Discussion AMD Ryzen MegaThread! FAQ and Resources

[warmon6]

This leak about a new 16 core zen states that the "gaming issues" have been "Ironed out" by new silicon revisions:

https://forums.anandtech.com/threads/ryzen-strictly-tec...

Poster goes on to say that these new silicon revisions will be available right down through all SKU's

That link is broken, and a search in the forums didn't come up with it. Could you post again, please?

From looking at the url, im guessing it somewhere in this thread. "Ryzen: Strictly technical"

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

Although im guessing it links to a single post but im not sure which post it would be (bad link or deleted post/comment).

 

[jaymc]

here sry try it now: https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-34#post-38797798

An here:
http://www.nag.co.za/2017/03/17/amd-reportedly-has-a-second-zen-launch-for-high-end-desktops-coming-soon/

 

[Rogue Leader]

here sry try it now: https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-34#post-38797798

An here:
http://www.nag.co.za/2017/03/17/amd-reportedly-has-a-second-zen-launch-for-high-end-desktops-coming-soon/

As a Moderator to help the discussion along I think I need to step in here and nip this in the bud

OrangeKhrush is a former member here who went by a different name, I would take any information from him with a grain of salt. I wouldn't waste any effort on information from him.

Thanks all

 

[juanrga]

I have been watching that AT thread: People in that thread has been promising us fixes since launch. First a BIOS update was going to fix gaming, then changed to a SMT fix, latter a W10 scheduler patch,...

Reviews (PcPer, TechSpot, Hardware.fr, computerbase.de) and even AMD have disproved everything what they say up to now.

The last promise is coming from OrangeKrush. He is promising us silicon revisions based in the nonsense that the RyZen chips in the market are rebranded engineering samples (sic), that everyone that has purchased RyZen is a "beta tester" and that the real final silicon is coming next month. Here you have an example of he posting nonsense

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-29#post-38793748

 

[cdrkf]

@Juan, I do find this hysteria about ryzens 'bad' gaming performance slightly confusing.

The only areas where I think there are a problem are the handful of titles that drop performance with smt on (but don't do the same on Intel), and the couple of well threaded titles that simply don't scale on ryzen but do on the 6900k (e.g. AOTS). I think those are edge cases, but still things that are likely to get fixed down the line (if not in existing titles I at least expect the behavior to become non existent in the future).

Still, a few edge cases isn't the end of the world. Generally ryzen is pretty strong- latest tests on mass effect Andromeda for example put ryzen just behind the 7700k, which is perfectly acceptable imo.

 

[gamerk316]

@Juan, I do find this hysteria about ryzens 'bad' gaming performance slightly confusing.

The only areas where I think there are a problem are the handful of titles that drop performance with smt on (but don't do the same on Intel), and the couple of well threaded titles that simply don't scale on ryzen but do on the 6900k (e.g. AOTS). I think those are edge cases, but still things that are likely to get fixed down the line (if not in existing titles I at least expect the behavior to become non existent in the future).

Still, a few edge cases isn't the end of the world. Generally ryzen is pretty strong- latest tests on mass effect Andromeda for example put ryzen just behind the 7700k, which is perfectly acceptable imo.

Here's what happened:

A few months prior to release: Expectations: Haswell
A few days prior to release: Expectations: Better then Kaby Lake
Release: Actual Performance: Haswell
Expectations: Missed

Hype is a double edged sword.

 

[-Fran-]

Here's what happened:

A few months prior to release: Expectations: Haswell
A few days prior to release: Expectations: Better then Kaby Lake
Release: Actual Performance: Haswell
Expectations: Missed

Hype is a double edged sword.

And who the hell was expecting it to be better than Kaby Lake? lol

Blowing stuff out of proportion and hyperbole are also problematic things, you know?

I never once saw a slide from AMD blowing things out of proportion nor saying Ryzen was going to be better than Intel in all disciplines. I'm even more inclined to think people that blows or hypes are usually from the other camp, ironically.

Cheers!

 

[cdrkf]

@yuka the main offender was the slide AMD showed thremselves that pitched the 1800x vs the 7700k in games, showing it just behind or just ahead in a range of titles.

The slide wasn't wrong, just based on 4k numbers, which inadvertently set the expectation that the 1800x would maintain that relative position in all situations.

