AMD's Ryzen CPUs (Ryzen/TR/Epyc) & Vega/Polaris/Navi GPUs

I would wait a bit and see how things go. But I expect you have read you up on this due you are in front for buying.

 

Gigabyte claims that their Aorus X370 can now use DDR4 3600 RAM, RAM update for ASUS Crosshair incoming.

The Aorus has better gaming performance than the Crosshair due to more updated BIOS too by the way.

 

Tiny Tom Logan review. Nice and in depth. He got good results with a 4.0ghz OC

(time stamp link to results)

 

@ajc9988 As chew* states, if you are thinking of a pure overclock, then perhaps the R5 (low leak) Ryzen should be the focus. I suspect there to be more headroom for higher frequencies with those. I would much rather see you with a Ryzen, with your knowledge, and maybe wait a month or two for some bios updates, than with a Skylake-X. Just my .02 cents.

chew* states Lower leakage = better air water. Higher leakage = extreme cooling

 

"Binned Ryzen CPUs coming soon to Silicon Lottery!" -
https://siliconlottery.com/collections/pga-1331

AMD 1800X Boxed Processor (ETA March 6th)
Sold Out
Click here to be notified by email when AMD 1800X Boxed Processor (ETA March 6th) becomes available.
https://siliconlottery.com/collections/frontpage/products/1800x

AMD 1700X Boxed Processor (ETA March 6th)
Sold Out
Click here to be notified by email when AMD 1700X Boxed Processor (ETA March 6th) becomes available.
https://siliconlottery.com/collections/frontpage/products/1700x

AMD 1700 Boxed Processor with Heatsink (ETA March 6th)
Sold Out
Click here to be notified by email when AMD 1700 Boxed Processor with Heatsink (ETA March 6th) becomes available.
https://siliconlottery.com/collections/frontpage/products/1700

 

I really don't like this kind of service, if is it a lottery you cant buy the winner "thing" and I also Know that the best binned is already preordered by their loyal customer.
And the ****ty cpu left is already sell at minor price because an open box? no.. So dont like neither in this case.
I hope that will be closed soon or later, and please stop makin' free ads for them

 

You clearly have many misconceptions about how the SiliconLottery service works.

Spend a bit of time getting to know more about how they do business before passing such harsh judgement on a perfectly awesome idea.

The answers are in their forums, and you can always call them up, email them, or post your own questions to them directly.

There is nothing nefarious, unfair, or unseemly about what they do, and your imagination appears to be taking over from common sense or actual informed experience.

I think it's awesome that Silicon Lottery are going to make the same service available for the new AMD CPU's.

 

I don't do business with SL, I don't work for them, they don't pay me in any form directly or indirectly.

My interest in SL is simple admiration for their enthusiast oriented niche market service that was desired and discussed by many of us for years, and I'm thrilled to see such a service now available for *anyone* to participate without a huge personal investment.

In the past anyone interested in finding the highest performance parts had to do it themselves - or get together as a group - to purchase a lot of parts on their own, test it themselves, and many individuals and companies have done this for many years.

Your questions about how SL does what they do is off topic here, and you can post and ask questions directly of SL, I'll let them defend themselves with their detailed answers to your questions.

And, I will add that I hope SL have a long and successful enterprise, and with their success that interest increases enough that other entrepreneurs can enter the market with their offerings to provide alternatives to SL.

 

I wonder how high of a frequency they can hit though, most seem to be topping out about 3.9-4.1 Ghz and 1 4.3Ghz under water

Could be interesting in what they can find out in the wild...

 

Exactly, since SL are motivated to make a business of binning AMD CPU's like they do with Intel CPU's, SL will be incentivised to figure out how to do that, benefiting the community by sharing their findings - which is one of the awesome things SL does.

If you look at the Intel SKU's, SL publishes the % of CPU's they find at each bin level from their last binning run, which gives us all a good idea what the odds are to get a OC CPU from a random purchase.

For example:

Intel 7700K @ 5.2GHz Boxed Processor
https://siliconlottery.com/collections/all/products/7700k52g
As of 2/22/17, the top 7% of tested 7700Ks were able to hit 5.2GHz or greater.

So if you only need a 7700K that can OC to 5ghz, your odd's of getting one are much higher, and you could try to get it by buying a random sample:

Intel 7700K @ 5.0GHz Boxed Processor
https://siliconlottery.com/collections/all/products/7700k50g
As of 2/22/17, the top 59% of tested 7700Ks were able to hit 5.0GHz or greater.

