Does that tell us anything whatsoever? Surely the Titan X is the bottleneck at 2160p?
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
*** AMD "Zen" thread (inc AM4/APU discussion) ***
- Thread starter Boomstick777
- Start date
Does that tell us anything whatsoever? Surely the Titan X is the bottleneck at 2160p?
Caporegime
Apparently a fake according to some in the AT forums Ryzen thread - apparently a person with the same username had done the same sort of thing before.
3.6 base 4turbo. Not bad as long as we can overclock to 4700+and costs less than 6900k.
Also someone above said something about 10nm...
AMD goes straight to 7nm end of 2018 for all products. GloFo already setting up the manufacturing process for mass production, ignoring 10nm as waste of time and resources.
yes because rumors talk about intel's 10nm cannonlake to be released Q4 2017, that leaves AMD with about 6months, AMD also said that Ryzen will last 3 years, so no new CPU from AMD untill 2020(7nm) aside from APUs.
Caporegime
Rumours say 2018 for Canon Lake desktop NOT 2017. You might find mobile chips in late 2017 though. People keep falling for the same old delaying tactic rumours - when is the last time Intel actually replaced a major desktop CPU launch within 12 months??
Seriously we had the same stuff spouted IB would replace SB in under 12 months, Haswell would replace IB in under 12 months,etc. Haswell would be replaced in 12 months,etc. It never happened. The only time when something got delayed in Broadwell,they never bothered releasing it in any meaningful way for consumer desktop.
Kaby Lake is a major launch - instead of bringing forward Canon Lake on an immature process node,which is more advantageous in its first run for laptops,it would be cheaper for them to drop prices for a few months.
Intel pushed cutting edge process nodes to mobile first since it is required there. Its been the case for the last few years.
Moreover,I like how you conveniently ignored what has been mentioned in this thread by AMD - they hinted at a one year cadence for uarch updates. They said Ryzen updates would be "tock,tock,tock" and that the uarch would be replaced in 2020. Looking at previous roadmaps we know Ryzen version one was delayed by probably three to six months. They already said mobile Ryzen APUs are out at the end of the year,so it would not surprise me that is the first "tock" for them,followed by the CPUs getting the core updates next year.
The same thing happened on 32NM for AMD. The APUs got PD first followed by the CPUs. The only reason AMD gave up on the FX CPUs,was because 28NM bulk was a flop for a high frequency design like BD/PD - Kaveri took yonks to actually ship at similar clockspeeds to PD APUs on 32NM. By that time the whole line was a dead end,so they re-orientated the design to save power which 28NM was better at.
Intel does the same thing with HEDT being behind consumer CPUs in uarch updates.
Last edited:
Caporegime
https://www.extremetech.com/computi...ils-overclocking-plans-emphasizes-hard-launch
AMD said the uarch would last four years and the core will be updated three more times.
Note the term "core",meaning there is no indication we won't see core updates to both the APU and CPU lines.
Edit!!
Plus there is no mention of AMD NOT making the core on a new process node.
Whether AMD makes Ryzen on a new process node will be down to the availability of that process node within the next four years. If it is,then I can see AMD releasing Ryzen on a new node too.
After all the K10(Phenom) uarch was released on 65NM,45NM and 32NM.
The BD uarch was released on 32NM and 28NM.
I would be very surprised if Ryzen stays on 14NM for four years and that will only happen if GF or Samsung fail miserably in releasing one AMD can use.
AMD said the uarch would last four years and the core will be updated three more times.
Note the term "core",meaning there is no indication we won't see core updates to both the APU and CPU lines.
Edit!!
Plus there is no mention of AMD NOT making the core on a new process node.
Whether AMD makes Ryzen on a new process node will be down to the availability of that process node within the next four years. If it is,then I can see AMD releasing Ryzen on a new node too.
After all the K10(Phenom) uarch was released on 65NM,45NM and 32NM.
The BD uarch was released on 32NM and 28NM.
I would be very surprised if Ryzen stays on 14NM for four years and that will only happen if GF or Samsung fail miserably in releasing one AMD can use.
Last edited:
God I hope not, or a least if they do I hope they make CPU only versions of all their chips available. I mean, there is a reason that Intel's stock 2011 coolers are sold separately.
Soldato
Intel still has issues with their 10nm process stating even on their last paper two weeks ago, they believe will be able for their Coffee Lake end Q4 2017.
Last edited:
Deleted member 66701
D
Deleted member 66701
Ditto - I don't want to pay for something I wont use.
Soldato
They will bring out "new" cpus. They have stated already with Zen+ there will be improvements and there are still a reasonable amount of improvements to make.
Last edited:
yes they will do refresh but no new architecture untill 2020
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Which is the same for Intel. Coffee/Cannon are just small iterations and they are iterations on an old architecture. A brand new architecture will have (95% of the time) significantly more room for fixing and improving areas of the design. I'd expect Zen + to offer a significantly bigger boost to IPC than Cannon or Coffee lake will as a result.