I still think a higher thread count cpu isn't a bad bet for longevity, especially where it's not proving a major bottleneck in lower thread stuff. Games are making use of more threads as time goes on, I could see 12 being useful in the not too distant future though I accept 16 is probably overkill for the lifespan of this generation of cpu.

 

[-Fran-]

And that was a 1700, not a 1800X IIRC; I really don't remember AMD testing the 1800X against the 7700K other than in the demos where they put a lot of stress to the CPUs to show how the 1800X could pull ahead, which is not that un-realistic (streaming scenario), but I will concede their testing demos and slides are just as bad as any other company. Those are the evil shenanigans of the Marketing Department.

This lunch has been populated with the usual hyperbole from adversaries and blowing some thing out of proportion by the usual suspects in every forum. I'm not even counting raging fanbois, but *supposedly* level headed people. Agendas and all.

That being said, I still have issues seeing where AMD made the "hype train" go faster. This is a million times better than the BD release and people still have issues accepting that Ryzen is a good product. Ensue Jackie Chan meme.

Cheers!

 

[gamerk316]

I'm not disagreeing. Ryzen actually did a bit better then my expectations. But did anyone else pay attention to the forum in the week leading up to release? "Is Intel doomed?" "How long until AMD is worth more then Intel", and so on and so forth. The hype ran out of control based off numbers AMD themselves put out, and it backfired. AMD still hasn't regained it's lost stock price over "missed" expectations.

 

[cdrkf]

I'm not disagreeing. Ryzen actually did a bit better then my expectations. But did anyone else pay attention to the forum in the week leading up to release? "Is Intel doomed?" "How long until AMD is worth more then Intel", and so on and so forth. The hype ran out of control based off numbers AMD themselves put out, and it backfired. AMD still hasn't regained it's lost stock price over "missed" expectations.

Yeah I agree- it did get out of hand. Still I wouldn't worry about the stock price- it's been bouncing between $13 and $14 ever since the Ryzen launch (it peaked at $14.5 I think?) so it hasn't dropped much. Nothing like it was the last few years (I mean at $2 - $3 AMD was seriously undervalued).

My guess is that the stock might climb a bit higher after the R5 parts launch. I'm then expecting it to drop a bit based on the graphics side based on the info coming out on the 'RX 500' series all being re-brands (although on an improved process at least) and the fact that Vega is unlikely to be faster than the 1080ti. Then it will probably climb again if the R3 parts are any good.

Also what do we all make of the supposed AMD 'high end' platform with a 16 core / 32 thread part with quad channel memory? It sounds like wishful thinking to me but I'd be interested if anyone thinks there's some validity to the claims (I mean it's not totally impossible given AMD will have 16 core 'naples' server parts, I'm just not convinced they'll release them on another platform as well)...?

 

[-Fran-]

Ok, that is fair to say; I also saw a lot of that behavior, but then again we all know those comments are flame-bait. No need to take them seriously, right?

That is why I'm having issues seeing where AMD made things worse this time around; the marketing (for better or for worse) was at the same level as the rest of the companies out there putting their product in a good light, but never once I saw an over-promise from them.

Cheers!

 

[blazorthon]

I had a 965C at 3.9Ghz and still have my A8 3850 ~2.5Ghz... The 965 was indeed faster, but due to clocks only. At the same speed, they were the same CPU and the A8 felt faster with a GPU attached (4890 at the time). You can even go back to the benchies at Anand; the differences are mostly due to clocks and not the lack of L3. I remember having a *very* lenghty discussion with someone at the time with several benchmarks from different places and there were things for both arguments; so for me at the end they were the same.

That is to say, I do think having L3 is better, but not a necessity. If there's not much L3 usage, it's wasted space anyway.

Cheers!

idk, some workloads didn't mind the lack of L3, some had a double digit performance difference in my testing a few years back. It wasn't huge, it wasn't even necessarily noticeable in many cases, but it happened in my testing at least. I don't have an Athlon II anymore to rerun things with, but I could try powering up my old Phenom II system if you want fresh benches.

 

[gamerk316]

Remember: On-board cache is just a main memory speedhack. If it's not utilized, its literally just a power-hungry die wasting amount of space.