In the past I've recommended not spending extra buying a SL binned CPU when the % is high enough to indicate you have a good chance getting a CPU OC that fit's your need / desires from a random sample.

It will be very interesting to see the results of SL binning of AMD's CPU's, and what the %'s are for each binning SKU.

In the case of AMD CPU's, it may be SL won't be able to make a business out of it due to the limited range of OC available at a safe voltage.

SL may find they need to advocate to customers that water or other extreme cooling methods and higher than recommended voltages be used to attain the higher binned performance.

All good info for the benefit of all of us, even if we aren't customers of SL

 

I spent hours reading on reviews, user experiences, etc. yesterday. Some have even suggested that the problem is that old bioses were used (more than Gront below, some reviewers tested before the Feb. 28th event, meaning they didn't try the bios updates on the day of release, while others outright said they could not obtain beta bioses from the companies at the time of writing, but that they would update when that became available). So it is mainly to see what pans out soon, including, hopefully, some really nice boards. Also, the 5.2 is not the top for LN2, as Der8aur has done some benches at over 5.4 and got the top frequency to 5.8. Considering how the scores stack against HW-E and BW-E 8-cores, where you get a better score than those with a lower multiplier, I'd say the silicon lottery is at play (also going on the voltage variance in reviews).

http://hwbot.org/hardware/processor/ryzen_7_1800x/

Firmware is forthcoming to hone all boards, so don't put too much weight into any board being better than another at the moment. That is going to be in flux for awhile (especially depending on the ram update which is expected upwards of two months from now).

Actual numbers in the benchmarks matter more to me than raw multiplier/bclk frequency achieved. I learned the OC basics on AMD, so I've got a soft spot for them and want them to do well. I've seen the numbers and Ryzen looks fun to play with (more so after more options are opened up in the months to come). Now, I'm planning sub-ambient with my chiller (5C is the bottom limit on the 5,000 BTU beast I have until I figure out how to take the limiter off of it), and maybe delving into DICE and LN2 with my next platform in the future. This is why I'll be looking for binned chips and keeping my options open.

I did read Chew's write-up and saw the voltage leakage part. I will be wanting to see how the binning goes before making a decision, as well as what firmware updates change and if higher quality boards are put out. I also estimate, because skylake had a couple hundred MHz on haswell chips, that you will see 100-200MHz on skylake over the BW-E avg air OC, but with slightly higher IPC. That being the case, it will be on top until a Zen refresh either slides in right behind it. Otherwise, you have to hope Zen+ comes in with the IPC gains and refinement of the SMT to better utilize what is out there. (if they also do a die shrink with it, then cannonlake is in trouble)!

It's more a timing thing than anything else.

I break down that same analysis and find that buying SL second highest binning is usually the sweet spot. Same with Caseking (I believe they are the Euro binners). Otherwise I scour HWBot forums for the sales posted by the top OCers selling the chips they don't want. Those numbers are a larger influence on what chip I get overall....

 

@ aaronne
People who don't want to bin their chips on their own time and dime like Silicon Lottery. What's not to like? It's a fair service. Like any other business in this world, the owner is free to sell the best to his friends for mates rates, hardly something exclusive to them.

And no.. I can assure you @hmscott doesn't work for them...he went BGA

Spoiler: Spoiler

 

@Raiderman - tell chew that to get around the core 0 dropped on SMT diisabled, you have to start up the OC app inside of windows and raise the core 0 to match the overclock of the rest. It has been reported and they are working on a bios fix for it. I can't find the article I read that in last night, but it is a common problem.

Edit: nevermind, read it from his post last night. I really need to wake up this morning and organize my sources for what I've seen to date... Ignore my post...

 

@ajc9988, I think he knows this. He has posted this bug in his second post outlining these that he has come across.


SMT disabling bug 1700 model. when disabling SMT and posting you lose overclock. Don't panic click the cores tab and you will see all cores but 0 are still at there overclocked speed. Open ryzen master. clock core 0 up. Problem reported and pending bios update.

 

So, reposting a couple posts from OCN:

Undervolter -

Quote:
Originally Posted by IRobot23

So far-
- Ryzen is worst AMD overclocking
- RYZEN beat all expectations (ST,IPC,MT, high clocks, cost)
- IMC is very efficient
- There is no DDR4 that would run stable in CR1 mode at higher clockspeed.
- MB have hard time, not full support
- Games,windows, apps need optimizations for RYZEN.

