OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
No, just left it stock with a 2600 Kit, doesn't need much more help, its a bit of a pocket rocket
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Zen 2 IPC, Clocks & Gaming Performance Improvements
- Thread starter erek
- Start date
No, just left it stock with a 2600 Kit, doesn't need much more help, its a bit of a pocket rocket
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
The only BS is you and your ignorance. Target was always 6-8Ghz (touted by GloFo as a possibility BEFORE AMD even started work on BD) hence the relaxed front end and again the cache and memory control.
And you need links to quotes as I am sure it is well after the design was in work. (yep a review which is well after AMD would have known what the real clocks would be aka: engineering sample).
And I don't give a flying FK about your argument in lala land elsewhere over 5.7Ghz. You only mention it here as if we are all supposed give you accolades and well that sure as hell aint gonna happen. Just because you manage to find some moron else where gives you no cred here. That ship has sailed.
Edit: I will add a good deal of my Knowledge on BD is from the STILT and I will take his word over yours 1E934286 of the time
Last edited:
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
so much love for Juan, he needs a buddy, I will invite him to my christmas party.
https://hardforum.com/threads/zen-2...ormance-improvements.1943227/#post-1043202989
And vague mention to The Stilt, doesn't change the fact that your 6--8GHz continues being unadultered BS.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
just because you can't understand how target is associated in the chain doesn't change the facts. I KNOW BD and all the tidbits of info leading up to its release and why the design path was chosen. Just because your knowledge is limited to crap information doesn't change the history.
how about a quiz.
What is the max safe voltage on BD? ( hint: trick question)
Why was the architecture design chosen? (more cores is not a correct answer.)
those 2 require knowledge and really can't be looked up. Of course I don't expect a response as neither of those really have a negative, though number 2 you could probably spin.
Your question is not tricky, it is completely useless, first because the relation between working voltage and working frequency is not a constant; second because one can have different f_max for the same voltage/node.
I know why the microarchitecture was chosen, who was the main engineer, how it is based in a former (cancelled) prototype from another engineer, whom I also know, and even know the original prototype was initially developed outside of AMD...
Your 6--8GHz continues being BS. Why you insist after AMD engineers confirmed the target was just "30% increase in clock speed over Phenom II" is beyond reason.
Last edited:
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
This is my last response to you in regard to this.
It is obvious you know nothing. You cant answer any question straight, always skirting it and trying to turn it around and make it about something else. Why? Because you don't know. But here I will give you the short answers. Why? Because I KNOW them.
The max Voltage was never stipulated, even when asked they skirted the answer, AMD. Maybe they didn't know. At any rate the consensus is... there is no real max, at least reasonably, obviously 3V would fry them. However the max is temp. Many of us ran our chips at >1.6V and reaching 1.75V FOR YEARS and for the most part still running. Mine isn't right now because I have its replacement 1800X. The limitation is temp, so If you could keep temps below 65C, preferably below 60C because higher clocks lost stability above 60C. These are proven facts and KNOWN BY ALL WHO CLAIM TO HAVE FULL KNOWLEDGE OF THE BD ERA.
The architecture was chosen back leading up to 2009. Akin to the GPU AMD had a vision which was based on the direction and early software in development that was to incorporate multithreaded applications, obviously Gaming as well. A DX12-like API was in early development but suddenly ended with no reason. Unfortunately with its end it was too late for any changes in AMDs BD architecture and add to that GloFo 6-8Ghz possibility, AMD soldiered on. Obviously the outcome fell short of the original expectation from clock to Software threading. Fact of life and not everything turns out how we may wish it.
It is sad that an individual will try so hard to pretend to be so knowledgeable even in an area they have an overt disdain for. I refuse to put anyone on ignore as watching the lions share of you prove yourselves ignorant and obtuse is rather laughable. But this is the end of this discussion. You have proven yourself ignorant of anything practical when it comes to AMD and it use in real world. All the advice I have is for others and the hope they too see thru your hollow and baseless lies guised as facts.
As most gentlemen would say: "Good day to you sir, I said GOOD DAY".