Nowadays, most stuff will use it to some extent, but for applications not bound by memory bandwidth, you won't see nearly as measurable a performance benefit.

 

[aldaia]

See the thing that's wrong with that statement is Intel isn't selling me 1 core compared to lets say 6 from Amd they are selling me 4 cores and Amd would be selling 6. If games use one core at higher usage compared to the other 5 cores then we might have a issue with the game not getting the same numbers compared to a faster per core processor. Most games are starting to finally use 4 cores at least. Anything over a 1060/480 would bottleneck a I3 and you won't be getting the full usage out of your GPU

I don't get why some people in this thread are trying to convince me that something more expensive and with less performance (for my needs) is better, My top priority is not gaming, gaming is secondary, i never spend more than $150-180 on a GPU, any CPU is good enough to get full usage of that. On the other hand I compile a lot, and that is an area where Ryzen shines. I also run simulations that in all aspects are perfectly parallel with zero data communication. I can even spawn the exact number of threads and fix them to specific CPUs to avoid thread migration (simulations run for several days). So my expectations are to do at least as good as MT Cinebench if not better. Why the hell should I buy an i7 7700K for $350 when a $250 R5 1600X will give me more performance where it matters and exactly the same while gaming? And if I can spare $350, what I should buy is R7 1700.

No one is trying to convince you i said several times that Ryzen is superior to Intel in everything but gaming for the money.

In your case Ryzen is amazing in almost all cases but gaming Ryzen is at least not bad. Once i get more $$$ a 1700 is becoming my rig that handles my handbrake machine and some other tasks.

If you are doing work like you said i'd never recommend a I5 over a 6 core Ryzen part, if anything i'd try and tell you to get a 1700 at least for your work try and fit it in the budget

Lame fact about me, i'm having more fun from BOTW on my switch then i have been on my PC for 2 years. Power is not everything(also why i'm not as active on the forums).

Sorry If I misunderstood you.

I agree that 1700 is probably the better choice for my needs. Though, I think that, the 6 core parts are the sweet spot in terms of performance/$. 1600X should be about 75% the performance of 1800X (maybe more since L3 cache size is the same) for half the price. 1600 should also be 75% the performance of 1700 for 66% the price.

4 core parts don't seem to offer same level of performance/$ 1500X is aprox half 1700X but price is also half of it. 1400 seems to be 50% of a 1700 but for more than 50% price.

 

[daerohn]

See the thing that's wrong with that statement is Intel isn't selling me 1 core compared to lets say 6 from Amd they are selling me 4 cores and Amd would be selling 6. If games use one core at higher usage compared to the other 5 cores then we might have a issue with the game not getting the same numbers compared to a faster per core processor. Most games are starting to finally use 4 cores at least. Anything over a 1060/480 would bottleneck a I3 and you won't be getting the full usage out of your GPU

I don't get why some people in this thread are trying to convince me that something more expensive and with less performance (for my needs) is better, My top priority is not gaming, gaming is secondary, i never spend more than $150-180 on a GPU, any CPU is good enough to get full usage of that. On the other hand I compile a lot, and that is an area where Ryzen shines. I also run simulations that in all aspects are perfectly parallel with zero data communication. I can even spawn the exact number of threads and fix them to specific CPUs to avoid thread migration (simulations run for several days). So my expectations are to do at least as good as MT Cinebench if not better. Why the hell should I buy an i7 7700K for $350 when a $250 R5 1600X will give me more performance where it matters and exactly the same while gaming? And if I can spare $350, what I should buy is R7 1700.

No one is trying to convince you i said several times that Ryzen is superior to Intel in everything but gaming for the money.

In your case Ryzen is amazing in almost all cases but gaming Ryzen is at least not bad. Once i get more $$$ a 1700 is becoming my rig that handles my handbrake machine and some other tasks.

If you are doing work like you said i'd never recommend a I5 over a 6 core Ryzen part, if anything i'd try and tell you to get a 1700 at least for your work try and fit it in the budget

Lame fact about me, i'm having more fun from BOTW on my switch then i have been on my PC for 2 years. Power is not everything(also why i'm not as active on the forums).

Sorry If I misunderstood you.