Intel and AMD both have tocks coming but no new architecture till 2020-2021 time frame.
As for Cannon lake on 10nm this year, nope. Coffee lake MIGHT hit this year in Q4, if it does it is solely 45W and below dual cores and I believe not even quad cores. Cannon lake is an additional tock on 14nm which looks set to bring an extra 2 cores to the i7 range. This is pretty likely to come in 2018 rather than late this year. Also before it manages to launch AMD will have both 8 core cpu and quad core APU options available. Now if you want pure CPU, the 8 core is almost certain to be faster than the Cannon lake hex core and adding the extra cores is if anything going to mean hex core Cannon lake clocked slower than the quad core skylake.
If you want APU for the GPU performance, then AMD quad core APU is going to trash Cannonlake hex cores... because if you want GPU power the extra two cores won't do anything for you. AS such Cannon lake doesn't really bring much to the table, it can't match an AMD APU on gpu performance and it can't match AMD on mainstream CPU count and it will also likely cost a lot more than either. Keep in mind skylake igpu performance(without the insanely expensive iris pro versions which aren't really available for desktop) can basically not beat AMD's 5 year old GPU and CPU architecture based Bristol Ridge chips on 28nm... while Intel uses their most up to date CPU and GPU architecture on 14nm.
AMD gpu performance will likely be in the region of 2.5-3x higher than currently and Intel won't come close to matching that.
Back to Coffeelake and 10nm, if you think back to Broadwell, it was supposed to come over a year earlier than it did but 14nm delays and trouble with the early process caused issues. What was supposed to be a full product stack ended up only being the mobile dual cores, then they finally released quad core desktop Broadwell a year later a couple months before Skylake released.
Intel are going the same way with 10nm, significant delays, TWO added 14nm tocks(Kaby and Cannon, with Kaby being architecturally identical but with higher clocks), they've moved to only pushing dual core mobile products for the first release at 10nm and likely desktop quad/hex cores till a year or so later.
Intel won't have a 10nm higher power desktop part till late 2018 at the earliest now, Cannon Lake looks like Skylake architecture again but with hex core added to the mix and even so it offers nothing compelling, with 8 core CPU or twice the GPU performance from a quad core APU from AMD.
Soldato
- Joined
- 19 Feb 2011
- Posts
- 5,849
Massive Salt required...
So i just read this
https://www.reddit.com/r/Amd/comments/5rudia/amd_ryzen_model/
Listed also on that thread was
R7 1800X >$499
R7 1700X >$399
R7 1700 >$349
Now if you look at the chart, R7 1700X is same line as 7700k / 6800k, but the R7 1700X is a 8/16 chip @ 3.0-3.6ghz, a lot of its in chinese so i obviously dont really understand what is being said. I thought the 8/16 was going to have a 3.4 base? according to Lisa Su....
Looks like a load of old cobblers to me tbh. Mind you if that R7 1800X was $499 for 6900k performance, i can see them flying off the shelves.
Also if AMD need more cores to compete with Intels current offerings thats a bit lackluster to me personally... like needing a 6/12 to compete with an i5...
So i just read this
https://www.reddit.com/r/Amd/comments/5rudia/amd_ryzen_model/
Listed also on that thread was
R7 1800X >$499
R7 1700X >$399
R7 1700 >$349
Now if you look at the chart, R7 1700X is same line as 7700k / 6800k, but the R7 1700X is a 8/16 chip @ 3.0-3.6ghz, a lot of its in chinese so i obviously dont really understand what is being said. I thought the 8/16 was going to have a 3.4 base? according to Lisa Su....
Looks like a load of old cobblers to me tbh. Mind you if that R7 1800X was $499 for 6900k performance, i can see them flying off the shelves.
Also if AMD need more cores to compete with Intels current offerings thats a bit lackluster to me personally... like needing a 6/12 to compete with an i5...
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
No one anywhere has said base clock speed, the ES shown with I believe the Doom demo in December had a 3.4Ghz base clock, the ES shown in CES had a 3.6Ghz base clock and 4Ghz boost. The final clock speeds are unknown, though there was a hint the other day of a 3.7Ghz base clock on likely the highest end chip.
AMD historically have more than one high end chip that is unlocked.
Intel actually have LOADS of i7 and i5 variants of Skylake chips, but only one a piece that is a 7xxx K version chip and thus only one particularly overclockable and relevant to the likes of us.
AMD however has usually had at least 2 versions of any high end chip. So in Intel terms AMD would have a 7700k with a 4.4Ghz boost speed, but also likely a 7650k that is still 4 core/8 thread, but with 3.8Ghz boost and 3.4Ghz base clocks and price it lower, but still be unlocked and overclockable making it a far cheaper and better value chip for enthusiasts. It's also possible that they'll simply sell a downclocked 65W edition of the chip as they have done lately. IE 95W at max clocks possible, 65W version downclocked say 15% because in general with silicon you don't get linear gains, so you can usually save a huge amount of power with only relatively small clock speed reductions, ie drop power 35-40% with only 15-20% lower clocks.