Gideon
2[H]4U
- Joined
- Apr 13, 2006
- Messages
- 3,558
The only one in denial is you bud. You have no clue what 14nm+ is yet you act like you know when you quote me... Instead you went to a few news sites and saw they were calling it the next Zen and you went with that, when in fact AMD has said nothing about it. The next version of Zen is Zen2 and will be 7nm and that is the only thing AMD has said period. 14nm+ on that one chart is all you have for a reason and it was the only thing they showed with 14nm+. Pinnacle Ridge is the 7nm Zen2 and is a 2018 product, if you notice on that chart it shows it launching at the tail end of 2018 and all of 2019. 2020 is Zen3 and likely a different motherboard and socket will be used for it. For once you should just admit you have no clue what your talking about as you seem confused why your getting ripped apart on here by facts and by several different members.
gigaxtreme1
2[H]4U
- Joined
- Oct 1, 2002
- Messages
- 3,577
14nm+ might be the FX 14nm process Rambus and Glofo are fabricating the HBM2 PHY on. It is the most recent press release on the Glofo site.
Gideon
2[H]4U
- Joined
- Apr 13, 2006
- Messages
- 3,558
That may be possible as well, just AMD has not clarified what it means. Anything is possible as AMD has been unusually tight lipped with their plans and road maps. Heck even I am not certain on the Pinnacle Ridge being the 7nm Zen2 but that is from rumors I have heard. Either way I am certain I will have a ZEN 2 in my hands before Intel can lunch a 10nm desktop processor.
CaptNumbNutz
Fully [H]
- Joined
- Apr 11, 2007
- Messages
- 24,990
I think I got cancer reading this thread and the bickering.
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
I doubt PR is on 7nm Zen2, I am sure I read AMD confirming it is 14nm+ and due Q1 2018. I am hoping PR sees say a 200-300mhz headroom gain with some tweaks to IMC performance maybe in the tune of 10-20% performance uplift, if that adds some efficiency great if not no major train smash.
I think Zen2 would be a legitimate IPC jump for AMD and significant performance gain while PR strikes me more as an AMD tock.
This new round of personal attacks and ruminations doesn't change the fact that AMD engineers confirmed that the target wasn't that 8GHz that you insist on repeating, neither your posts can negate that Glofo node was aimed at the 5GHz range.
Also "Software threading" wasn't the cause of BD fiasco. Typical server applications are well-threaded and can use lots of cores/threads and it was in the server space where BD failed more.
Last edited:
I know that "Zen in 14nm+" (aka Pinnacle Ridge) is not "Zen2 in 7nm".
The sequence is
1) Zen on 14nm (2017)
2) Zen on 14nm+ (2018)
3) Zen2 on 7nm (2019)
4) Zen3 on 7nm+ (2020?)
Now continue saying I am "in denial", whereas you are the only one negating the existence of Zen on 14nm+
Last edited:
What about 14nm+ being a marketing name for a more mature and tweaked 14LPP?
It makes no sense that AMD ports Zen design to a new process node. AMD doesn't have the money for news masks, validation,... Just consider that the ZP dies used in RyZen and ThreadRipper are dies that failed validation for EPYC.
AMD is reusing the same die from RyZen to EPYC precisely to save costs. It doesn't make any sense to design a separate die now only for Pinnacle Ridge, especially when there is no EPYC version of it.
The transition Summit Ridge --> Pinnacle Ridge is like the transition Trinity--> Richland.
Pinnacle Ride is a Richland-like refresh of Summit Ridge. Same muarch, same everything, but slightly higher clocks (I expect 200--300MHz extra) thanks to a more mature 14LPP process.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
How about you back that up with some link to something somewhere. Again your word has as much value as donkey snot. This isn't the same as before now that we have Infinity Fabric. So past APUs may not be indicative at all of future implementations therefore we can not rely on how they were updated.
Pieter3dnow
Supreme [H]ardness
- Joined
- Jul 29, 2009
- Messages
- 6,784
Don't get what all the fuss is about , it is simple PR person says it will be better , internet goes wild ? If PR person said it sucks donkey balls PR person would be looking for another job ...
Then it remains to be seen where the advantage is coming from improved clock, process or improved IPC/architecture optimization at this point in time rather trivial since you can't buy them yet.
Then it remains to be seen where the advantage is coming from improved clock, process or improved IPC/architecture optimization at this point in time rather trivial since you can't buy them yet.
Why do you mention "Infinity Fabric"? It has zero relevance to my arguments about dies, masks, and process nodes.
The comments on every post I've ever seen there are like cancer. Sometimes there's a few good ones, but it's probably best to be safe and avoid them.
jenosmaverick
Limp Gawd
- Joined
- May 24, 2017
- Messages
- 142
I just ventured here after reading that Asus will finally have a Strix am4 mitx. Was looking for more information on zen+ and zen2 since I'll be having my Dan case on January. Looking in on either a Ryzen or Icelake build. From a casual gamer pov the thread has too much toxicity lol.