Anything else?

I would like to know why people are bringing CCX unit as a botttleneck in games?
Whats the problem to have 8MB of L3 shared between only 4 cores? Whats could cause the bottleneck?

As far as I know CCX0-> Cores wont use data stored in CCX1-L3$.

Can someone please explain?


Good summary, but the main RAM problem, will probably remain the intolerance for high speed dual rank RAM. There is probably a new wave of DDR4 tested for Ryzen underway:

G.SKILL Announces Flare X Series and FORTIS Series DDR4 Memory for AMD Ryzen
https://www.techpowerup.com/231204/...s-and-fortis-series-ddr4-memory-for-amd-ryzen


^ Photo shows DDR4 3466 T1.


IMHO: Ryzen overdelivered, BUT, the release was premature. Here, finding a motherboard is near impossible. In a price finding engine that has virtually all the known e-shops in the country, you enter " AM4 motherboard" and you get 10 results with AM3+ motherboards and 3 ASUS B350M.

Then, you have Ryzen having problems with:
- SMT, HPET on, balanced power profile in Windows.
- RAM frequency and ranks.
- motherboard BIOSes that some reviewers said made major difference are released one day before or on the day of the official launch.
- Reviewers keep having serious bugs.
- AMD says that sooner or later they will release a driver that will solve the balanced power profile and core unparking issue.

NOW you have RAM companies that come out to say "oh, now we will release RAM that works well with Ryzen".

All this, sounds like there was haste to make the launch, but the time wasn't enough. Imagine, if Ryzen were launched, with all BIOS related issues that affect performance fixed, with perfect support for current RAM kits, with AMD's Windows driver ready.

It would have been much better. Once a review is written, it's over. First impressions are important.Gamers only see FPS.

EDIT: About CCX, read this:

http://www.overclock.net/t/1624566/...-recommendations-for-the-future#post_25890150

Crazy elf-

Because it's a big barrier.

Communication between the 2 CCXs is very slow - it's more like have 2 CPUs with 4 cores than 1 CPU with 8 cores (like on Haswell E). DDR4 is essentially the last level cache. The Infinity Fabric (that's what AMD calls the interconnect) is only 22GB/s.

Right now Windows (and I assume Linux) is treating the CCX like 1 CPU, which means that data is being moved between the 2 CCX at a big performance penalty since it has to hit the DRAM. We need a patch that will treat them as if we have a 2P socket.



Memory overclocks therefore might also have a bigger impact on Ryzen than Intel because it acts like a Last Level Cache, making the poor RAM speeds a bigger problem than other architectures (when it was first released, due to the slow speeds and loose timings, the 6700K did worse in many games than the 4790K, despite Haswell being slightly slower than Skylake). The workstation benchmarks prove that this CPU has a lot of potential, but it is being held back by the RAM, the slow inter-CCX connection, and perhaps the cache (there's no uncore/core separation).

Cool fact: the 4 core APUs and CPUs won't have this problem because they have 1 CCX.
Edited by CrazyElf - Today at 10:01 am

I take this very differently than others. Cannard PC got a developers board and chip back in last July or August, then leaked benchmarks in December. The board manufacturers likely had boards being developed then. Ram manufacturers only now are saying they will release different lines. Windows does not operate to fully utilize the chips, causing numerous issues, which means they did not spend the time to make the drivers work well. If you add this all up, ALL related companies thought Ryzen would not make the impact it did and blew off AMD to varying degrees. By doing so, they ignored what matters, consumers. Now that consumers have spoken, you have some reporters saying AMD didn't have enough supply. They shipped 1M chips on launch day and they are sold out just about everywhere. I'd say that trying to blame them on supply when everyone thought the demand would be low in a down PC market is to ignore the lack of support by all other parties. This was not a forced release, other than adjacent companies not giving the support or attention they would give Intel. As such, the platform should evolve and mature as time moves forward.

 

Thanks @TBoneSan, good one, what do *I* need LGA SL service for?

I'm happy saving money, enjoying low thermals, low fan noise, and high performance *without* LGA

 

Edit: BTW, notice MB manufacturers claimed they only had 5-6 weeks to develop the firmware. How long does it take to design a board and put it into mass production? Look at the timeline and access to early silicon for these companies. Do you really believe it is AMD's fault that they did not have the time to nail out firmware? Or do you think they, instead, blew off actually putting out a product that works?

