I had given up on AMD… until today - Ryzen 9 3900X & Ryzen 7 3700X Review

[Valentyn]

9 minutes ago, TheNaitsyrk said:

I'm a little disappointed. It's previous generation all over again, except slightly better.

 

Still not as good at gaming even though the die is 2x smaller. Clocks still low at 4.3 at the maximum if OCed and OCing hurts performance often. Games that love high clocks is a no no and 9900K still smokes it.

 

I really wanted this CPU to properly match 9900K in gaming, but it's not quite there yet.

Seems there are some BIOS issues afoot, that stopped the CPUs from boosting to their rated Speeds,

Anandtech already saying they're going to rerun rests.

Will be interesting to see if there's any real difference.

[TheNaitsyrk]

6 minutes ago, Valentyn said:

Seems there are some BIOS issues afoot, that stopped the CPUs from boosting to their rated Speeds,

Anandtech already saying they're going to rerun rests.

Will be interesting to see if there's any real difference.

I trust GamersNexus more as they have most scientific test imo. Their tests are apparently fine.

 

BUT...

 

If this is true, then I might reconsider in 2 months if things are sorted out.

[tautvydas]

just been wondering, is all testing done after people noticed that there were problems with bios?
or the scores are with bios holding chips back.

[Zashius]

Id really like to see a comparison between the 1950x, 2950x, and 3950x kinda like an evolution of zen 

[TommyTwoPockets]

I was really excited to watch this, especially to see you getting more screen time Anthony! Anyway, glad to see AMD getting really close. My friend who is looking to upgrade his AMD 3+ finally is excited for these Ryzen chips. Wondering if and how Intel will respond besides price dropping their CPUs. 

[BZP]

Hopefully this question is OK for this topic.
I want to begin doing some work with 1080p video in Premiere. Should I get a Ryzen 3900X setup or get one of these refurbished systems?

$1800 Xeon 2135, Quadro P4000, 32GB RAM, 1TB SATA HDD 7200rpm
$2100 Xeon 4114, Quadro P4000, 16GB RAM, 512GB SATA SSD Class 20

Thanks,
Brian

[Fatbelly]

I have watched your videos on and off for quite some time, thanks for the work put in. I never had any reason to register, but I had a lot of time on my hands recently and I wrote an "article" of complied review oddities for the ryzen 3000 reviews, and posted it on forums i frequent to see if anyone could clear up any of the points.

https://www.reddit.com/r/hardware/comments/ccaote/ryzen_3000_hype_and_review_oddities/

 

AMD actually commented on the part I referenced from your video about huge run variations when affinity was not manually set for the 3900x

 

" There's nothing to optimize against for the cluster or chiplet design. The scheduler is already optimized for the CCXes, and the chiplets introduce no additional complexity from the perspective of a game or scheduler. The topology appears monolithic d/t its implementation. "

 

As this is my mind does not explain the issues LTT experienced and mitigated by manually assigning cores (but rather could be the cause of it? "The topology appears monolithic"), I asked for a followup and will add any reply.

 

Did you re-test this on latest windows bios or possibly another motherboard? If the issue persists it really does not look good. I kind of want to buy a 3900x but i am waiting to decide until this issue is cleared up as it seems pretty major if your results were correct and no optimization is coming (we put a large L3 cache, solved! (?)).

 

The rest is is pretty minor (I do not care about power efficiency nor temperatures as long as it performs as it should).

[Emily Young]

4 hours ago, Fatbelly said:

" There's nothing to optimize against for the cluster or chiplet design. The scheduler is already optimized for the CCXes, and the chiplets introduce no additional complexity from the perspective of a game or scheduler. The topology appears monolithic d/t its implementation. "

This is patently ridiculous for AMD to say. Of course it introduces more complexity for a scheduler - Yeah, it's CCX aware, but if you're in a situation where you've got a choice between two CCXes for your extra game threads and one of them happens to be on another CCD/chiplet from your main thread(s), there's a whole lot more latency involved in crossing that boundary than just the CCX boundary, and from what I've seen anyway, Windows currently doesn't have any awareness of that. All of our testing was done with the most up to date patches and BIOS revision available at the time, including the CCX-aware optimizations in 1903.