As for more cores, who said AMD need more cores, Intel offer more cores as well... having more cores doesn't equate to needing more cores.
AMD historically have more than one high end chip that is unlocked.
Intel actually have LOADS of i7 and i5 variants of Skylake chips, but only one a piece that is a 7xxx K version chip and thus only one particularly overclockable and relevant to the likes of us.
AMD however has usually had at least 2 versions of any high end chip. So in Intel terms AMD would have a 7700k with a 4.4Ghz boost speed, but also likely a 7650k that is still 4 core/8 thread, but with 3.8Ghz boost and 3.4Ghz base clocks and price it lower, but still be unlocked and overclockable making it a far cheaper and better value chip for enthusiasts. It's also possible that they'll simply sell a downclocked 65W edition of the chip as they have done lately. IE 95W at max clocks possible, 65W version downclocked say 15% because in general with silicon you don't get linear gains, so you can usually save a huge amount of power with only relatively small clock speed reductions, ie drop power 35-40% with only 15-20% lower clocks.
As for more cores, who said AMD need more cores, Intel offer more cores as well... having more cores doesn't equate to needing more cores.
Last edited:
Soldato
- Joined
- 19 Feb 2011
- Posts
- 5,849
Heh
http://www.amd.com/en-us/innovation...ew-horizon&utm_medium=redirect&utm_source=301
@ 27.20 and onwards
Lisa Su "Lets talk clock speeds, i can tell you that Ryzen will have Base Clock Speeds of 3.4GHZ or Higher" they even show these on the slide behind her....
So yeah..AMD did say 3.4ghz base, Lisa Su said 3.4ghz Base
So i guess that Ryzen presentation is a fabrication then DM?
Caporegime
That chart you showed says base clockspeeds of under 3.4GHZ,and Lisa Su said clockspeeds above 3.4GHZ though.
So at this point you have just basically said the chart is wrong!!
Soldato
- Joined
- 19 Feb 2011
- Posts
- 5,849
Yeah read my title of my post where i put up the link to the chart, i said massive salt required, i in no way believe that chart, its utter nonsense Cat...
The rumor mill for Ryzen is hilarious, one week were only getting an 8/16, the next were getting an 8/8, were not getting a 6/12 the following week were getting a 6/12 etc etc
Nobody but AMD know the truth right now, and probably select partners who are churning out mobos etc for it.
All im interested in is solid info on core count, performance and price.
I want a 6/12 or 8/16 that offers at minimum the same performance as my 4770k in single threaded stuff, as that means in multithreaded stuff i will get a nice boost. And i dont want to have to sell one of my kids to buy it
Soldato
Been confirmed my Ryzen samples, but won't be till after launch ;X
Soldato
Ashes of the Singularity Crazy 4K benchmark on 4GHz Ryzen CPU leaked!
http://wccftech.com/amd-ashes-ryzen-4-0-ghz-benchmarks/
Well the benchmark of 4GHz R7 8C/16T Ryzen with Titan X Pascal at 5,100 is not very impressive.
Stock 4C/8T Haswell 4790K with a GTX 1080 will smash it at higher score of 5,400:
http://www.ashesofthesingularity.co...-details/95e673f9-2824-4da5-8571-f6addf6368ee
Stock 4C/8T Haswell 4790K with a Titan X Pascal score 5,800:
http://www.ashesofthesingularity.co...-details/c03f9aba-c562-460e-95ae-6bc08a89c1b8
Stock 4C/8T Skylake 6700K with a Titan X Pascal score impressive 6,900:
http://www.ashesofthesingularity.co...-details/4ea54e6f-ce58-45dd-aec2-43d2a0f56d95
Shame nobody benchmarked it with Kaby Lake 7700K and uploaded result yet.
http://wccftech.com/amd-ashes-ryzen-4-0-ghz-benchmarks/
Well the benchmark of 4GHz R7 8C/16T Ryzen with Titan X Pascal at 5,100 is not very impressive.
Stock 4C/8T Haswell 4790K with a GTX 1080 will smash it at higher score of 5,400:
http://www.ashesofthesingularity.co...-details/95e673f9-2824-4da5-8571-f6addf6368ee
Stock 4C/8T Haswell 4790K with a Titan X Pascal score 5,800:
http://www.ashesofthesingularity.co...-details/c03f9aba-c562-460e-95ae-6bc08a89c1b8
Stock 4C/8T Skylake 6700K with a Titan X Pascal score impressive 6,900:
http://www.ashesofthesingularity.co...-details/4ea54e6f-ce58-45dd-aec2-43d2a0f56d95
Shame nobody benchmarked it with Kaby Lake 7700K and uploaded result yet.
Last edited: