China mobo manu leaks | 15% ipc gain zen 2

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
9,743
From Wccftech they report:

Zen 2 arch.
Motherboard manufacturers leak.

15% IPC Gain over Zen +
Samples arriving at mobo firms w/ 4.5ghz

Improved memory controller

Core counts unknown

Boost or all core unk.

Looks like it might be the next Intel stomper? All speculation. Its Wcc so.... salt is required.

Tried very hard not to be click baity since 1. I make zero money with you clicking this and 2. Because someone on this site will bitch about it being click bait. Never damn fails.

Link: https://wccftech.com/amd-ryzen-3000-15-percent-ipc-uplift-4-5-ghz-x570-pcie-gen-4-40-lanes/
 
My 3770k is also getting a bit long in the tooth. I know there are already a ton of options that will blow it away but it's providing adequate enough performance to hold out for Zen 2
 
For me, the jump from Phenom II 955 (turned into beyond 980 spec if not near top of what most Ph II x4 I have seen charts on, "maxed out non exotic cooled"

to Ryzen 2600x is "estimated" ~230% "improvement" on base clock alone, jump to 7nm and/or 7nm+ EUV towards 2020 will be another nice sizeable chunk ahead.

my poor little pheny II needs to go to bed, it did its job LOL......

They have been "tight lipped" about Ryzen before launches (for damn good reason) but Dr Lisa has been very straight to the nitty gritty about how how "the" company is doing overall, if they are meeting ramps, yields, drivers on point and so forth.

Last I heard was with the current 7nm TSMC offers (for high volume production, is performing above base spec requirements) and can go for something like 25% more transistors at 5% less power or something like that
AMD does a "round average" this way here they are not being "it WILL BE" like Intel and Nv have done time and time again last few years and not always meet with "ahem 14nm++++ with 10nm "ready for the limelight before Q2 at latest Q3 this year"

going to be very interesting indeed
 
Did someone finally kill juanrga or what?
He used to eat these threads for breakfast.

Edit: Checked his twitter and seems he got a girlfriend and became a politcal activist/troll.
Good for him(us).
 
Did someone finally kill juanrga or what?
He used to eat these threads for breakfast.

Edit: Checked his twitter and seems he got a girlfriend and became a politcal activist/troll.
Good for him(us).

He didnt eat nothing. He just turbo trolled while on trucker speed.
 
Did someone finally kill juanrga or what?
He used to eat these threads for breakfast.

Edit: Checked his twitter and seems he got a girlfriend and became a politcal activist/troll.
Good for him(us).
Lol yeah he probably got pussy whipped into supporting her political ideology of choice. He was pretty much 'paid level thread derailing' so doesn't surprise me to see him go with the sheckels for another job. He gave up last year after lots of us called him out for it. He'd just come in, thread crap and drag it off topic if it was anything positive for AMD, or steer the conversation to Intel (in an AMD forum lol). I bet he now works for shareblue or some other pathetic political disruption/shilling/trolling/'organic discourse' company on facebook/forums/cuckchan etc.
We have a new version of him, that trans guy/girl/it whatever that tries to make every thread about sexuality, patriarchy, fucking white males and women/gays dindu nuffin lmao. Never blocked anyone here but they're getting awfully close.
And there are others who are a watered down juan, always a few shareholders floating around here too. Usually they are easy to spot though because when AMD does something good they are not happy and usually know nothing about what they are crapping on.

Anyyyway back OT..
This is inline with expected IPC improvements. Upper end though, i was expecting ~10-12% on most workloads if all goes well. 15% might be only specific workloads if true.
4.5GHz must be all core. I would expect higher in single-dual core under PBO3 or whatever they'll have by then because we have current 2700x already reaching around there.
 
  • Like
Reactions: c3k
like this
15% IPC plus clock speed improvement, nice. 40 pcie lanes? That would be very nice and would make TR less needed overall. 16core + 40pcie lanes all version 4 (y)

Well if the X570 is that much of an improvement - sign me up
 
15% IPC plus clock speed improvement, nice. 40 pcie lanes? That would be very nice and would make TR less needed overall. 16core + 40pcie lanes all version 4 (y)

Well if the X570 is that much of an improvement - sign me up

I wonder if they will have or offer a trade up program (lol pipe dream I know) for current Threadripper owners for the next gen. Its kind of a stab to the face in a way that we spent $1000 on our 2950x's that in just a year or less they come out with a desktop part that is going to have 16 faster cores @ 7nm.

I guess that is the risk you have with technology.

I might ditch threadripper if the new Zen 2 is significantly faster in multithread workloads even at dual channel memory.

noko I could be wrong but it seems that Intel is already shitting coal in a panic with this press release from a few days ago

https://www.digitaltrends.com/computing/intel-promises-10nm-2019/
 
I can't see AMD cannibalizing Threadripper sales for Ryzen3.