 

It IS true that there isn't any additional complexity from the game's perspective... At least not in theory. It's all transparent, after all, hence the "topology appears monolithic" comment, and ironically this is also why CCX scheduling was a problem: If it's not known what CCX a thread is on, then it's not possible to ensure any related threads are also on that CCX. To say that there's no optimizations that can be made to the scheduler to avoid this, at least from my observations, is completely false. Why else would forcing a game to run solely on one CCD improve performance if it's not that it was crossing the CCD boundary before? Now, if the scheduler is supposed to be CCD-aware, then it's currently broken and the "optimization" would necessarily just be fixing it.

 

At any rate, I wouldn't say that wrangling this issue is a showstopper. Ryzen 1st and 2nd gen also had scheduler issues that prevented them from reaching their full potential out of the gate, and that Ryzen 3rd gen runs as well as it does is a good endorsement in my eyes. I just don't know why AMD would be damage controlling this rather than looking into it more closely. I mean, maybe it is our setup that's the problem, but to date nobody from AMD has approached me to talk about our results.

[Fatbelly]

6 hours ago, GabenJr said:

This is patently ridiculous for AMD to say. Of course it introduces more complexity for a scheduler - Yeah, it's CCX aware, but if you're in a situation where you've got a choice between two CCXes for your extra game threads and one of them happens to be on another CCD/chiplet from your main thread(s), there's a whole lot more latency involved in crossing that boundary than just the CCX boundary, and from what I've seen anyway, Windows currently doesn't have any awareness of that. All of our testing was done with the most up to date patches and BIOS revision available at the time, including the CCX-aware optimizations in 1903.

 

It IS true that there isn't any additional complexity from the game's perspective... At least not in theory. It's all transparent, after all, hence the "topology appears monolithic" comment, and ironically this is also why CCX scheduling was a problem: If it's not known what CCX a thread is on, then it's not possible to ensure any related threads are also on that CCX. To say that there's no optimizations that can be made to the scheduler to avoid this, at least from my observations, is completely false. Why else would forcing a game to run solely on one CCD improve performance if it's not that it was crossing the CCD boundary before? Now, if the scheduler is supposed to be CCD-aware, then it's currently broken and the "optimization" would necessarily just be fixing it.

 

At any rate, I wouldn't say that wrangling this issue is a showstopper. Ryzen 1st and 2nd gen also had scheduler issues that prevented them from reaching their full potential out of the gate, and that Ryzen 3rd gen runs as well as it does is a good endorsement in my eyes. I just don't know why AMD would be damage controlling this rather than looking into it more closely. I mean, maybe it is our setup that's the problem, but to date nobody from AMD has approached me to talk about our results.

 

I looked through other reviews that checked 99th percentile minimum fps and they did not publish similarly bad results - so either they ignored outliers and published median/good results, or something could be off with your setup - would be interesting if you retested, maybe with a different motherboard/switch other components such as ram/gpu just in case of some incompatibility? (did you get the same inconsistent results in other games or just battlefield btw?)

(But assuming your result was correct I guess the answer can be interpreted as "we slapped on a huge L3 cache to mitigate issues and consider the architecture monolithic, hope latency is ok!" ).

 

To me personally, this makes me hold off upgrading to a 3900X, at least until this issue is cleared up (also, I play destiny2 and it seems destiny2 can't even start on ryzen3000, but a bios fix for that is supposedly in the works, so I need to wait either way...).

 

[lysergicknight]

Hey, so this is mostly unrelated to this video's topic but, what game is being played during the L3 Cache Comparison from 2:40 - 2:53 in the video? 

-lys

[Spotty]

13 hours ago, lysergicknight said:

Hey, so this is mostly unrelated to this video's topic but, what game is being played during the L3 Cache Comparison from 2:40 - 2:53 in the video? 

-lys

Battlefield V.