Ridge uses a 14nm+ node, which has been relabeled as '12nm'. Last known real roadmap showed Pinnacle Ridge has up-to 8-cores.

1800X is already in the TDP limit of the AM4 socket. Those 95W are a marketing label, real TDP is 128W.

12LP is 14LPP+ (aka an improved 14LPP). I can expect slightly higher clocks for 12LP, maybe 10% higher clocks at isopower. But expecting 10% higher clocks plus 50% moar cores on the same power slot? Very difficult to believe.

Where exactly are you getting this TDP "limit" for AM4 socket? I am right now, while running BOINC 16 threads for HardOCP, pulling 170W avg over the last 14 hours. By your statement, something should be on fire or exploded by now right?

 

[Nightfire]

Lol, I managed to hit 240 watts at the wall (SFX Gold) using an overclocked A12-9800! By my estimates, the APU was drawing about 200 watts all by itself. So far the, B-350 TUF is doing just fine.

 

[tangoseal]

with a base clock of 4.6 it will blow by anything intel has to offer at this point.

Im happy with my 7820x and 8600k unless the new Zen smashes them then I will be swapping them out yet freaking again. But this time I am going to hold off to see what INtel has snaking out of their pant legs once AMD drops a kill or be killed next chip.

 

[Nightfire]

$330 for a 2700? Damn, I like that!

The highest binned 8 core Ryzens (1900x) have legitimatley hit 4.3 ghz already:

Check out the CB run at 4:01
Every 1900x review I have seen (as few as they are) overclock to at least 4.1 GHz

The next gen chips should be even higher than that. Consistant 4.5 GHZ would be great.
Even with no IPC gains, that should be very near 2000 CB points.

I was goint to wait for Raven Ridge, but with Pinnacle Ridge so close now, I will probably get a 2700 or 2600 with the cheapest GPU I can find on Craigslist.

 

[mnewxcv]

I just hope they can bring ryzen 5 1400 performance to the mobile platform.... Intels high performance mobile offering cost a metric fuck ton. It's definitely an untapped market.

 

[Araxie]

Where exactly are you getting this TDP "limit" for AM4 socket? I am right now, while running BOINC 16 threads for HardOCP, pulling 170W avg over the last 14 hours. By your statement, something should be on fire or exploded by now right?

I guess he it's referring to the Stock TDP which it's the marketing label of 95W.

 

[juanrga]

Yes, we know 12 LP is just 14 or whatever. You must hunt down every thread that mentions 12 LP

Either way, 10% will be huge for gaming purposes.

I only wrote a remind that 12LP cannot provide 50% moar cores and a whole ~GHz in the same thermal slot.

10% will be 10% for everything, from games to Blender.

Where exactly are you getting this TDP "limit" for AM4 socket? I am right now, while running BOINC 16 threads for HardOCP, pulling 170W avg over the last 14 hours. By your statement, something should be on fire or exploded by now right?

I am getting it from AMD technical docs. It is the limit defined by the spec for the AM4 platform. It is not the limit that can get individual mobos. Motherboard makers can design mobos for overclocking. The limits of the spec are surpassed when overclocking . Not a mystery here.

Adding 50% moar cores and ~1GHz would increase the limit to around 350W. There is no way in this Universe that 12LP can reduce this number by a factor of ~3x to fit it within the AM4 spec. So it is crystal clear that the slide is fake.

 

[N4CR]

As I said, 12 cores doesn't seem possible, not only for reasons juanrga gave but also the fact that they wouldn't have time to redesign to that level, for what is basically a re-spin and tweak.
The base clocks I'd say are more than possible judging by official sources, so don't throw the baby out with the bath water. This could be one of those leak tests e.g. partially accurate but with different numbers passed out to each potential leaker.

We'll see 12 cores on 7nm likely early '19 at this point.

 

[IdiotInCharge]

Well, 12 cores shoudl be possible, but stock clocks would be buried, no different than the high core count Intel HEDT parts.