Anyway to have my mind clear and have abit more picture on what are the coming cpus this year is there really a zen+ or zen2 and will it be released early next year or late next year around the same time as icelake? I would really love to have either zen+ or icelake over coffeelake. As much as enthusiasts will like coffeelake it just doesn't make sense to me that i'll buy it when icelake zen+/2 is just around the corner after it's release. I'd rather get ryzen 1700 by then.
Anyway to have my mind clear and have abit more picture on what are the coming cpus this year is there really a zen+ or zen2 and will it be released early next year or late next year around the same time as icelake? I would really love to have either zen+ or icelake over coffeelake. As much as enthusiasts will like coffeelake it just doesn't make sense to me that i'll buy it when icelake zen+/2 is just around the corner after it's release. I'd rather get ryzen 1700 by then.
The next CPUs will be Zen+ (see the post by juanrga) aka Pinnacle Ridge. I would think they would launch in March similar to Ryzen launch last year but we will see. So far I haven't seen any items related to this as far as leaked zen+ chips in the wild. If I remember it was around Dec/Jan when the ryzen was leaked at 3.5 max ghz? Someone with a better memory can help out there.
I just explained in my posts that there is no Zen+ anymore because it was renamed to Zen2.
I also explained that Pinnacle Ridge uses Zen cores, and the roadmap I provided clearly says that Pinnacle Ridge uses Zen cores.
So... They changed the name from the 2015 roadmap they put out. In the real world, what does that mean? Also, what is up with you guys and these nebulous 'roadmaps' from years gone by? Admittedly, I'm not the most knowledgeable in when these companies make these 'roadmaps', I bet they revise their roadmaps every year, so why wouldn't you use the most current roadmap to illustrate your points? My bet is that the current roadmap doesn't illustrate your point at all, or I could be wrong, honestly, who cares.
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
true but in reality the roadmaps are never gospel. In my other thread, an article about Threadripper, AMD execs state the roadmaps don't always come to fruition or as planned. They even mention naming conventions don't hold as TR is a prime example.
Renaming Zen+ to Zen2 means you can find lots of confused people in forums believing that "Zen+" is Zen in 14nm+, whereas others believe that what is really coming is Zen2 in 7nm.
AMD is renaming lots of things those days, Zen+ and Zen++ weren't the only projects renamed:
Summit Ridge --> Ryzen
Naples --> EPYC
SP3r2 --> TR4
N4CR
Supreme [H]ardness
- Joined
- Oct 17, 2011
- Messages
- 4,947
This.Roadmaps seems to also be used to misguide the competition with enough vagueness and wiggle room to mean many different things. Intel roadmaps went to hell a long time ago
Only thing we can actually count on is the CEO confirming 7nm tape out for 2H 2017. If you lie to your investors, you get raped in court, so this is the most reliable info we have.
https://hothardware.com/news/amd-confirms-7nm-tape-out-2h-2017-navi-zen-2
2018 mass production is likely. If we'll see that in retail channels Q4 '18 or Q1 '19, is up to how smoothly things go. Considering it's a die shrink on a process proven by IBM (moved to 5nm now), it's likely to not be beyond that time-wise. Definitely would be unlikely to be a 2 year tape out - retail gap like Ryzen.
Remember when the usual suspects were claiming Ryzen would barely break 3Ghz? I'd say AMD has finally got their CPU shit sorted and the cash to do so.
Pieter3dnow
Supreme [H]ardness
- Joined
- Jul 29, 2009
- Messages
- 6,784
I'm just hoping that if it stays on 14nm there still enough benefits, maybe push out new APU on 7nm in 2018 Q1 which should be less of a problem with the process node if things don't work out. And see a refresh of Zen to 7nm if things pan out ...
gigaxtreme1
2[H]4U
- Joined
- Oct 1, 2002
- Messages
- 3,577
It looks like 7nm will first be Vega 20 I assume next year, maybe Q1 according to Digitimes. Pinnacle Ridge APU maybe Q3.
http://www.digitimes.com/news/a20170908PD210.html
Q1 may see Raven Ridge on a tweaked 14nm, FX-14 nm that Global foundries and Rambus are working on for the HBM2 PHY.
https://www.globalfoundries.com/new...rates-25d-high-bandwidth-memory-solution-data
http://www.digitimes.com/news/a20170908PD210.html
Q1 may see Raven Ridge on a tweaked 14nm, FX-14 nm that Global foundries and Rambus are working on for the HBM2 PHY.
https://www.globalfoundries.com/new...rates-25d-high-bandwidth-memory-solution-data
JustReason
razor1 is my Lover
- Joined
- Oct 31, 2015
- Messages
- 2,483
WAIT!!! Why did you go and link to relevant facts, that just isn't how it is done right? lol.
