Discussion AMD Ryzen MegaThread! FAQ and Resources

[Redneck5439]

Still nothing that I can find that one could call official, but things are still looking promising.

http://wccftech.com/amd-ryzen-7-1800x-overclocked-benchmarks/

https://www.forbes.com/sites/antonyleather/2017/02/28/amd-ryzen-new-benchmark-leaks-world-records-intel-price-drops-and-what-to-expect-on-launch-day/#7eb06ce625df

 

[warmon6]

Still nothing that I can find that one could call official, but things are still looking promising.

http://wccftech.com/amd-ryzen-7-1800x-overclocked-benchmarks/

https://www.forbes.com/sites/antonyleather/2017/02/28/amd-ryzen-new-benchmark-leaks-world-records-intel-price-drops-and-what-to-expect-on-launch-day/#7eb06ce625df

Not laughing at you but at the leaked benchmark in the forbes article that has World of Tanks in it. Bless that person soul for doing a benchmark but WoT isn't that demanding of a game 😀.

 

[french toast]

Asus is faltering in the bang/buck department, although if I can get Sonic Radar at a good price (unlikely, as it's a ROG feature), that may win out.
I'm not sure I'm ready to drop $400 on a CPU, but Ryzen may really be the case where "moar cores!!!" will be worthwhile, so I don't think I'll be going below the 1500.

I'm somewhat partial to Gigabyte and Asus boards. I've not had either have one go out on me. And for OC'ing, I do like Asus on their higher end boards. My Xeon X3470 (which may be replaced by a Ryzen setup) is going along fairly happily at 4.2ghz on a P7F7-E WS motherboard. Well, maybe not completely happily as stock speed is 2.9ghz. But it doesn't twitch at all.

As for the 'moar cores', that IS part of the advantage of the Ryzen equation. They don't quite have the IPC up to snuff - Intel has them beat on that. BUT - their core counts make up for that deficit in spades, and it's bolstered immensely by the price/core side of things. The cores are quick enough, and they're pricing them VERY aggressively. IMHO if you don't buy at least a hex-core, you should just stick with Intel since they have a higher IPC per core. But once you get past the quad core into hex core territory, the Ryzen chips will be the much better buy.

I'm not sure I'd agree on that- if the leaks are true you have choice of 4 real cores vs 2 core + smt on the low end... and then 4 core + smt vs 4 core without on the mid range. Given the performance is so close I'd still argue I'd rather have a true quad over the i3 (although I admit that the i3 copes remarkably well as a gaming chip). I guess that position does depend on the benchmarks of course, however based on what we've seen so far I'm fairly happy the Ryzen R3 should be ahead of an i3 due to 4 full cores (with turbo) vs 2...

I agree its not that big of a difference now if the leaks are true and i'm sure they are at this point. 10-15% more IPC isn't worth getting when you get half the amount of cores. That is coming from a person who would take a I3 skylake-haswell over a 8 core fx for my gaming.

It gets complicated when you think that the 1800X matched the i7 6900k on a heavily-threaded benchmark. It means that yes, Ryzen has lower IPC, but when threads are on use, the AMD can scale as well as an Intel. So, I'd say the R3 is always a better choice over the i3. And that's coming from a person who would take an 8-core FX over an i3 for everything else, and stick to adequate 60-fps gaming.

Videocardz.com leaked the official cinebench results. AMD was sandbagging the 1800X the whole time. All the chips they showed were the 1700X. The 1800X matches the 6900K in ST and blows the doors off in multithread by 150 points.

https://videocardz.com/66216/firefly-was-a-great-show

The numbers for the 6900k in that slide are a bit strange

Juan Anandtech uses a water cooler in their 6900k reviews, AMD uses stock 140w Intel cooler, if you were to water cool the 1800x what score would it get?

IF Anandtech used a watter cooler for the 6900k and AMD uses the stock air cooler for the 6900k, then WHY Anandtech got 153 score, whereas AMD got a higher score of 162?