Sent from my SM-G900P using Tapatalk

 

Thanks for linking that video. Great to have everything in perspective.

 

Not AMD's fault, I think OEM's were hesitant to put out their best effort for them. Crappy build quality on the motherboards, terrible bios', and all around giving the middle finger to AMD once again.
Asus, one of the leaders, IMO, should be ashamed, as they seem to be the worst culprit. I used to respect Asus a lot, and would go for one of their boards before anything else. I think Asrock is the new board manufacturer that at least puts out an effort. If it is not associated with Intel, then deliver less than adequate products. I think most board manufacturers were caught a little off guard at Ryzen. Meaning they just expected another AMD flop of a release. It will get bettter, as I was reading at XS.org, that AMD, and board manufacturers have always been slow starters with newly released platforms. I remember the first thoroughbreds were very hot until some bios revisions came down the pipeline. Deneb was decent at first, but didnt hit its stride until C3's hit the market. Manufacturing process coupled with bios revisions will make Ryzen a much better OCer, and gamer.

 

AMD "Ryzen gets better in games with time" Swedish so use google translate. Or see the links in the article
"We hear people on wanting to see improved 1080p performance and we fully expect that Ryzen performance in 1080p will only get better as developers get more time with “Zen”. We have over 300+ developers now working with "Zen" and several of the developers for Ashes of Singularity and Total Warhammer are actively optimizing now."

 

I'd avoid posts from Undervolter, I catch him many times spreading fud or blasting threads getting his own narrative across.

Ill repost what I found on reddit yesterday that summarizes the issues in a smaller concise package.

NOT MY CONTENT but I thought I would share some thought from Reddit

Causes of poor gaming relative to CPU performance of Ryzen:

1. Windows is load-balancing across CCXes. This means that a thread is being moved around on the CPU - which is normal - so that a single core isn't used more than others. On Ryzen, that needs to happen ONLY within a CCX, otherwise you will incur a massive penalty when that thread no longer finds its data in the caches of the CCX.
   
   
2. SMT hurts single threaded performance due to shared structure. Ryzen statically partitions three structures to support SMT: Micro-op queue (dispatcher), Retirement queue, Store queue. This means that, with SMT enabled, these resources are cut, potentially, in HALF (mind you, these are just queues that impact throughput of a single thread).
   
   
3. Memory latency quirks still not worked out. Gaming can be quite sensitive to memory latency and bandwidth. These issues will be, most likely, remedied with BIOS updates.

Its basically growing pains of a completely new offering in a world where everyone tailors to Intel

Lastly, I think AMD was probably trying to appeal to the developers with its work horse CPU lineup right now while maintaining the interest of the gamer crowd. As we all expect the node to improve and overclock headroom to improve with 4-6 core variants that are due 2H of this year.

 

Quoted from chew* at xtremesystems. He was rebutting a users disappointment in Ryzen performance.



Im not and i will explain why. In order to beat or be competive a reviewer needs to use intel 8/16 or 10/20 chip and board etc etc.

To beat it in ipc you need another chip and board. Pick one you cant have both.

To be best in games another chip

On this platform disable smt you get both in one socket at far less cost. One socket one platform.

So to game you need x chip on intel. But...at the cost of workstation performance. And vice versa

Ryzen is middle of the road happy balance.

 

IDK about you guys, but I don't think I've ever gotten a new to market MB that ran out of the box without downloading an updated BIOS

AMD is always rushing to get things out so they can monetize their offerings as soon as possible - all mfgrs have time sensitive windows of opportunity and none can afford to keep MB's in house until "all the bugs" are worked out.

They'd never ship anything

You gotta stick with what you have and work with the vendor to get updates, keep interacting and feeding back features and bugs you need fixed - research other MB's to find features you'd like to have on your vendors MB too, it's an interactive thing that is part of the fun of being a first adopter.

In fact it's the only thing fun about being a 1st adopter, the rest is a lot of work

 

Im well aware, but in order to have competition people actually have to buy these components and people are painting the picture that its an underwhelming Bulldozer 2.0 over on OCN at least even though its completely asinine.

Its basically all perception and facts were deteriorating pretty quick.

 

There is more than that. Take for example the Windows problem with core parking. I do agree, take everything with a grain of salt, but I had to repost his to repost the later response in the interaction. If I gave only the second part, the first part would not have given context (which was further addressing a four bullet point on what errors are known).

