Jump to content
 
Sign in to follow this  
Malaphax

AMD Ryzen 3 (Matisse)

Recommended Posts

I think you're misunderstanding my point.  They're having to remove existing features from x370 boards in order to support new 3000 series processors, they ran into limitations on the bios size.  I've heard certain raid settings and older processor support was removed in the new firmware updates.  I don't see this as a problem, mostly because you can choose to run the old firmware and still use those features, but it's definitely something to be aware of. 

 

Share this post


Link to post
Share on other sites

This is a really useful article talking about the differences in Precision Boost, Precision Boost Overdrive, and auto-OC. 

tl;dr AMD and the mobo AIB use multiple variables to determine when the cpu should boost/overclock - biggest one being thermals.  While this is by far more efficient from a power/performance perspective it's also a bit inconsistent.
In general it means that manually overclocking provides very limited performance advantages versus using solid cooling and letting the CPU figure it out on it's own. 

https://www.gamersnexus.net/guides/3491-explaining-precision-boost-overdrive-benchmarks-auto-oc

 

 

Share this post


Link to post
Share on other sites

The 3900X generally loses vs a 9900K in gaming performance.  But you can get cheeky by disabling SMT (hyperthreading) and then... well you actually end up with the 3900X winning.  As a note, you can do the same thing on a 9900K and also get better performance on most games, although at that point you're better off buying a 9700K, which is what most reviewers recommend for pure gaming performance anyways. 
Current guess is that when SMT is disabled the cores generate lower thermals and thus are able to boost higher.  I suppose you could also see an increase in performance from just using better cooling - which you can see in this article that tests just that. 

https://www.gamersnexus.net/news-pc/3492-ryzen-cpu-thermals-matter-coolers-and-cases

amd-ryzen-3900x_cold-scale_all.png

Some people are suggesting that you can just run SMT off and reboot into bios and switch it on if you need to do any productivity stuff like rendering files.  That's sorta silly but if you're on a fast SSD/nvme boot drive I guess it's an option. 

 

Share this post


Link to post
Share on other sites
28 minutes ago, Jedi2155 said:

https://www.techspot.com/review/1877-core-i9-9900k-vs-ryzen-9-3900x/

Best summary of gaming performance across a wide range of games.

Stock

 

Overclocked

 

I understand why they are used in reviews but, for me personally, I have become less and less interested in gaming benchmarks in CPU reviews being run at 1080p. I have been gaming at 1440p for years and I'm pretty sure that applies to everyone else in our group too. I am far more likely to be GPU limited in the games I play which ends up with graphs that look more like this (see below).

image.png

Share this post


Link to post
Share on other sites

I'm still CPU limited because I almost always run multiple applications at the same time as gaming. I almost never play just a single game in Full Screen. Its why I'm still mostly focused on Ryzen compared to Intel because MOAR CORES.

Share this post


Link to post
Share on other sites

I'm a little surprised siliconlottery.com did not have much luck beyond 4.2Ghz all-core for multiple models. I would have expected the wall to be more at 4.5-4.6Ghz considering that some of the models have that as their precision boost overdrive frequency.

https://siliconlottery.com/collections/matisse

I guess you could say it would be reasonable to expect everything to be capable of 4.0Ghz all-core.

Share this post


Link to post
Share on other sites

It must be part of the CCX binning that was mentioned previously somewhere where the current chips have 1 good, and 1 okay CCX, while the later 3950X might have 2 good CCX controllers.

Share this post


Link to post
Share on other sites
31 minutes ago, Jedi2155 said:

It must be part of the CCX binning that was mentioned previously somewhere where the current chips have 1 good, and 1 okay CCX, while the later 3950X might have 2 good CCX controllers.

That still does not line up with the fact that siliconlottery's currently highest listed 3900X is @ 4.20GHz @ 1.25Vcore. There are a few reviews that saw the 3900X boost up to 4.3Ghz out of the box with precision boost overdrive with all cores loaded. Anandtech got their 3900X overclocked @ 4.3GHz @ 1.35Vcore all-core.

Share this post


Link to post
Share on other sites
On 7/19/2019 at 12:27 PM, Malaphax said:

The current overclocking strategy is actually to overclock one CCX, and leave the other CCX near stock.  I find this hilarious because that's the same strategy many phone SOCs used with the big/little style of cores.  While there is no current proof of this just yet, some people are suggesting that the 3900x has one "great" CCX and one "decent" CCX which is why all core overclocks aren't great, but single CCX overclocks seem to be promising.  Since AMD themselves mentioned binning the 3950X, there's a possibility that it will have 2 "great" CCXs and might hold up better under an all core overclock - I'll believe it when I see it.  Even then, I'm fairly certain that will only matter to people trying for benchmark records with LNO2. 

Feel free to read more on that here: https://www.reddit.com/r/Amd/comments/cefwjg/ty_der8auer_for_per_ccx_oc_recommend_massive/

Basically the 3000 series doesn't actually like all core overclocking, because most of the time at least 1 ccx is holding back the rest.  Also the boost clocks are really only ever for 1-2cores - the boost speeds heavily scale based on the number of cores and cooling available, it's not quite as simple as saying "overclock everything to the boost clock" on the 3000 series like it was on previous zen chips.  Per CCX overclocking looks super interesting and Der8auer has done a really good job showing how that works.  Unfortunately there's no way to handle this at the bios/startup level yet, but I suspect if this gets big enough some of the mobo AIBs might try and put these options in later BIOS updates.  I'm not sure if silicon lottery is trying to maintain lower core voltages or if they really did have a bad batch.  I know some people were joking on the amd subreddit that silicon lottery would have to retest/rebin all their chips on a ccx level once the derbauer stuff came out - they might very well be doing that. 