I wonder if the 40 PCI-E lanes is really 20 PCI-E 4.0 lanes that have the bandwidth of 40 PCI-E 3.0 lanes?

That would still be double the I/O bandwidth of Ryzen2 but at the same time be less desirable than a Threadripper2 (64 PCI-E 3.0 lanes) let alone a PCI-E 4.0 Threadripper3. It is also simpler for the sockets because the Ryzen's already have 20 lanes (although it is entirely possible AMD thought of that and left pins available in the original AM4 spec socket to add more PCI-E lanes).

Ryzen3 will probably still be dual-channel RAM so it would be another way to keep costs down (CPU and motherboard) and unlikely to challenge Threadripper2/3. As a bonus it would be less likely to alienate early Threadripper adopters (who probably already feel burned by the Threadripper2).

Maybe AMD will be pulling an Intel Xeon with the PCI-E lanes, where different CPUs have different number of lanes even though they all fit in the same socket? If AMD did that it would allow AMD to create a middle enthusiast level (below Threadripper but above the Ryzen2) and give buyers more options. On the other hand some enthusiasts are unhappy that lower Xeons are crippled at the PCI-E lanes level, such a tactic by AMD may alienate the very buyers they are trying to attract. Also it may make more sense for AMD to keep the lines simple as it reduces the number of SKUs (especially motherboard chipsets).

Don't get me wrong, I want a big 16-core Ryzen3 with lots of I/O as I'm planning on replacing my home hypervisor and I would prefer to avoid the premiums you pay for Threadripper. I just think AMD is smart enough to not bite the hand that feeds them now that they have a competitive architecture again - how many Threadripper buyers are sitting on their hands right now waiting for Ryzen3 to drop to see if a platform a third of the price of Threadripper would do the job? If AMD doesn't give Threadripper buyers a real reason to buy Threadrippers they are not going to sell a lot of them.
 
15% IPC plus clock speed improvement, nice. 40 pcie lanes? That would be very nice and would make TR less needed overall. 16core + 40pcie lanes all version 4 (y)

Well if the X570 is that much of an improvement - sign me up

I find these numbers always weird because in the end what do you do with your computer? We saw the 29% number when dealing with financial programs.

How many times have we seen something approach 10% gains over last generation at same frequency (not the compiler optimized crap you see all the time). That makes me wonder if that 7nm process is that great to carry such a large amount of the improvements even if they say it is working up to 4.5ghz?

With DDR5 (supposedly much faster) being introduced next year (if i'm not mistaken) the x570 even with all the PCI E lanes seems a bit to much.
 
I find these numbers always weird because in the end what do you do with your computer? We saw the 29% number when dealing with financial programs.

How many times have we seen something approach 10% gains over last generation at same frequency (not the compiler optimized crap you see all the time). That makes me wonder if that 7nm process is that great to carry such a large amount of the improvements even if they say it is working up to 4.5ghz?

With DDR5 (supposedly much faster) being introduced next year (if i'm not mistaken) the x570 even with all the PCI E lanes seems a bit to much.

In the past you seen 32nm to 28 to 22 to 18 to 14+++++++++++++++++++++ to 12 and now to 7nm. The biggest shrink possibly ever in modern computing may very well be 12/14 to 7nm. That's a 50% reduction in process size in one iteration from the last.

Is there not be any expected benefit to dropping the node size literally in half?
 
In the past you seen 32nm to 28 to 22 to 18 to 14+++++++++++++++++++++ to 12 and now to 7nm. The biggest shrink possibly ever in modern computing may very well be 12/14 to 7nm. That's a 50% reduction in process size in one iteration from the last.

Is there not be any expected benefit to dropping the node size literally in half?
That is correct in a way but never ever have I seen AMD being able to capitalize on it so well (that is if the 15% holds water). There is a slide with the exact numbers for the improvement of the 7nm (vs 14nm) process I think it was 1.35 (performance)
Have you seen anything like this from Intel ?
 
Last edited:
In the past you seen 32nm to 28 to 22 to 18 to 14+++++++++++++++++++++ to 12 and now to 7nm. The biggest shrink possibly ever in modern computing may very well be 12/14 to 7nm. That's a 50% reduction in process size in one iteration from the last.

Is there not be any expected benefit to dropping the node size literally in half?

Mandatory juanrga reminder that intels non-existing 10nm is superior to tsmc's 7nm and that 7nm is a lie.
 
This next Ryzen might be my move from Intel to AMD.

8700k is out because it doesn't make sense, 9900k is out because it's already priced stupid and I've gotta work in a waterblock.

My 7700k is a good 4-core with HT but that configuration is making it feel pretty dated lately. XD
 
This next Ryzen might be my move from Intel to AMD.

8700k is out because it doesn't make sense, 9900k is out because it's already priced stupid and I've gotta work in a waterblock.