I also refused to adopt the view as my own, instead giving source so that weighting by others can be provided. But, noted....

Sent from my SM-G900P using Tapatalk

 

I understand, I am quick when I see that username.

I may also have problem reading space out text like that, Im too accustomed to my bullet points, its a note taking uni thing and ive been grinding that stone for 6 months now so I apologize.

 

No problem. I was too lazy to put quotes and quotes in quotes (properly formatting) for the post. I do apologize for the laziness. I could over-do it like my legal outlines!

Sent from my SM-G900P using Tapatalk

 

Someone is selling the R7 1800X for $450 on Amazon. A little too good to be true.

 

It would not be surprising, benchers only need it for a short time to get some subs on hwbot and move on

 

The person had like 15+ in stock at one point and they are marked as brand new....

Another red flag is that the delivery date estimation is between five to 20 business days. I may buy one to see where this goes.

 

And it's a fantastic effort from a company that has basically come from nowhere to be so close.

But as the AMD guy admitted on the call to the Steve guy from gamers nexus, Ryzen is 7% down on Kaby Lake IPC, and then there is the clock speed deficit as well. Anything worse (as seen in 1080p benchmarks) is a gap that can be bridged by optimisation.

People forget there are 6 and 4 cores to come. Also that Kaby Lake IPC beats Broadwell-E as well.

Fact remains it's a theme that AMD knows too well... Worse but cheaper. I wish these reviews put together a frames per dollar or bench points per dollar comparison to show just how much Ryzen kicks butt at current prices.

Or pair a Ryzen platform with a BW-E on same budget I.e. do the dollars spent on GPU serve you better for gaming?

I think everything so far is just super rush job and time will tell for things to shake out

 

Let's not forget that it is 7 to 8 % in FPS , not IPC, as it smokes kaby lake in most workstation benches. Anything multi core, Ryzen really shines, and that will only get better. They were discussing FPS, not IPC.

Sent from my SM-G935T using Tapatalk

 

The TechReport actually does this ( link). In fact, they did it for heavily multi-threaded applications in the first graph on the page and for lightly multi-threaded ones (i.e. games) in the second graph. It is clear that if your main usage is something heavily multi-threaded, Ryzen is the CPU for you. For lightly-threaded loads like games, the Core i7-7700K is still better and it's cheap.

Basically, this is a clear shot at Broadwell-E (i.e. the i7-68X0 and i7-69X0) . Those CPUs never made a great deal of sense to begin with, but now they make practically none at all. The Ryzen 7 1800X is at roughly the same level of multi-threaded performance for half the price and it's actually not that far off from them in lightly threaded (because they're Broadwell rather than Kaby Lake and because they don't hit the same clock speeds as the i7-7700K). As a bonus, all of those chips are 140W whereas the 1800X is only 95W.

 

AMD responds to 1080p gaming tests on Ryzen
https://www.pcper.com/news/Processors/AMD-responds-1080p-gaming-tests-Ryzen

"By far one of the most interesting and concerning points about today's launch of the AMD Ryzen processor is gaming results. Many other reviewers have seen similar results to what I published in my article this morning: gaming at 1080p, even at "ultra" image quality settings, in many top games shows a deficit in performance compared to Intel Kaby Lake and Broadwell-E processors.

I shared my testing result with AMD over a week ago, trying to get answers and hoping to find some instant fix (a BIOS setting, a bug in my firmware). As it turns out, that wasn't the case. To be clear, our testing was done on the ASUS Crosshair VI Hero motherboard with the 5704 BIOS and any reports you see claiming that the deficits only existed on ASUS products are incorrect..."

Spoiler: AMDs response

"AMD responded to the issues late last night with the following statement from John Taylor, CVP of Marketing:

“As we presented at Ryzen Tech Day, we are supporting 300+ developer kits with game development studios to optimize current and future game releases for the all-new Ryzen CPU. We are on track for 1000+ developer systems in 2017. For example, Bethesda at GDC yesterday announced its strategic relationship with AMD to optimize for Ryzen CPUs, primarily through Vulkan low-level API optimizations, for a new generation of games, DLC and VR experiences.

Oxide Games also provided a public statement today on the significant performance uplift observed when optimizing for the 8-core, 16-thread Ryzen 7 CPU design – optimizations not yet reflected in Ashes of the Singularity benchmarking. Creative Assembly, developers of the Total War series, made a similar statement today related to upcoming Ryzen optimizations.