Boost clocks for a few cores could also be improved, though...

 

[KazeoHin]

I don't believe this, as much as I want to.

We know that memory channels are not tied to CCXs, so a 3-CCX part could be made with 12 cores and two memory channels, but I doubt it would have high all-core turbo speeds. Sure, there's no stopping AMD from allowing it to boost to 4.0+ on one or two cores, but no way they can maintain over 3.0 on all cores under the 95 watt restrictions of the socket. On OC mainboards or just boards with a half decent VRM, you could OC the base line further, but you have to design these chips to work on ALL boards, including the lowest bidder Foxconn OEM trash that Dell and HP use for their desktops.

 

[OrangeKhrush]

I don't believe the core count, from most things said prior seem to point at PR being the same core flavour with everything you expect from a "tock" cycle. I could see higher clocks, tweaked IMC and a few other bits and bobs but a 5ghz 12 core....nah

 

[tangoseal]

Again who cares about TDP except manufacturers? Really is your 95 watt or 65 or 35 watt CPU going to matter to you? Are you designing a laptop from scratch? Getting millions in investment and trying to battle the likes of apple? Or are you just gonna build a desktop and overclock the shit out of it anyways and give two whopping pisses less about TDP ...

man the TDP concerns of folks here is redunkulous

Literally the only time I take any concern with TDP is if I were trying to save every watt for a high power savings 24.7 ITX server NAS like thingy. Other than that let the watts be watts.

 

[mnewxcv]

Again who cares about TDP except manufacturers? Really is your 95 watt or 65 or 35 watt CPU going to matter to you? Are you designing a laptop from scratch? Getting millions in investment and trying to battle the likes of apple? Or are you just gonna build a desktop and overclock the shit out of it anyways and give two whopping pisses less about TDP ...

man the TDP concerns of folks here is redunkulous

Literally the only time I take any concern with TDP is if I were trying to save every watt for a high power savings 24.7 ITX server NAS like thingy. Other than that let the watts be watts.

Tdp matters for mining. Which ryzen cpus are pretty good at.

 

[tangoseal]

Tdp matters for mining. Which ryzen cpus are pretty good at.

A niche market, but ok it matters for that obviously when you try to balance power comsumption vs. performance vs. payout but for the 99% of gamers/power users, etc... it should have zero impact on anything you do.

But people really use a cpu to mine anymore? Really? GPUs are like all the rage. I think CPUs are just too slow in hash count to matter anymore from what I understand.

Side note I find it funny how many people are all about Global Warming awareness and saving the whales and the polar bears etc... while burning more carbon running their 500 GPU coin farms in their warehouses. I love the hypocrisy. Just talked to one of those the other day and had to leave before I laughed any more.

 

[mnewxcv]

A niche market, but ok it matters for that obviously when you try to balance power comsumption vs. performance vs. payout but for the 99% of gamers/power users, etc... it should have zero impact on anything you do.

To an extent. Higher Tdp means more expensive cooler and or louder fans. Also means hotter room, depending where you live this is good or bad. My gpu is already a 200w heater, don't need to double that.

That being said, if performance to Tdp scales with current generation, I'll take the performance and extra heat. But last summer I had my rig in my bedroom and don't have air conditioning. Playing games late at night then trying to go to sleep in a hot ass room sucked.

 

[tangoseal]

To an extent. Higher Tdp means more expensive cooler and or louder fans. Also means hotter room, depending where you live this is good or bad. My gpu is already a 200w heater, don't need to double that.

That being said, if performance to Tdp scales with current generation, I'll take the performance and extra heat. But last summer I had my rig in my bedroom and don't have air conditioning. Playing games late at night then trying to go to sleep in a hot ass room sucked.

I agree with your usage scenario. I am referring to a subset of individuals on this forums that go waay to damn far with the whole TDP thing.

Well unless they offer a higher spec, new version of PCIe, then TDP MUST remain constant and can't get any higher than the current highest allowable on the market. Otherwise the TDP will be violating the specification of the very bus that is supposed to support it whether it be cpu or add in cards.

Again, not for the customer to worry about, the manufacturers have to heavily weigh TDP. I mean if TDP wasn't a concern to a manufacturer then nVidia/AMD/Intel could just slam a chip out with a 5 kw TDP and it would do more shit than you could ever dream of in a picosecond but due to limitations of physics, power supplies, regulations, safety, and realistic approaches TDP will always remain a factor reserved for the concern of the manufacturer. As a consumer we should just consider it's impact in our individual scenario.

It just gets old hearing people always complaining about TDP this or that and literally missing out on awesome performing platforms because they truly do not understand to what segment TDP numbers are even published for, the manufacturers.

 

[IdiotInCharge]

It just gets old hearing people always complaining about TDP this or that and literally missing out on awesome performing platforms because they truly do not understand to what segment TDP numbers are even published for, the manufacturers.

TDP is still useful for bracketing power usage and heat output, but here, we're looking at a part whose required TDP would prevent it from being used with many current motherboards.

Of course that point is moot now, given that the rumor is bunk, but if AMD were to do something like this without a massive increase in efficiency we'd be looking at another 8700k quandry in terms of motherboard support.

 

[N4CR]

Really, TDP only matters when it's AMD.
No one praised Ryzen for being far more efficient than its' competitor. Ryzen 1700x/1800x is almost 1/3rd of the power of a 7900x and only 25% less performance.. max OC to Max OC or stock to stock, regardless it makes the 7900x look like a power sucking chip.

Zen when run before it's 2nd critical (~3.6GHz) is extremely efficient. I'm interested to see how the LP+? '12nm' revision does in this regard.

 

[IdiotInCharge]

Really, TDP only matters when it's AMD.
No one praised Ryzen for being far more efficient than its' competitor. Ryzen 1700x/1800x is almost 1/3rd of the power of a 7900x and only 25% less performance.. max OC to Max OC or stock to stock, regardless it makes the 7900x look like a power sucking chip.

Zen when run before it's 2nd critical (~3.6GHz) is extremely efficient. I'm interested to see how the LP+? '12nm' revision does in this regard.

You're going to complain about bias and then compare an HEDT part to a consumer part?

You're the pot calling the kettle black here.

 

[JustReason]

You're going to complain about bias and then compare an HEDT part to a consumer part?

You're the pot calling the kettle black here.

If Intel had a 8 core consumer part then he could have used it for reference. But since Intel has been raping its customers and holding back the industry as far as consumer access is concerned then HEDT vs consumer is all you got.

 

[IdiotInCharge]

If Intel had a 8 core consumer part then he could have used it for reference. But since Intel has been raping its customers and holding back the industry as far as consumer access is concerned then HEDT vs consumer is all you got.

So an 8700k that's as fast or faster than any R7 doesn't count, because the Ryzen CPUs have more cores, just slower?

Not buying it, sorry.

 

[juanrga]

Really, TDP only matters when it's AMD.
No one praised Ryzen for being far more efficient than its' competitor. Ryzen 1700x/1800x is almost 1/3rd of the power of a 7900x and only 25% less performance.. max OC to Max OC or stock to stock, regardless it makes the 7900x look like a power sucking chip.

Zen when run before it's 2nd critical (~3.6GHz) is extremely efficient. I'm interested to see how the LP+? '12nm' revision does in this regard.

TDP did matter a lot when certain people believed the marketing "95W" was real. I recall correcting then the wrong claims made by chiparchitect, when he took that 95W marketing value to pretend that RyZen 1800X was more efficient than 140W Broadwell-E. In reality RyZen is less efficient than Broadwell-E.

No one praised RyZen for being more efficient than "its competitor", because it is not more efficient. The 7900x is more efficient than the 1800X and ties with the 1700X, contrary to your claims,