Are you by any chance also sensitive to PWM (pulse width modulation) strobing? :-D
But like you said, this is not something that majority of people would notice.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Ryzen 3000 hype and review oddities
- Thread starter Morkai
- Start date
Are you by any chance also sensitive to PWM (pulse width modulation) strobing? :-D
But like you said, this is not something that majority of people would notice.
Yes I am.
It gives me a mighty headache too...
There's at least a fair chance this is the cross-chiplet latency in action. Try assigning affinity manually to the same ccx and see if it goes away?
Since this is the issue I was most suspicious about, I looked trough a lot of different forums both dedicated to various games and new/old ryzen and fixing issues with manual affinity (or ht disabling) has been a recurring theme.
Black desert online seems to have been one of the most problematic cpu intense games, and never fixed from what i gather. New 3000 series takes a huge performance hit in it so it is probably a good test case, and also free.
Last edited:
On the surface, that makes sense. However, I don't think that's it actually. This couldn't have been the case with the Athlon 64, X2, Phenom, Phenom II or Bulldozer. Intel systems have virtually always felt smoother, whether they are as fast or not. If this issue existed before that, I don't recall. I started reviewing motherboards back in the late Pentium 4 days. This was the first time I had the opportunity to go back and fourth between AMD and Intel in the same room and back to back like that. It's something Morry, Kyle and I had talked about back in the day when we were all running Pentium 4's. Obviously, it never stopped us from building and enjoying AMD systems. It's something you notice at the time but its not significant enough that you spend a lot of time thinking about.
As I said, we had always attributed it to Hyperthreading back when we only had single-core CPU's. Kyle even brought it up in the conclusion of one of our motherboard reviews back in 2006 or 2007. Long before chiplets or CCX complexes were a thing.
DuronBurgerMan
[H]ard|Gawd
- Joined
- Mar 13, 2017
- Messages
- 1,340
One thing Intel has almost always done better than AMD is cater to edge cases and smooth out the rough edges of a design. Usually there is more polish in an Intel product (and frankly, the same with Nvidia vs AMD/ATI). About the only time I can recall when this wasn't the case, off hand, is the launch of the original K7 Athlon. Even then, the chipset was a weakness - AMD's Irongate chipset was lacking in features, and later Via chipsets were a cluster. But K7 itself was a very polished product. It dominated in everything. Even the Athlon 64 and X2 releases - which were still great - weren't that dominating.
IMHO, this is because AMD is almost always fighting from behind for marketshare. Shoestring R&D budgets, pushing their tech to the edge, etc... In fact, AMD has made something of an art of pushing their designs right to the edge and somehow not crossing it (Intel, OTOH, when pushed has crossed that line - like with the famous P3 1.13 GHz). Precision boost really pushes Zen+ and Zen 2 right up to the edge. Very effective (though disappointing for us overclockers) AMD is occasionally brilliant at operating within severe monetary constraints. Of course, this is tempered by occasional turds too (K5, Bulldozer).
I think sometimes this becomes apparent in the end product in weird ways. Sketchy drivers for Radeons, sometimes. Finicky memory compatibility or bad BIOS pushes for the CPUs. Some vague sense of less smoothness here and there.
This doesn't change the fact that I really like what AMD has done with Zen. In typical AMD fashion they found a cheap shortcut around a significant problem - that being the ability to scale core count relatively cheaply - since single core performance was approaching a wall. It's the kind of near-hacky solution Intel would not be inclined to use unless there were no other options (although the Pentium D and Core 2 Quad did this - I think even Intel wanted to get away from that as quickly as possible). The CCX and CCD/Separate IO concepts, though? Typical AMD. Just like x86-64. Typical AMD to find a way to cheaply solve a problem.
There was a joke about how NASA spent millions trying to create a pen that would write in space. Whereas the Russians just used a pencil. At first the latter seems better - it's certainly clever - but NASA learned a great deal about fluid dynamics in space in the process. So both approaches had validity. In this, AMD is the Russians, basically.
The market benefits from both approaches, I think. Reliable, polished products, and hacky ideas to cut costs/push a tech to the edge. On my home workstation, I'm all about using Zen products - when the 3900X pops back into stock, I'll get one (unless stock is so thin the 3950X comes out first lol). I'll take the occasional reliability blip/weird vague issue in the interests of saving boatloads of money relative to the Intel competition.
But if my money was no concern, I'd stick with Intel. Fortunately for AMD, money is almost always a concern.
BrotherMichigan
Limp Gawd
- Joined
- Apr 8, 2016
- Messages
- 376
Meanwhile, I can point you to one four way blind test between a 5960X, 1800X, 7700K, and 1700X with three different users that couldn't determine whether any one system was smoother than another.
For reference:
Almost seems like an opinion!
Meanwhile, I can point you to one four way blind test between a 5960X, 1800X, 7700K, and 1700X with three different users that couldn't determine whether any one system was smoother than another.
For reference:
Almost seems like an opinion!
As I said before, it is ONLY something you see when you have the two absolutely side by side and work with them allot doing different things. I am currently using an AMD CPU in my main rig. It's obviously not that big of a deal to me.