CPU benchmarking deficits to the competition in certain games at 1080p resolution can be attributed to the development and optimization of the game uniquely to Intel platforms – until now. Even without optimizations in place, Ryzen delivers high, smooth frame rates on all “CPU-bound” games, as well as overall smooth frame rates and great experiences in GPU-bound gaming and VR. With developers taking advantage of Ryzen architecture and the extra cores and threads, we expect benchmarks to only get better, and enable Ryzen excel at next generation gaming experiences as well.

Game performance will be optimized for Ryzen and continue to improve from at-launch frame rate scores.” John Taylor, AMD"

As long as I was there, I checked out other pcper Ryzen info, this article has some interesting links and reader comments:

Ryzen shine! It is time for your AMD roundup
https://www.pcper.com/news/Processors/Ryzen-shine-It-time-your-AMD-roundup

"March 3, 2017 | 07:39 AM - Posted by Anonymous (not verified)
Oh yeah, 1700 is where it's at.

All the information about OC so far indicates that 1700/1700x/1800x all clock to the same 4.0/4.1GHz and it does not depend on the model.

As long as you have basic know-how of OC, 1700 will be at the same level with the rest at almost 200 less moneys."

"March 3, 2017 | 08:51 AM - Posted by Jarlen (not verified)
1600x is what matters, cheap and allround exceptional for gamers both now and long term and will probably clock 200-300 mhz more than 1700 and 1800 series. 8 cores always overclocks worse than 4-6 cores. look at intel hedt.

if i really need 8c/16t i would take 1700 non-x and oc it, they all reach same clocks pretty much. 1800x is a halo product, waste of money like 6950x"

"March 3, 2017 | 11:34 AM - Posted by StephanS
Not sure any Ryzen will clock higher. But the 1600x will need nimble cooling.

I personally find the R7 1700 gaming result more then needed.
That I get 140FPS vs 170FPS at 1600x900 with a GTX 1080 is irrelevant to me.
I care that I can get 60FPS at 1440p on a 1070 class GPU.

And the 1700 seem to deliver the perfect balance of gaming and raw horsepower."

Benchmarking Ryzen & GPU Bottlenecks

 

It seems that Windows has a problem with Ryzen's threads. Hopefully a Windows update fixes that soon.

From Agent-47 of the Anandtech forums:
"Windows has a bug affecting Ryzen where it allocates both logical and physical cores as if they were physical cores while incorrectly guessing processor cache size. This is a potential cause of SMT performance issues, as the processor may not behave as expected by the windows scheduler when a thread hits a non-physical core that is expected to be the real deal."

 

Probably a windows 10 problem I would laugh my arse off if Windows 7 worked correctly out of the box!!

 

Windows hasnt been updated to account for AMD so yes, its an issue at the moment. Just like Bulldozer and Dual core CPU back in the XP days.

 

I created a meme for those who are arguing with idiots on other tech sites. Feel free to use it

 

.......and here is another for ya!

 

Ryzen is a great architecture for servers but less so for consumers.

The current ryzen 8 cores is horribly unsuitable for my workloads and thus I will continue using intel 8 cores and 10/12 cores in the future but I am very excited for the 32 cores octo channel.

 

That's an interesting use case. Can you specify the applications you are thinking of or are they secret? Off the top of my head, I can't think of a load where Ryzen is noticeably inferior to the 8-core i7s. Is it something to do with memory bandwidth and/or PCIe lanes?

 

Memory bandwidth wise, ryzen is terrible.

There is a bunch of apps that needs faster memory bandwidth I believe. I know CFD applications does a lot.

 

Ya, not suited for consumers, and better for servers?.....haha.....can anyone say imbecile? Buh-Bye! I like this ignore list thing.

 

From chew* @XS

 

My theory is these suffer similar issues that the old core 2 quads had. That CPU was non NUMA and 2 separate actual dies. The old scheduler used to take a non optimized thread and run it between the two cores causing the chip to load the cache twice. This could cause a 10% or greater performance hit to individual non optimized threads. Windows 8.0 on a fresh install fixed the scheduler.

It should be known that for W7 there was a Bulldozer patch that could give 5% of the performance loss back to the old C2Q.

The Ryzen is two silicon chips as seen here.

 

Yeah, I think he misread my post on the ryzen master. I was saying nevermind because the information originated from him, not because it didn't work. What brought it up was him talking of being tired and trying many different BIOS versions. I, too, was tired, hence not realizing my source of that information was him in the second post. A lot of miscommunication there (with me misreading causing the issue to begin with).

As to his abilities, I meant no offense. Thought I was addressing something he hadn't seen, not commenting on ability. I've seen his name floated places before and his reputation and regard with the overclocking community. So, please convey I meant no offense, just a miscommunication.

Edit:

Since I'm waiting for a binned chip, could you ask him to confirm the following info someone posted:


Well It turns out, its windows dault for the low performance!!!

Logical Processor to Cache Map:
*--------------- Data Cache 0, Level 1, 32 KB, Assoc 8, LineSize 64
*--------------- Instruction Cache 0, Level 1, 64 KB, Assoc 4, LineSize 64
*--------------- Unified Cache 0, Level 2, 512 KB, Assoc 8, LineSize 64
*--------------- Unified Cache 1, Level 3, 16 MB, Assoc 16, LineSize 64
-*-------------- Data Cache 1, Level 1, 32 KB, Assoc 8, LineSize 64
-*-------------- Instruction Cache 1, Level 1, 64 KB, Assoc 4, LineSize 64
-*-------------- Unified Cache 2, Level 2, 512 KB, Assoc 8, LineSize 64
-*-------------- Unified Cache 3, Level 3, 16 MB, Assoc 16, LineSize 64
--*------------- Data Cache 2, Level 1, 32 KB, Assoc 8, LineSize 64
--*------------- Instruction Cache 2, Level 1, 64 KB, Assoc 4, LineSize 64
--*------------- Unified Cache 4, Level 2, 512 KB, Assoc 8, LineSize 64
--*------------- Unified Cache 5, Level 3, 16 MB, Assoc 16, LineSize 64
---*------------ Data Cache 3, Level 1, 32 KB, Assoc 8, LineSize 64
---*------------ Instruction Cache 3, Level 1, 64 KB, Assoc 4, LineSize 64
---*------------ Unified Cache 6, Level 2, 512 KB, Assoc 8, LineSize 64
---*------------ Unified Cache 7, Level 3, 16 MB, Assoc 16, LineSize 64
----*----------- Data Cache 4, Level 1, 32 KB, Assoc 8, LineSize 64
----*----------- Instruction Cache 4, Level 1, 64 KB, Assoc 4, LineSize 64
----*----------- Unified Cache 8, Level 2, 512 KB, Assoc 8, LineSize 64
----*----------- Unified Cache 9, Level 3, 16 MB, Assoc 16, LineSize 64
-----*---------- Data Cache 5, Level 1, 32 KB, Assoc 8, LineSize 64
-----*---------- Instruction Cache 5, Level 1, 64 KB, Assoc 4, LineSize 64
-----*---------- Unified Cache 10, Level 2, 512 KB, Assoc 8, LineSize 64
-----*---------- Unified Cache 11, Level 3, 16 MB, Assoc 16, LineSize 64
------*--------- Data Cache 6, Level 1, 32 KB, Assoc 8, LineSize 64
------*--------- Instruction Cache 6, Level 1, 64 KB, Assoc 4, LineSize 64
------*--------- Unified Cache 12, Level 2, 512 KB, Assoc 8, LineSize 64
------*--------- Unified Cache 13, Level 3, 16 MB, Assoc 16, LineSize 64
-------*-------- Data Cache 7, Level 1, 32 KB, Assoc 8, LineSize 64
-------*-------- Instruction Cache 7, Level 1, 64 KB, Assoc 4, LineSize 64
-------*-------- Unified Cache 14, Level 2, 512 KB, Assoc 8, LineSize 64
-------*-------- Unified Cache 15, Level 3, 16 MB, Assoc 16, LineSize 64
--------*------- Data Cache 8, Level 1, 32 KB, Assoc 8, LineSize 64
--------*------- Instruction Cache 8, Level 1, 64 KB, Assoc 4, LineSize 64
--------*------- Unified Cache 16, Level 2, 512 KB, Assoc 8, LineSize 64
--------*------- Unified Cache 17, Level 3, 16 MB, Assoc 16, LineSize 64
---------*------ Data Cache 9, Level 1, 32 KB, Assoc 8, LineSize 64
---------*------ Instruction Cache 9, Level 1, 64 KB, Assoc 4, LineSize 64
---------*------ Unified Cache 18, Level 2, 512 KB, Assoc 8, LineSize 64
---------*------ Unified Cache 19, Level 3, 16 MB, Assoc 16, LineSize 64
----------*----- Data Cache 10, Level 1, 32 KB, Assoc 8, LineSize 64
----------*----- Instruction Cache 10, Level 1, 64 KB, Assoc 4, LineSize 64
----------*----- Unified Cache 20, Level 2, 512 KB, Assoc 8, LineSize 64
----------*----- Unified Cache 21, Level 3, 16 MB, Assoc 16, LineSize 64
-----------*---- Data Cache 11, Level 1, 32 KB, Assoc 8, LineSize 64
-----------*---- Instruction Cache 11, Level 1, 64 KB, Assoc 4, LineSize 64
-----------*---- Unified Cache 22, Level 2, 512 KB, Assoc 8, LineSize 64
-----------*---- Unified Cache 23, Level 3, 16 MB, Assoc 16, LineSize 64
------------*--- Data Cache 12, Level 1, 32 KB, Assoc 8, LineSize 64
------------*--- Instruction Cache 12, Level 1, 64 KB, Assoc 4, LineSize 64
------------*--- Unified Cache 24, Level 2, 512 KB, Assoc 8, LineSize 64
------------*--- Unified Cache 25, Level 3, 16 MB, Assoc 16, LineSize 64
-------------*-- Data Cache 13, Level 1, 32 KB, Assoc 8, LineSize 64
-------------*-- Instruction Cache 13, Level 1, 64 KB, Assoc 4, LineSize 64
-------------*-- Unified Cache 26, Level 2, 512 KB, Assoc 8, LineSize 64
-------------*-- Unified Cache 27, Level 3, 16 MB, Assoc 16, LineSize 64
--------------*- Data Cache 14, Level 1, 32 KB, Assoc 8, LineSize 64
--------------*- Instruction Cache 14, Level 1, 64 KB, Assoc 4, LineSize 64
--------------*- Unified Cache 28, Level 2, 512 KB, Assoc 8, LineSize 64
--------------*- Unified Cache 29, Level 3, 16 MB, Assoc 16, LineSize 64
---------------* Data Cache 15, Level 1, 32 KB, Assoc 8, LineSize 64
---------------* Instruction Cache 15, Level 1, 64 KB, Assoc 4, LineSize 64
---------------* Unified Cache 30, Level 2, 512 KB, Assoc 8, LineSize 64
---------------* Unified Cache 31, Level 3, 16 MB, Assoc 16, LineSize 64