Just kidding. Thanks for the links and good to see you have backup for your claim on that HBM info, I hadn't seen that anywhere else as of yet, and very cool if it works out.
gigaxtreme1
2[H]4U
- Joined
- Oct 1, 2002
- Messages
- 3,577
Interesting that this FX-14nm process is on the news site and has not been added to the Mosis schedule.
https://www.mosis.com/db/pubf/fsched?ORG=GF
I think we can look for changes to the schedule as the process comes out of the experiment stages(FX-14nm= Foundry Experimental?) It would be nice to see what kind of run days the process has as it ramps.
She didn't mention Navi neither Zen2, good luck on the Court. Moreover, wouldn't be the first time she lies to investors. She and Rory Read were sued for "false and misleading statements to investors about the manufacturing and subsequent launch of, as well as the demand for, its Llano microprocessor", and a pair of days ago AMD agreed to paid $29.5 millions to investors to end the litigation.
We are still awaiting for the 14nm Power9 chip. And some IBM systems will not get Power9 until in 2018. But we must believe IBM "proved" 7nm and it is coming in the same year. Also we must consider the historic record at Glofo; until now all their nodes were either delayed or canceled: 10nm, 12nm, 14nm, 20nm, 28nm, 32nm, 45nm. But we must believe that Glofo will be on schedule for 7nm.
I remember knowledgeable people mentioning, before launch, that RyZen would hit a wall at 4GHz, and other people pretending that it could do 5GHz on air with easyness. Imagine who was right and who was wrong.
Your own source claims that
Is this the first hint that Glofo will fail on the 7nm node as well?
gigaxtreme1
2[H]4U
- Joined
- Oct 1, 2002
- Messages
- 3,577
I doubt they are ready for production on that node either. TSMC has a jump on the processes considering their diverse customer base. Neither foundry has complex designs on that process yet. Mostly SRAM.
Gideon
2[H]4U
- Joined
- Apr 13, 2006
- Messages
- 3,558
Cant do any worse then Intel has at 10nm... Were still waiting for that node shrink that never comes....
First 10nm Intel chips have been tested and performance results leaked.
And I can find worse situations without too effort... For instance Glofo 10nm process was coming in 2015
Then Glofo canceled both 14XM and 10XM, and acquired 14nm tech from Samsung.
Latter Glofo announced the development of a new 10nm process, the 10LP, which was canceled past year because it was not competitive, forcing Glofo to go directly with the 7LP node.
https://www.kitguru.net/components/...s-7nm-and-10nm-process-technologies-in-house/
https://www.techpowerup.com/225095/globalfoundries-to-skip-10-nm-and-jump-straight-to-7-nm
Thus Glofo has the honor of canceling 10nm twice: 10XM and 10LP.
7nm process on something as big and complex as a GPU right out the gate? TSMC? No GloFlo? This just does not strike too much confidence in me at all or of the reliability or accuracy of the article. Now it would be nice if 7nm process was ready for prime time and mass production. I just don't see it but what do I know.
Gideon
2[H]4U
- Joined
- Apr 13, 2006
- Messages
- 3,558
First 10nm Intel chips have been tested and performance results leaked.
And I can find worse situations without too effort... For instance Glofo 10nm process was coming in 2015
Then Glofo canceled both 14XM and 10XM, and acquired 14nm tech from Samsung.
Latter Glofo announced the development of a new 10nm process, the 10LP, which was canceled past year because it was not competitive, forcing Glofo to go directly with the 7LP node.
https://www.kitguru.net/components/...s-7nm-and-10nm-process-technologies-in-house/
https://www.techpowerup.com/225095/globalfoundries-to-skip-10-nm-and-jump-straight-to-7-nm
Thus Glofo has the honor of canceling 10nm twice: 10XM and 10LP.
Yep and 10nm from Intel was due in 2015 and yet here we are in 2017 and no 10nm.... Reminds me of a song..