Simple juan, Amd was decent enough to go into bios and turn on turbo boost 3 @4ghz, whilst using stock coolers.
This helped 6900k ST scores whilst the 6900k MT score was slightly throttling, this is intels fault for creating a crap air cooler.
Using anandtechs water cooler numbers compared to ryzens stock air still puts ryzen ahead.

Now assuming ryzen used xfr to boost to 4.1ghz (ive heard it was NOT used but ill digress) we get Ryzen being around 2% slower ipc than broadwell,which is margin of error stuff.
Now cinbench being the benchmark you said would determine ipc above all others i assume you now accept Ryzen ~ broadwell ipc?

If you take other tests such as leaked gaming/passmark/cpuz/CB MT as well as efficiency we can determine with realistic confidence that Ryzen is a better core than broadwell - all things considered.
The latency numbers are troubling yes, but with such buggy bios right now its hard to get a concrete picture, the only thing we do know is Ryzens cache system is robust enough for it not to matter in in real scenarios, evidenced by leaked gaming benchmarks which are latency sensitive and show ryzen beating broadwell.
Then there is the fact that ryzens IMC seems to be more efficient than intels extracting more bandwidth out of same memory than even skylake.

All in all we can conclude AMD has made a better more efficient version of broadwell, on a worse process no less

 

[french toast]

CPCHardware confirmed me there is a problem with the IMC in RyZen and it affects gaming

https://mobile.twitter.com/CPCHardware/status/836346777267761155/actions

Check also the benches given here

http://www.tomshardware.co.uk/forum/id-3327589/amd-ryzen-megathread-faq-resources/page-8.html#19350306

Also it seems reviewers are confirming problems with overclocking

https://www.reddit.com/r/Amd/comments/5wkcbi/computerbasede_editor_hints_details_about_their/

For those that don't know which is the video that Computer base is talking about. It is a video taken at RyZen Tech Day

https://www.youtube.com/watch?v=vFYlP_3dC1U

Check around 19:10. When AMD tried to show how RyZen overclocks the system crashed on live. It crashed when the frequency got 4.1GHz. It was a watter cooled chip. You can see reviewers face, some even laugh.

We have to wait for reviews

Thanks for the links juan.
Also remember scan have pre overclocked r7 1800x at 4.2ghz, they dont traditionally do max overcking for obvious reasons, as was the case foe for kabylake being 200mhz lower.
Now assuming Ryzen gets only 4.2 max, that has to be compared to 1050$ 6900k only achieving 4.3ghz

your correct we need reviews.

 

[Redneck5439]

One key thing to remember as Ryzen launches and "buggy" behavior is discovered here and there is AMD is dealing with a whole new architecture. Hopefully any bugs can be quickly worked out via software/ bios updates and aren't anything to worry about. I'm sure Intel and Intel friendly reviews will try to make the most out of any bios issues AMD may have at launch, however hopefully we can all see through that for what it is.

Right now, with all the early/ leaked benchmarks showing basically the same thing across the board, I think it is safe to assume that Ryzen will perform as predicted with no bombshell issues. A big question is if Ryzen will get fair and unbiased reviews or will some yield to Intel's pressures. One can only guess what Intel will "discuss" with reviewers before they write their articles, for those reviewers who yield to the pressure and contact Intel before publishing their reviews.

If Ryzen is everything I expect it to be I wonder how long it will be before Intel rolls out a new architecture. I personally believe that Kaby Lake is pushing their current architecture to near limit of what it is capable of.

 

[french toast]

One key thing to remember as Ryzen launches and "buggy" behavior is discovered here and there is AMD is dealing with a whole new architecture. Hopefully any bugs can be quickly worked out via software/ bios updates and aren't anything to worry about. I'm sure Intel and Intel friendly reviews will try to make the most out of any bios issues AMD may have at launch, however hopefully we can all see through that for what it is.

Right now, with all the early/ leaked benchmarks showing basically the same thing across the board, I think it is safe to assume that Ryzen will perform as predicted with no bombshell issues. A big question is if Ryzen will get fair and unbiased reviews or will some yield to Intel's pressures. One can only guess what Intel will "discuss" with reviewers before they write their articles, for those reviews who yield to the pressure and contact Intel before publishing their reviews.

If Ryzen is everything I expect it to be I wonder how long it will be before Intel rolls out a new architecture. I personally believe that Kaby Lake is pushing their current architecture to near limit of what it is capable of.

Current roadmaps point to icelake on 10nm in Q1 2019.
Maybe they can push that to Q4 2018 if really really pushed, excepting lower yeilds in the process.

Who knows what this will provide performance wise? Assuming they go wider with HT x4 then could be massive jump.

Then again zen++ on 7nm is thought to be H1 2019

 

[Eximo]

Must be an old process map.

Cannonlake, 8th gen should be on 10nm sometime this year. Likely see ultrabook processors for the holiday season as per usual.

 

[TechyInAZ]

Must be an old process map.

Cannonlake, 8th gen should be on 10nm sometime this year. Likely see ultrabook processors for the holiday season as per usual.

10nm is gona be interesting. I wonder if it will be just as resilient to voltage as 14nm finFET is.

 

[gamerk316]

Must be an old process map.

Cannonlake, 8th gen should be on 10nm sometime this year. Likely see ultrabook processors for the holiday season as per usual.

10nm is gona be interesting. I wonder if it will be just as resilient to voltage as 14nm finFET is.

Be careful talking process. Intel's is a "true" 10nm node, everyone else, not so much. It's getting quite arbitrary really.

Intel won't do much to counter; you'll maybe see six/eight core chips in the enthusiast market, but I doubt Intel is going to change coarse. Core is remarkably optimized, and there's really only so much low hanging fruit left to clean up. As I've said for a while now: As we reach the end of die shrinks, we reach "peak computing". I fully expect everyone to end up stuck at more or less the same performance level while we work out how to replace current CPU design philosophies.

 

[Eximo]

Still 2018 for the desktop processor, high hopes for Z270 support. Then I might pick one up.

Got burned on Broadwell with my Z87 board. Not that Broadwell was that interesting, but the option would have been nice.

 

[french toast]

Must be an old process map.

Cannonlake, 8th gen should be on 10nm sometime this year. Likely see ultrabook processors for the holiday season as per usual.

Nope i was referring to mainstream desktop, cannonlake 10nm is for mobile Q4 2017.
Coffeelake 14nm+ (50% more threads) either Q4 2017 or Q1 2018.
Icelake new uarch+ 10nm either Q4 2018 or Q1 2019 .

There is no cannonlake 10nm mainstream desktop as far as i know, and rumours point to no cannonlake 10nm HEDT 2018 either.

Likely Coffeelake X or something.

 

[010TheMaster010]

Speaking to the review boxes sent out, they definitely are mostly different. I've seen three now and each one has had a different(or no) cooler/motherboard. Also, a reminder, for large multiquotes, use spoilers or it's really hard on mobile users.

 

[svengalis]

dont think its been posted but new benchmarks!

[video="https://www.youtube.com/watch?v=ax3i-K6-VaA"][/video]

http://www.dsogaming.com/news/more-amd-ryzen-7-1700x-versus-intel-i7-6800k-gaming-benchmarks-leaked-online/

 

[Redneck5439]

Must be an old process map.

Cannonlake, 8th gen should be on 10nm sometime this year. Likely see ultrabook processors for the holiday season as per usual.

10nm is gona be interesting. I wonder if it will be just as resilient to voltage as 14nm finFET is.

Be careful talking process. Intel's is a "true" 10nm node, everyone else, not so much. It's getting quite arbitrary really.

Intel won't do much to counter; you'll maybe see six/eight core chips in the enthusiast market, but I doubt Intel is going to change coarse. Core is remarkably optimized, and there's really only so much low hanging fruit left to clean up. As I've said for a while now: As we reach the end of die shrinks, we reach "peak computing". I fully expect everyone to end up stuck at more or less the same performance level while we work out how to replace current CPU design philosophies.

Intel no doubt about it are in the drivers seat. They have built a well deserved reputation on power and performance and while Bulldozer stymied in the mid range, iCore vaulted into high performance history. AMD has everything to prove with Ryzen, whereas Intel really have nothing to prove to anyone. That being said, I really don't think a die shrink to 10nm is going to be enough if Ryzen debuts where it looks like it will and AMD can get get more performance with their second generation. Ryzen is in first generation so it should, in theory, be able to be pushed much harder and optimized much more than Intel's iCore which is entering 8th generation. Intel's process is extremely optimized and can't be pushed much further. AMD has a real chance of recovering some market share before Intel can release a new architecture.

 

[Embra]

I remembered reading somewhere (sorry I do not have a link) that intel has had problems with 10mm and that AMD may be skipping to 7mm.

 

[010TheMaster010]

This is true, but will (likely) be a 7/10 hybrid mode, as their 14 is a 14/20 hybrid(as far as I know)

 

[french toast]

Yes, globalfoundries moves to 7nm H2 2018, it is expected (not confirmed ) that amd will use this 7nm for zen 3 in H1 2019.

Intels 10nm yeilds suck, hence why there will no cannonlake desktop in 2018, they will use 14nm+ Coffeelake instead.
It is expected Globalfoundries 7nm will be equal to intels 10nm, so in 2019 they will have process parity for the first time in 10 years.

 

[french toast]

This is true, but will (likely) be a 7/10 hybrid mode, as their 14 is a 14/20 hybrid(as far as I know)

No this has nothing to do with Samsungs 14nm Lpp process which they use now, it is a ground up design using the purchased/aquired IBM semiconductor team/fabs/IP.

Should equal intels 10nm, at least.

 

[jaymc]

I guess well find out tonight... Does that mean it will fall in different zones one by one..?

For example first in UK & Ireland and hours later in the states the plot thickens..

I think history would be a good indicator?

As in, when they announced Bulldozer, was it US Pacific Time, Atlantic, Central, UK-London time? And so on.

EDIT:

Bulldozer was:
Toms Review date: October 12, 2011 at 7:10 AM
AMD Release date: Under NDA I think.

Cheers!

Pretty sure i remember refreshing toms page over and over again at 11:59pm eastern time until i saw bulldozer's review at 12am. Was so exited i skipped to the conclusion and basically cried lol

lol... haha.

Still no reviews.. A watched kettle never boils eh.. I was checking at five this AM an all.. :-(

 

[Redneck5439]

I guess well find out tonight... Does that mean it will fall in different zones one by one..?

For example first in UK & Ireland and hours later in the states the plot thickens..

I think history would be a good indicator?

As in, when they announced Bulldozer, was it US Pacific Time, Atlantic, Central, UK-London time? And so on.

EDIT:

Bulldozer was:
Toms Review date: October 12, 2011 at 7:10 AM
AMD Release date: Under NDA I think.

Cheers!

Pretty sure i remember refreshing toms page over and over again at 11:59pm eastern time until i saw bulldozer's review at 12am. Was so exited i skipped to the conclusion and basically cried lol

lmao... haha. Still no reviews.. A watched kettle never boils eh.. I was checking at five this AM an all.. :-(

Starting to wonder if the 28th was the correct date, think I read something in a recent article that was quoting March 2nd. Either that, or there are a lot of reviewers waiting for actual retail samples and not using the samples AMD sent them. Was really expecting a couple actual reviews by now.

I'm starting to figure out what things I can keep from my current build and what things will have to be upgraded from the start. I'm hoping to use my existing power supply (1000W 80+ Gold) which has done fine feeding my FX 8370 at 5+ Ghz. I'm just finding it hard to believe that the R7 1800X is actually going to be 95W at stock when the FX 8370 was 125W at stock. Wish we could get some real reviews so we know what the actual power usage is and where the verifiable benchmarks put the R7 1800X.

 

[juanrga]

Now assuming ryzen used xfr to boost to 4.1ghz (ive heard it was NOT used but ill digress) we get Ryzen being around 2% slower ipc than broadwell,which is margin of error stuff.
Now cinbench being the benchmark you said would determine ipc above all others i assume you now accept Ryzen ~ broadwell ipc?

If you take other tests such as leaked gaming/passmark/cpuz/CB MT as well as efficiency we can determine with realistic confidence that Ryzen is a better core than broadwell - all things considered.
The latency numbers are troubling yes, but with such buggy bios right now its hard to get a concrete picture, the only thing we do know is Ryzens cache system is robust enough for it not to matter in in real scenarios, evidenced by leaked gaming benchmarks which are latency sensitive and show ryzen beating broadwell.
Then there is the fact that ryzens IMC seems to be more efficient than intels extracting more bandwidth out of same memory than even skylake.

All in all we can conclude AMD has made a better more efficient version of broadwell, on a worse process no less

Wouldn't we wait to proper reviews before assuming which is the final IPC or which is the efficiency?

About IPC, recall that traditionally CB has been a good measure of average IPC for AMD chips (and the reason why AMD used CB in talks about Kaveri, Carrizo, or Bristol), but "The Stilt" said recently in an AT forum that CB was favoring RyZen now. I interpret his words in the sense of CB giving performance above the average.

About efficiency recall that CPCHardware confirmed me that the real TDP of the QS far exceeds the displayed TDP. And the Computerbase editor is confirming the same for the retail chips: "That the new big cooler is designed for 125W is not a coincidence, even if the TDP of the big R7 actually says 95W."

My estimation of 105W, which CPCHardware confirmed, was for base clocks and turbo disabled. Adding ACT I obtain a real TDP of about 117W, if I did no mistake.

Let us wait to reviews.

CPCHardware confirmed me there is a problem with the IMC in RyZen and it affects gaming

https://mobile.twitter.com/CPCHardware/status/836346777267761155/actions

Check also the benches given here

http://www.tomshardware.co.uk/forum/id-3327589/amd-ryzen-megathread-faq-resources/page-8.html#19350306

Also it seems reviewers are confirming problems with overclocking

https://www.reddit.com/r/Amd/comments/5wkcbi/computerbasede_editor_hints_details_about_their/

For those that don't know which is the video that Computer base is talking about. It is a video taken at RyZen Tech Day

https://www.youtube.com/watch?v=vFYlP_3dC1U

Check around 19:10. When AMD tried to show how RyZen overclocks the system crashed on live. It crashed when the frequency got 4.1GHz. It was a watter cooled chip. You can see reviewers face, some even laugh.

We have to wait for reviews

Thanks for the links juan.
Also remember scan have pre overclocked r7 1800x at 4.2ghz, they dont traditionally do max overcking for obvious reasons, as was the case foe for kabylake being 200mhz lower.
Now assuming Ryzen gets only 4.2 max, that has to be compared to 1050$ 6900k only achieving 4.3ghz

your correct we need reviews.

If you check the third link you can see to the Computerbase editor commenting on those pre-overclocked systems: "Other reviewers (like us) mostly also don't manage that level of OC with their Ryzen-samples. We are all doing the AMD bluescreen."

 

[Redneck5439]

Now assuming ryzen used xfr to boost to 4.1ghz (ive heard it was NOT used but ill digress) we get Ryzen being around 2% slower ipc than broadwell,which is margin of error stuff.
Now cinbench being the benchmark you said would determine ipc above all others i assume you now accept Ryzen ~ broadwell ipc?

If you take other tests such as leaked gaming/passmark/cpuz/CB MT as well as efficiency we can determine with realistic confidence that Ryzen is a better core than broadwell - all things considered.
The latency numbers are troubling yes, but with such buggy bios right now its hard to get a concrete picture, the only thing we do know is Ryzens cache system is robust enough for it not to matter in in real scenarios, evidenced by leaked gaming benchmarks which are latency sensitive and show ryzen beating broadwell.
Then there is the fact that ryzens IMC seems to be more efficient than intels extracting more bandwidth out of same memory than even skylake.

All in all we can conclude AMD has made a better more efficient version of broadwell, on a worse process no less

Wouldn't we wait to proper reviews before assuming which is the final IPC or which is the efficiency?

About IPC, recall that traditionally CB has been a good measure of average IPC for AMD chips (and the reason why AMD used CB in talks about Kaveri, Carrizo, or Bristol), but "The Stilt" said recently in an AT forum that CB was favoring RyZen now. I interpret his words in the sense of CB giving performance above the average.