each zen thread is being registered as an individual core with its own L2 and L3 cache. i.e. totaling 136 MB cache!!. this is using Windows Sysinternals. This explains the SMT troubles in the event that a thread bounced to a HT thinking its the real deal.

Compare VS Intel

Logical Processor to Cache Map:
**------ Data Cache 0, Level 1, 32 KB, Assoc 8, LineSize 64
**------ Instruction Cache 0, Level 1, 32 KB, Assoc 8, LineSize 64
**------ Unified Cache 0, Level 2, 256 KB, Assoc 4, LineSize 64
******** Unified Cache 1, Level 3, 6 MB, Assoc 12, LineSize 64
--**---- Data Cache 1, Level 1, 32 KB, Assoc 8, LineSize 64
--**---- Instruction Cache 1, Level 1, 32 KB, Assoc 8, LineSize 64
--**---- Unified Cache 2, Level 2, 256 KB, Assoc 4, LineSize 64
----**-- Data Cache 2, Level 1, 32 KB, Assoc 8, LineSize 64
----**-- Instruction Cache 2, Level 1, 32 KB, Assoc 8, LineSize 64
----**-- Unified Cache 3, Level 2, 256 KB, Assoc 4, LineSize 64
------** Data Cache 3, Level 1, 32 KB, Assoc 8, LineSize 64
------** Instruction Cache 3, Level 1, 32 KB, Assoc 8, LineSize 64
------** Unified Cache 4, Level 2, 256 KB, Assoc 4, LineSize 64

 

You already ran test runs on Ryzen? Do tell

 

No. Its dual channel..... which is less memory bandwidth as quad channel...

 

It would still be worth testing for real data that can be compared and shared.

Otherwise, it's just an educated guess - which I have seen proven wrong a surprisingly high % of the time