I agree that 1700 is probably the better choice for my needs. Though, I think that, the 6 core parts are the sweet spot in terms of performance/$. 1600X should be about 75% the performance of 1800X (maybe more since L3 cache size is the same) for half the price. 1600 should also be 75% the performance of 1700 for 66% the price.

4 core parts don't seem to offer same level of performance/$ 1500X is aprox half 1700X but price is also half of it. 1400 seems to be 50% of a 1700 but for more than 50% price.

I do not think there will be so much difference in games. As there are not a lot of title nowadays that can use 8 cores 16 thread CPU in full range. A 6 core 12 thread CPU would nearly be as good as a 8 core 16 thread cpu at the same clock speeds. I guess there will be no or slightly slower results .

for the 4 core 8 thread parts, well depends on the game engine. however it should also be close to 8 core one provided that the game engine can not scale more than 4cores. Still will be lacking some performance due to background operations. the difference can be as low as 10% and in some heavily threaded games %50.

 

[aldaia]

See the thing that's wrong with that statement is Intel isn't selling me 1 core compared to lets say 6 from Amd they are selling me 4 cores and Amd would be selling 6. If games use one core at higher usage compared to the other 5 cores then we might have a issue with the game not getting the same numbers compared to a faster per core processor. Most games are starting to finally use 4 cores at least. Anything over a 1060/480 would bottleneck a I3 and you won't be getting the full usage out of your GPU

I don't get why some people in this thread are trying to convince me that something more expensive and with less performance (for my needs) is better, My top priority is not gaming, gaming is secondary, i never spend more than $150-180 on a GPU, any CPU is good enough to get full usage of that. On the other hand I compile a lot, and that is an area where Ryzen shines. I also run simulations that in all aspects are perfectly parallel with zero data communication. I can even spawn the exact number of threads and fix them to specific CPUs to avoid thread migration (simulations run for several days). So my expectations are to do at least as good as MT Cinebench if not better. Why the hell should I buy an i7 7700K for $350 when a $250 R5 1600X will give me more performance where it matters and exactly the same while gaming? And if I can spare $350, what I should buy is R7 1700.

No one is trying to convince you i said several times that Ryzen is superior to Intel in everything but gaming for the money.

In your case Ryzen is amazing in almost all cases but gaming Ryzen is at least not bad. Once i get more $$$ a 1700 is becoming my rig that handles my handbrake machine and some other tasks.

If you are doing work like you said i'd never recommend a I5 over a 6 core Ryzen part, if anything i'd try and tell you to get a 1700 at least for your work try and fit it in the budget

Lame fact about me, i'm having more fun from BOTW on my switch then i have been on my PC for 2 years. Power is not everything(also why i'm not as active on the forums).

Sorry If I misunderstood you.

I agree that 1700 is probably the better choice for my needs. Though, I think that, the 6 core parts are the sweet spot in terms of performance/$. 1600X should be about 75% the performance of 1800X (maybe more since L3 cache size is the same) for half the price. 1600 should also be 75% the performance of 1700 for 66% the price.

4 core parts don't seem to offer same level of performance/$ 1500X is aprox half 1700X but price is also half of it. 1400 seems to be 50% of a 1700 but for more than 50% price.

I do not think there will be so much difference in games. As there are not a lot of title nowadays that can use 8 cores 16 thread CPU in full range. A 6 core 12 thread CPU would nearly be as good as a 8 core 16 thread cpu at the same clock speeds. I guess there will be no or slightly slower results .

for the 4 core 8 thread parts, well depends on the game engine. however it should also be close to 8 core one provided that the game engine can not scale more than 4cores. Still will be lacking some performance due to background operations. the difference can be as low as 10% and in some heavily threaded games %50.

I was not talking about games. I was talking about my particular needs. If you read the full conversation you quoted you will see that "I compile a lot, and that is an area where Ryzen shines. I also run simulations that in all aspects are perfectly parallel with zero data communication."

 

[juanrga]

@Juan, I do find this hysteria about ryzens 'bad' gaming performance slightly confusing.

The only areas where I think there are a problem are the handful of titles that drop performance with smt on (but don't do the same on Intel), and the couple of well threaded titles that simply don't scale on ryzen but do on the 6900k (e.g. AOTS). I think those are edge cases, but still things that are likely to get fixed down the line (if not in existing titles I at least expect the behavior to become non existent in the future).