Gideon
2[H]4U
- Joined
- Apr 13, 2006
- Messages
- 3,548
Might be more noticeable on a Threadripper. I never much noticed any difference on my friends computers and all of them ran Intel to my AMD system. Biggest issue I noticed is some hardware just plain refused to work on my system, past that my rig felt just as smooth as theirs in games. Tho half the fun back in those days was getting every computer to work and everyone into the game.
It might for sure have been because of smt back on single/(dual) cores - running anything on the virtual threads would have to share recources with the main rendering thread etc.
But it seems unlikely that should remain an issue forever, with optimizations to smt and more cores.
For a long long time amd had a LOT weaker cpu cores (at the very least since sandy bridge -> now), and too weak single core cpu cores generally determines minimum fps and produces visible stutter when it drops low enough, so could be different reasons for different periods in time.
hyperthreading -> weak cores -> chiplets, or some combo.
Hardware unboxed/techspot did a test of smt on/off on 3900x and it did not significantly impact the 1% minimum fps, sometimes it improved it, sometimes decreased it, but generally a few % here and there. Nothing like the 30-50% drops black desert and other games and the LTT video experiences. Really hard to determine from random info and black magic fixes, but there are widespread issues and witch doctor optimizations all over.
https://www.techpowerup.com/review/amd-ryzen-9-3900x-smt-off-vs-intel-9900k/
D
Deleted member 134608
Guest
I have experienced the exact same thing as well. Hyperthreading was a large part of that, but AMD also had the worst subsystems in existence. Poor USB means interrupts on your key strokes and mouse movement. Crap ATA and SATA controllers resulting in poor hard drive performance, resulting in longer indexing time which slowed windows down. All of that contributed to the AMD systems not feeling smooth. Intel simply decimated AMD in that area.
Huge improvements happened (FINALLY) in those areas in the Athlon X2 era.
Can't speak to Zen though. Haven't tried any other those systems yet.
DuronBurgerMan
[H]ard|Gawd
- Joined
- Mar 13, 2017
- Messages
- 1,340
To be fair, a good chunk of AMD's weird issues back in the Athlon era was due to Via's god-awful chipsets. AMD's own chipsets were usually fine, if uninspiring in performance. And the nForce chipsets were usually okay. But Via? Worst. Shit. Ever. Made.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
Chipset and chipset drivers, man. VIA got better at it and then straight up disappeared; I was surprised that AMD didn't use them for Ryzen. Or even Realtek.
And we can't be too sure how AMD's new X570 chipset is going to fair in daily use either though so far they've been pretty stellar.
DuronBurgerMan
[H]ard|Gawd
- Joined
- Mar 13, 2017
- Messages
- 1,340
I've had no issues with X370. Chipset side of things has been excellent. Memory compatibility has been a bitch, but that's more the fault of the CPU's memory controller than the chipset proper. Everything else is A+. Not sure if X570 will continue that trend, but I don't see a big reason for it not to.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
To a degree, you can divorce memory support from the chipset. They do however go hand-in-hand in terms of design targets, firmware development, and so on.
Well, it's new- that's the only reason- and with AMD that's enough for a degree of concern. It's nice that it's been mostly unwarranted with respect to the periphery despite the teething issues that the system has otherwise experienced.
Rockenrooster
Gawd
- Joined
- Apr 11, 2017
- Messages
- 955
I've had more issues getting Intel systems to work right compared to my Phenom II, Pildriver, Ryzen 1600, Ryzen 1700. (Compared to X5472 dual socket, i7-940)
All my work systems are Intel so I'm not counting those.
That i7 940 was real finicky with RAM, and I had BSOD issues with the X5472 (BIOS setting ended up fixing it)
I just stuck with the rated speeds as far as RAM speed (nothing too crazy).
Always do your research before buying somthing if you intend to spend lots of money on faster components (Faster RAM, Expensive CPU, Mobo) or you might set yourself up for failure.
Look at QVL lists, they're there for a reason. ( I didn't buy QVL listed RAM for my Ryzen (Hynix 3000 16GB sitcks, but I did research it)
Intel and AMD systems are inherently different and have their quirks, if you have had Intel systems all your life, then you might have problems with a brand new AMD system for some reason.
But then again I don't mess with RAID on consumer anything, only dedicated devices for those (NAS, RAID controllers, servers).
I think with Ryzen 3000, 3200Mhz is the official max supported RAM speed, anything above is "overclocking" and not supported by AMD.
All my work systems are Intel so I'm not counting those.
That i7 940 was real finicky with RAM, and I had BSOD issues with the X5472 (BIOS setting ended up fixing it)
I just stuck with the rated speeds as far as RAM speed (nothing too crazy).
Always do your research before buying somthing if you intend to spend lots of money on faster components (Faster RAM, Expensive CPU, Mobo) or you might set yourself up for failure.
Look at QVL lists, they're there for a reason. ( I didn't buy QVL listed RAM for my Ryzen (Hynix 3000 16GB sitcks, but I did research it)
Intel and AMD systems are inherently different and have their quirks, if you have had Intel systems all your life, then you might have problems with a brand new AMD system for some reason.
But then again I don't mess with RAID on consumer anything, only dedicated devices for those (NAS, RAID controllers, servers).
I think with Ryzen 3000, 3200Mhz is the official max supported RAM speed, anything above is "overclocking" and not supported by AMD.