My 7700k is a good 4-core with HT but that configuration is making it feel pretty dated lately. XD

I have a hunch that the 5.1ghz model of Ryzen 2 is going to faster than the 9900k if not by a teeny bit, possibly, a large portion, in single threaded. But well see in couple months time.
 
I have a hunch that the 5.1ghz model of Ryzen 2 is going to faster than the 9900k if not by a teeny bit, possibly, a large portion, in single threaded. But well see in couple months time.
Honestly, even if the next Ryzen is a little slower than Intel in single core, I'll probably still bite. If you want eight cores and sixteen threads from Intel they want your kidney, and then you have to direct a firehose at the IHS to keep the thing cool.
 

Hell why not? Intel is still at 14++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Amd is shrinking to 7nm and is promising a 15% IPC gain over previous Zen + and Zen + was 11% slower on average in gaming than the 9900k at 5ghz So you tell me whats possible. Intel can't reign forever.

That 11% number is from HF, Anand, Gamers Nexus, Pauls, youi name it ... it averages 11% slower in games on all those review sites together.
 
Hell why not? Intel is still at 14++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Amd is shrinking to 7nm and is promising a 15% IPC gain over previous Zen + and Zen + was 11% slower on average in gaming than the 9900k at 5ghz So you tell me whats possible. Intel can't reign forever.

That 11% number is from HF, Anand, Gamers Nexus, Pauls, youi name it ... it averages 11% slower in games on all those review sites together.
I was running AMD a long time ago and with every new announcement it was always the same hype how this new gen was going to beat Intel....still waiting and I bet is NOT happening on this one either
 
Zen + was 11% slower on average in gaming than the 9900k at 5ghz So you tell me whats possible. Intel can't reign forever.

That 11% number is from HF, Anand, Gamers Nexus, Pauls, youi name it ... it averages 11% slower in games on all those review sites together.

Which includes GPU limited tests, a game that is 90% GPU limited can make a CPU that is 50% faster perform only 5-10% better in a average FPS test.

Even if it is still slower it may be fast enough that it performs well enough for most to get above the required FPS in current games and at high res in most games this will make the GPU the bottlneck so that there is no visible difference between the fastest available CPU (whatever that may be) and one that is slower.

I believe AMD can beat Intel in single threaded performance in a program like Cinbench which doesn't care about memory latency or perhaps in games when sticking to Intels stock memory frequency of 2666 with a c16 kit.
But with both overclocked using high speed RAM kits unless AMD has made some massive improvements to the memory controller and it's latency it will probably still fall behind in any software that is badly threaded and sensitive to memory latency like games, photoshop, lightroom.

The main advantage Ryzen 2700X had over 1800X in games was it's 7.6% reduction in memory latency but at best AMD is still ~60ns vs ~37ns for Intel when both have overclocked b die.
So even if the core is faster than Intel if it spends time idle while it is waiting on data from the RAM to be fed into the cache it will end up slower.

This is why I was hoping they would use that gap on the CPU for a L4 Cache if they could use something with half the latency of the RAM and at least 256MB like Intel used on the 5775c then they would have a true competitor to Intel.
 
Last edited:

Let’s be honest with ourselves. The Athlon glory days were a long time ago. Fact is he is right it’s been that way since Phenom. Remember the excuses? It was too advanced for Windows to use properly? Lol I do think Ryzen has a very real chance to be extremely competitive, by that I mean better in some areas and worse in others but there’s a large gap between Athlon 64 and Ryzen where the AMD fans were led to believe they were going to regain the performance crown and it never happened. Most of the times it wasn’t because of AMD but because of the fans themselves
 
AMD is already faster than Intel in several areas.

Wtf are you guys even talking about with all this dour and negative bleh blah?

2700x is only 9% slower clock for clock than the 9900k in avg.

2700x is also capped at around 4.2 ghz. Intel is another 7 to 800 mhz faster.

AMD is already faster per clock when weighed evenly against Intel.

But you guys dont want to see the inconvenient truth.

Edit.. fixed my typos sigh
 
Last edited:
waiting for that 12 core 5Gz stock CPU at a reasonable price. Still running my 6 core, 12 thread aging 2.6Ghz X5660 Xeon. Double the Cores, Double the stock frequency. ( Yes, I know IPC improvements have left my baby in the dust) Should make for an upgrade I can "feel" when I finally get around to it. Until then, My little "Beast" keeps plugging away. Drawing all the power in the world.
 
AMD is already faster than Intel in several areas.

Wtf are you guys even talking about with all this dour and negative bleh blah?

2700x is only 9% slower clock for clock than the 9900k in avg.

2600x is also capped at around 4.2 ghz. Intel is another 7 to 80p mhz faster.

AMD is already faster per clock when weighed evenly against Intel.