Still, a few edge cases isn't the end of the world. Generally ryzen is pretty strong- latest tests on mass effect Andromeda for example put ryzen just behind the 7700k, which is perfectly acceptable imo.

Here's what happened:

A few months prior to release: Expectations: Haswell
A few days prior to release: Expectations: Better then Kaby Lake
Release: Actual Performance: Haswell
Expectations: Missed

Hype is a double edged sword.

Right, that is the resume of what happened. The details were much more hilarious!

When reviews showed that RyZen doesn't match/beat 7700k on games, that same people, let me call them hysteric-people, didn't admit their mistake, but invented crazy excuses, and attacked to anyone that didn't say what they wanted to heard.

The first excuse was that reviews were biased. Some of that hysteric-people even send death threads to reviewers

https://twitter.com/GamersNexus/status/838221363991166981

When more reviews found the same, the second excuse was motherboard fault. It was a fault with Asus mobos, they said. And promised us that a BIOS update to be releases within days was going to fix gaming.

But reviewers found the same findings with non-Asus mobos and the BIOS update did never came. The third excuse from that hysteric-people was SMT. They said that AMD implementation of SMT was new and affecting games. They promised us a SMT fix coming soon.

Reviews (hardware.fr TechSpot,..) tested SMT to check it their clam was true or not (TechSpot even mentions that AT forum thread we are mentioning here)

http://www.techspot.com/review/1348-amd-ryzen-gaming-performance/

The conclusion was that SMT is working fine with only 1% (min) or 3% (avg) changes in performance over 16 titles.

Then the hysteric-people changed the excuse again. The problem now was the CCX-CCX latency and the windows scheduler bouncing threads between cores in different CCX. This was the new excuse why RyZen wasn't beating Kabylake. They claimed that the problem was on the W10 scheduler and pretended that Microsoft was working in a fix.

Again reviews tested this issue (PcPER, computerbase.de, Hardware.fr). The conclusion was that CCX-CCX latency affects one or two titles and nothing to the rest (less than 5% changes on average). It was also demonstrated that the W10 scheduler is working fine and that a new scheduler was not going to change anything substantially

https://www.pcper.com/reviews/Processors/AMD-Ryzen-and-Windows-10-Scheduler-No-Silver-Bullet

If you check the comment section you can find hysteric-people denying facts, and accusing PcPER of being paid by Intel: "You're literally showing to everyone around the world your true face, an ugly mug of a completely biased and undoubtedly bribed Intel shekel mongler." PCPer had to defended from unfair attacks with "We've been communicating with AMD throughout our testing process and they agree with our results."

A pair of days after AMD published an official communicate, where they confirmed reviewers' analysis and admitted that W10 scheduler works fine for RyZen and that SMT also works fine in general for games

https://community.amd.com/community/gaming/blog/2017/03/13/amd-ryzen-community-update?sf62107357=1

Guess what? That hysteric-people didn't admit the truth, didn't apologize for insults and attacks to reviews but continued with their hype agenda. PcPer even had fun with that and wrote a sarcastic page (check comments as well), which latter they deleted, but it is archived here

http://web.archive.org/web/20170313223822/https://www.pcper.com/news/General-Tech/AMD-Running-out-Intel-Sheckels-Renews-Contract-Defame-Own-Products

If you check the AT "technical" thread, you can find hysteric-people saying that the W10 scheduler issue is real, and that AMD is not admitting it officially because... "Wintel" (sic) is obligating AMD to deny the issue.

It is all comedy gold. You can find now to people as OrangeKrush pretending now that current RyZen chips are rebranded engineering samples and that the real RyZen with everything fixed is coming soon

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-29#post-38793748

And if you check the last pages of that weird AT thread, you can find people pretending again that Microsoft is releasing a magic patch for RyZen and that Ryzen gaming performance increases by up to 20% (some pretend it is more than 40%) with the Windows fix.

 

[Rookie_MIB]

I'm in the process of getting everything re-installed and sorted out on my rig.

Ryzen 7 1700
8GB (2x4GB) Corsair Vengeance LPX
Asus B350M-A motherboard
GTX960
Samsung 830 SSD
Win7 64b

It's been a little bit of a battle getting Win7 installed, but it's finally up and running. I had some issues where the installer wasn't recognizing the SATA drives - eventually it did start (I had tried to put it in RAID mode as I have twin 830's in RAID 0 for the boot drive - that might have been the issue) got everything finally installed (fortunately the mobo has PS2 ports for keyboard/mouse!).

Preliminary geekbench 3 with just everything at stock: Single core: 3481, Multi-core 24274.

I haven't had a chance to play with the memory (probably still running at 2133mhz) but I'll get around to some more updates once I get the system sorted out and my software side of things running (it will host my 12TB backup system...)

 

[jdwii]

Ok, that is fair to say; I also saw a lot of that behavior, but then again we all know those comments are flame-bait. No need to take them seriously, right?

That is why I'm having issues seeing where AMD made things worse this time around; the marketing (for better or for worse) was at the same level as the rest of the companies out there putting their product in a good light, but never once I saw an over-promise from them.

Cheers!

Quite frankly i'm a little happy that Amd did clarify that Windows 10 is handling Ryzen just fine. I suspect games will have to be programmed to take care of two quad cores in a system instead of 8 cores for Ryzen to start scaling better with next gen games.

I was more disappointed in the fans then Amd but i was a little disappointed with them to for saying we should include benchmarks that have the game GPU bound. What would Amd fans say if Intel did that and then blow off the most used resolution people play on in comments? I even showed a comment where someone in the horror of youtube comments wanted Jayztwocents to test in the least CPU bound case for gaming cause they claimed he is trying to make ryzen look bad lol.

 

[-Fran-]

Well, personally, I still think the scheduler can be tweaked, but I do agree the improvements won't be magic bullets for performance and won't make Ryzen leapfrog any Intel CPU it hasn't already beaten. Still, a 2-3 percent change in favor is still good and might even be quantifiable in day to day stuff, since the scheduler is the main responsible for having a snappy/responsive OS. I do remember the first days of the P4+HT, when the experience in the desktop was improved a lot after the schedulers were able to handle the extra thread correctly. Well, maybe nowadays it won't be as pronounced, but still.

I'm still waiting on AnandTech's part 2. I wonder if they'll put some new stuff into the light-spot.

Cheers!

 

[Rookie_MIB]

Well, personally, I still think the scheduler can be tweaked, but I do agree the improvements won't be magic bullets for performance and won't make Ryzen leapfrog any Intel CPU it hasn't already beaten. Still, a 2-3 percent change in favor is still good and might even be quantifiable in day to day stuff, since the scheduler is the main responsible for having a snappy/responsive OS. I do remember the first days of the P4+HT, when the experience in the desktop was improved a lot after the schedulers were able to handle the extra thread correctly. Well, maybe nowadays it won't be as pronounced, but still.

I'm still waiting on AnandTech's part 2. I wonder if they'll put some new stuff into the light-spot.

Cheers!

I think you'd be surprised what can be achieved through incremental updates. A while back when the FX-8000 series released, all the benchmarks had it about 10-15% behind the Sandy Bridge i5's. Next year, it was only about 5% behind. Fast forward two years, it's 5-10% ahead in the same game against the same SB i5. One of the major YT reviewers dug up that info.

I'd like to see the review pt 2 as well.

 

[vvacenovski]

Quick question: Considering benchmarks - Out of all R7's, best bang for the buck as a workstation (3ds Max, Vray), with secondary role in gaming and streaming, is the R7 1700 with aftermarket cooler? Am I wrong?

 

[-Fran-]

Quick question: Considering benchmarks - Out of all R7's, best bang for the buck as a workstation (3ds Max, Vray), with secondary role in gaming and streaming, is the R7 1700 with aftermarket cooler? Am I wrong?

Until the 1600X is not [strike]revealed[/strike] benchmarked, I'd say yes. If you care more about the extra 2C/4T the 1700 has, then even if the 1600X comes, then I'd say the 1700 will still be better for well threaded workloads.

Cheers!

EDIT: Self nitpick.

 

[Nope 1151]

Is anyone willing to benchmark a X6 Phenom against a 1600?