Sidenote derbauer did rip into AMD's marketing specifically the video where the head of marketing was talking about 4.7ghz boost, because he feels that those numbers are bullshit.  I do think AMD does a better job of marketing than intel, but sometimes they overpromise and then people rightfully call them on their bullshit.  The 3000 series is still really good, but it doesn't overclock well on all cores, and it requires some new overclocking methods to get the best out of it.  The good news is that stock it performs quite well. 

Share this post


Link to post
Share on other sites

That's not the point. With Zen1, yes, once you loaded all cores you were sitting at base clock even if you were using boost. With Zen1+ and Zen2 though, if you have the overhead, it will boost up above base clock even with all cores loaded. A number of reviews noted this behavior. That's what is bothering me. How did multiple reviews see a 3900x hit 4.3Ghz with just PBO and all cores loaded but SiliconLottery was not even able to get one 3900x to hit that?

Share this post


Link to post
Share on other sites

Here's a very detailed DRAM calculator to be used on all Zen processors.  This user has produced similar content before. 

https://www.overclock.net/forum/28064156-post4887.html

There was a nasty issue that prevented ryzen 3000 series from working with Destiny 2.  AMD has released a beta chipset driver that appears to work.  They'll post more news tomorrow morning on reddit. 
https://www.reddit.com/r/Amd/comments/cj4cz6/rejoice_guardians_destiny_2_now_works_on_ryzen/
The fault has to do with RDRAND call which uses a specific hardware based random number generator.  This code is typically used for cryptography, but Destiny 2 directly uses it and that's why it wouldn't launch on 3000 series. 
As a note, Bungo fucked up.  This code isn't really designed for simple random generation for games, it's slow as hell and a game doesn't need that level of complexity for it's random generators.  Seems like smart people on the internet are wondering just why Bungo is uses weird spaghetti code in Destiny 2.

Share this post


Link to post
Share on other sites

This is a simplified video regarding RAM speed/timings affecting ryzen 3000 series performance:

And this is a random dude on reddit showing some different testing and suggesting tighter timings can provide major performance uplift.

https://www.reddit.com/r/Amd/comments/cmcbc3/ocing_micron_edie_aka_subtimings_matter_a_lot_for/

I still think people are figuring out the best overclocking options for ryzen 3000 series.  As far as I'm concerned, I plan on running my PC close to stock initially, then exploring some options once more testing is done and people develop better recommendations.  It looks like 3200-CAS14 is very high performance (I bought a kit like that because I suspected it would perform well).  And using Precision Boost (not overdrive) still provides additional performance if the thermal headroom is there.  Once you start delving into 3600mhz with manual tight timings and CCX specific overclocking - you'll see additional performance gains but you're working very hard for that last 5-10%. 

Sidenote: There's the Pareto Principle (80/20 rule) which states that you get 80% of the benefit from the first 20% of the effort.  I tend to view that as a decent guideline for PC performance.  I think it's great that people on the internet will try and eke out that last bit of performance through really crazy overclocks and neat tweaks - but I'm a bit more interested in some easy, and more importantly, stable results. 

Share this post


Link to post
Share on other sites

Keep us updated on your ram experiences. I went with the E-Die set while yours is likely B-Die which to simplify is Rank 2 (2x4GB ) vs Rank 1 (1x8GB) densities. E-Dies have performance that closely matches B-Die but of course B-Die is best. Supposedly my ram might do 3600 at CL16 but time will tell,  This guy showed some serious performance boosts (at 1080p cpu limited resolutions of course) with manual timings.

image.png

BTW my CPU batch is 1930 (so week 30). Someone rumored that the newer batches are lower binned 3900x but I'm doubtful of that rumor. 
I've already put some of my system info in here for comparison purposes but I'll be following up once I get a chance.

https://docs.google.com/spreadsheets/d/1c3RTF_ZPjep-Zfimgoca2Ef1gSjZM0rSHVWLkknbfUI/edit#gid=0


I had some hardware swapping I had to do between my laptop (nVME 960 Evo) and desktop (SATA 860 Evo + 840 Evo) and I'm waiting on a new AM4 bracket for my H105 + some Grizzly Kryonaut coming in Friday. I have some spare conductnaut but I'm going to avoid liquid metal for now on the desktop.

My entire 2019 build is a simple update of parts from my 2014 machine:

CPU (3900X) + Motherboard (Aorus Ultra) + RAM (Ballistix DDR4 3200 CL16) ~ $875 + Tax = $950.

Share this post


Link to post
Share on other sites

New Agesa versions are being released by motherboard vendors - the beta versions are already out and should pass testing sometime this week.  This version fixes some of the lower boost clocks.  There was a minor kerfuffle when DerBaur posted a video with a survey of Ryzen 3000 owners that showed most people weren't hitting the advertised boost clocks.  Of course Intel then decided to take the results from this survey and put it in their marketing material... along with some other sleazy/misleading bullshit. 

This latest Agesa version appears to fix the "low" boost clocks and some minor thermal performance.  Most people are seeing between 25-75hz improvements on single core boost, which equates to single core performance of +2%, I'd consider that within the margin of error but w/e. 

Share this post


Link to post
Share on other sites

I'm still more annoyed by the constant ramp up/ramp down of the PCH fan but I haven't had time to really investigate the issue. I only get a few hrs a day to go on my computer and most of that is email chores + warship dailies. I will say that I saw my CPU hit 80C+ (i'm using Grizzly Kryonaut) and I'm not sure which PCH temperature i should use in HWInfo (there's 2 reported with one being about 15-20C higher than the other).

I'm on the F4a BIOS but will update next chance I get to the F6i? or something.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
Sign in to follow this  

×
×
  • Create New...