About efficiency recall that CPCHardware confirmed me that the real TDP of the QS far exceeds the displayed TDP. And the Computerbase editor is confirming the same for the retail chips: "That the new big cooler is designed for 125W is not a coincidence, even if the TDP of the big R7 actually says 95W."

My estimation of 105W, which CPCHardware confirmed, was for base clocks and turbo disabled. Adding ACT I obtain a real TDP of about 117W, if I did no mistake.

Let us wait to reviews.

CPCHardware confirmed me there is a problem with the IMC in RyZen and it affects gaming

https://mobile.twitter.com/CPCHardware/status/836346777267761155/actions

Check also the benches given here

http://www.tomshardware.co.uk/forum/id-3327589/amd-ryzen-megathread-faq-resources/page-8.html#19350306

Also it seems reviewers are confirming problems with overclocking

https://www.reddit.com/r/Amd/comments/5wkcbi/computerbasede_editor_hints_details_about_their/

For those that don't know which is the video that Computer base is talking about. It is a video taken at RyZen Tech Day

https://www.youtube.com/watch?v=vFYlP_3dC1U

Check around 19:10. When AMD tried to show how RyZen overclocks the system crashed on live. It crashed when the frequency got 4.1GHz. It was a watter cooled chip. You can see reviewers face, some even laugh.

We have to wait for reviews

Thanks for the links juan.
Also remember scan have pre overclocked r7 1800x at 4.2ghz, they dont traditionally do max overcking for obvious reasons, as was the case foe for kabylake being 200mhz lower.
Now assuming Ryzen gets only 4.2 max, that has to be compared to 1050$ 6900k only achieving 4.3ghz

your correct we need reviews.

If you check the third link you can see to the Computerbase editor commenting on those pre-overclocked systems: "Other reviewers (like us) mostly also don't manage that level of OC with their Ryzen-samples. We are all doing the AMD bluescreen."

Until the actual reviews come out its hard to say what to expect for overclocking and power management. I don't think we will see huge surprises from what the early leaked benchmarks have shown, but overclocking and power management might yield some surprises. For one thing I would find it hard to believe that right off the bat the bios for Ryzen won't need an update or two to work out the bugs.

I also find it very hard to believe that the actual TDP for the 8 core 16 thread R7 1800X is only 95W. To be honest 125W wouldn't really surprise me, but if AMD comes in at your estimated 105W - 117W that would still be amazing considering the i7 6900K is has a TDP listed at 140W. If AMD actually pulls off 95W though that would make for a "cool running" 8 core 16 thread processor and leave a little more headroom for overclocking, in theory.

 

[Redneck5439]

http://marketrealist.com/2017/02/amds-ryzen-versus-intels-core-a-performance-face-off/

Another interesting read awaiting actual reviews. I think the Cinebench scores quoted were provided by AMD, but still an interesting read.

 

[8350rocks]

What about this one, these guys supposedly broke a world record in Cinebench.

https://www.techpowerup.com/231047/amd-ryzen-7-1800x-cracks-5-20-ghz-oc-with-ln2-and-all-eight-cores-enabled
https://www.youtube.com/watch?v=OAQCWgqZXzk

I have seen that reported in multiple site. I don't understand it. I don't understand how 2363 points can be a world record when there are lots of scores above that. The top score and worldwide record (for 8-cores) is 2445

http://hwbot.org/submission/3233876_the_overclocking_knights_cinebench___r15_core_i7_5960x_2445_cb

First attempt was 2363

Second Attempt was 2426

Third Attempt was 2449

CPU/Memory caching coming into play? Scores went up each run, which infers something is getting cached somewhere.

Minor frequency bump as well...IIRC.

 

[010TheMaster010]

@French Toast
Why do you think it will be >= Intel's 10nm?
Given past history, between all major fabs that produce CPUs (SS, Intel, IBM, GF) Intel has the best node and is closest to the marketed number. GF/SS 14 is not a true 14nm and nothing (to me) points to true 7nm, or it equalling intels 10nm that will be slightly mature by that time.