But you guys dont want to see the inconvenient truth.
Even if Intel retains it's single-core speed advantage (which I'm not convinced it will,) the 3700X vs the 9900K makes the 9900K look like the huge ripoff that it is.

AMD showed one of their 8c/16t products narrowly beating a 9900K recently. Even if that's a little skewed, even if the 3700X is a tad slower in single thread, it's still a 12c/24t processor with the same clock speed, 150% the cores and threads, way lower TDP and a 7nm process, at 60-70% the price if speculation is accurate. Intel's about to get a kick in the teeth.
 
AMD is already faster than Intel in several areas.

Wtf are you guys even talking about with all this dour and negative bleh blah?

2700x is only 9% slower clock for clock than the 9900k in avg.

2700x is also capped at around 4.2 ghz. Intel is another 7 to 800 mhz faster.

AMD is already faster per clock when weighed evenly against Intel.

But you guys dont want to see the inconvenient truth.

Edit.. fixed my typos sigh

What does “faster per clock when weighed evenly” mean?
 
What does “faster per clock when weighed evenly” mean?

If you down clock the Intels aggressive 5.x ghz clock to AMDs peak 4.x ghz then suddenly you see the AMD is actually faster per cycle in IPC across the board than the Intel. The only advantage really Intel has at this point in the race is gross clock speed.

Possibly thier ringbus but small tweaks to infinity on AMDs part is rapidly closing that gap.

Lastly I too own Intel products as well so I'm not being lopsided here.
 
Last edited:
If you down clock the Intels aggressive 5.x ghz clock to AMDs peak 4.x ghz then suddenly you see the AMD is actually faster per cycle in IPC across the board than the Intel. The only advantage really Intel has at this point in the race is gross clock speed.

Possibly thier ringbus but small tweaks to infinity on AMDs part is rapidly closing that gap.

Lastly I too own Intel products as well so I'm not being lopsided here.

Nor am I since I’m holding out for Zen2 to finally retire my 3770k. Well not retire but to be demoted to my 2nd gaming machine.
 
If you down clock the Intels aggressive 5.x ghz clock to AMDs peak 4.x ghz then suddenly you see the AMD is actually faster per cycle in IPC across the board than the Intel. The only advantage really Intel has at this point in the race is gross clock speed.

Possibly thier ringbus but small tweaks to infinity on AMDs part is rapidly closing that gap.

Lastly I too own Intel products as well so I'm not being lopsided here.

Also if you cap both the 2700x and 9900k to their 105W and 95W spec'd TDP without extended turbos or anything like that. They perform nearly identically across the board.
 
Nor am I since I’m holding out for Zen2 to finally retire my 3770k. Well not retire but to be demoted to my 2nd gaming machine.

I wont blindly support AMD with fervor though.

I/we hope Zen 2 surpasses Intel. Not because of anything more than true competition which is good for we consumers.
 
I wont blindly support AMD with fervor though.

I/we hope Zen 2 surpasses Intel. Not because of anything more than true competition which is good for we consumers.
Amen. I hope AMD wipes the floor with Intel in 2019.

Then I hope Intel retakes their lead.

Then AMD again.

Rinse and repeat... We wind up with better processors for the competition.
 
Let’s be honest with ourselves. The Athlon glory days were a long time ago. Fact is he is right it’s been that way since Phenom. Remember the excuses? It was too advanced for Windows to use properly? Lol I do think Ryzen has a very real chance to be extremely competitive, by that I mean better in some areas and worse in others but there’s a large gap between Athlon 64 and Ryzen where the AMD fans were led to believe they were going to regain the performance crown and it never happened. Most of the times it wasn’t because of AMD but because of the fans themselves

I'll admit that it seems more recent than it was, at least in my head, but yes, it's been 10 years since they were on top. I was disagreeing with the assertion, that I've seen from more than one person, that AMD has never been, and will never be, competitive.

The younger crowd might only remember a time where AMD was beat hard by Intel, but they've been on top before, for years, and have the potential to be again.
 
If you down clock the Intels aggressive 5.x ghz clock to AMDs peak 4.x ghz then suddenly you see the AMD is actually faster per cycle in IPC across the board than the Intel. The only advantage really Intel has at this point in the race is gross clock speed.
Maybe in cinbench but Intel has a big advantage in games due to there sensitivity to memory latency unless your comparing Intel with 2666c16 vs AMD 2933c14.

@4GHz with 3200c14 RAM Intel is up to 10-15% faster in these tests but there is no way of knowing how GPU limited they are.
https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/page3.html

I haven't used AMD since Opteron 170 and before that a Duron 650 upgraded to Athlon XP2000.
But I am looking forward to pairing A new Ryzen CPU with my Radeon VII and seeing how it compares to my 6700K with some heavily overclocked b die.
 
Last edited:
Back
Top