Who's Happy So Far with Ryzen 3000 Series?

Having looked again I noticed that you run single thread bench, so my bad. I thought it was MT.


Good for you. At 4k you shouldn't notice any CPU bottleneck in low 1%. I noticed fair bit even in 1440p.
Well I do not notice it @1440p with any rAM timings, but i am old.
memory-chart-png.png
 
Well I do not notice it @1440p with any rAM timings, but i am old.
View attachment 178148
It’s not RAM timings, it’s CPU and more specifically a single thread performance. Try other games, like DE:MD or DirectX11 games, like GR:W.

For me and the games I play, 3700x without an overclock was a downgrade from 4790k @4.8GHz. Yes, extra cores helped, but smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations.
 
Not arguing with you,I am really fine playing on Ryzen system over Intel system.
 
It’s not RAM timings, it’s CPU and more specifically a single thread performance. Try other games, like DE:MD or DirectX11 games, like GR:W.

For me and the games I play, 3700x without an overclock was a downgrade from 4790k @4.8GHz. Yes, extra cores helped, but smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations.

WHEA bug might be your issue, AGESA 1.0.0.3ABB should fix it (Fps tanking)
even at 4Ghz a 3700x should be killing 4790K in singlethread performance even at 4.8Ghz easily, if something is very latency sensetive it's different story but we see that is very rare in pretty much every test.

So I'd patiently wait for the ABB (Nvidia users seems to be particularly affected) and see then :)
 
You lucky so and so! :-D

If my 3700x boosted that well I’d have kept it.
If my RAM worked at 3600MHz and not just at 3200MHz, I’d have kept it.
If AMD RAID worked reliably I’d have kept it.

Rant over.

EDIT:

Just watched your first video and noticed that yes two cores boosted well, but the rest were well under-clocked?!

That’s not I was expecting to see...

Having looked again I noticed that you run single thread bench, so my bad. I thought it was MT.


Good for you. At 4k you shouldn't notice any CPU bottleneck in low 1%. I noticed fair bit even in 1440p.

It’s not RAM timings, it’s CPU and more specifically a single thread performance. Try other games, like DE:MD or DirectX11 games, like GR:W.

For me and the games I play, 3700x without an overclock was a downgrade from 4790k @4.8GHz. Yes, extra cores helped, but smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations.

I changed my mind ,I want to argue and say I have not had the experience you had.Everything I quoted you saying I had a positive experience and I would say pretty much opposite of your experience with Ryzen.

So I took your suggestion and played Deus Ex Mankind Divided and set it to 2560x1440p Ultra and played 20 minutes.:D

I also reduced Clock speed of CPU to 4250Mhz and set V-sync on .
Looks to me 1% low is still pretty darn fantastic on a Ryzen CPU 3600X
for a game where your 3700X could not play smooth.Weird how that works. Wonder what it would be on an Intel machine with same settings ,opps I know what it would be lol.

My point is Ryzen CPU are capable CPU's for gaming and content creation and just as good as intel cpu's. I do not give a fuck what CPU you use .I can just test and come up with a result instead of saying Blah blah blah with Blah blah blah but smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations.Sorry also if I did not record your certain game location for the smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations .


Results
03-08-2019, 13:54:44 DXMD.exe benchmark completed, 77697 frames rendered in 1299.093 s
Average framerate : 59.8 FPS
Minimum framerate : 9.7 FPS
Maximum framerate : 60.9 FPS
1% low framerate : 47.1 FPS
0.1% low framerate : 13.8 FPS

Video with MSI Afterburner Running with the above results.
 
I changed my mind ,I want to argue and say I have not had the experience you had.Everything I quoted you saying I had a positive experience and I would say pretty much opposite of your experience with Ryzen.

So I took your suggestion and played Deus Ex Mankind Divided and set it to 2560x1440p Ultra and played 20 minutes.:D

I also reduced Clock speed of CPU to 4250Mhz and set V-sync on .
Looks to me 1% low is still pretty darn fantastic on a Ryzen CPU 3600X
for a game where your 3700X could not play smooth.Weird how that works. Wonder what it would be on an Intel machine with same settings ,opps I know what it would be lol.

My point is Ryzen CPU are capable CPU's for gaming and content creation and just as good as intel cpu's. I do not give a fuck what CPU you use .I can just test and come up with a result instead of saying Blah blah blah with Blah blah blah but smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations.Sorry also if I did not record your certain game location for the smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations .


Results
03-08-2019, 13:54:44 DXMD.exe benchmark completed, 77697 frames rendered in 1299.093 s
Average framerate : 59.8 FPS
Minimum framerate : 9.7 FPS
Maximum framerate : 60.9 FPS
1% low framerate : 47.1 FPS
0.1% low framerate : 13.8 FPS

Video with MSI Afterburner Running with the above results.


Thanks for proving my point.
 
Deus Ex MD is probably the most intensive game period. Not sure why it didn't become the Crysis of it's day.

I had a LOT of trouble running it when I had my Titan X Pascal even. This was on a 4K TV, but I had to drop to virtual 21:9 (1600P) and medium/high settings just to hit 60fps.

That game will melt your whole rig on ultra.
 
Eh, .1% probably right after a level loaded or something.

Thus the import of frametime analysis- you can toss those results as not representative of gameplay since you're not actually playing. But absent that analysis, we have no idea ;).
 
Eh, .1% probably right after a level loaded or something.
0.1% went as low as 1 FPS I think in the video on hacking the door.
1%Low of 47 FPS with Vsync on recording with shadow play is very good for DXMD game on 2560x1440 Ultra for 20 minute run.
 
Deus Ex MD is probably the most intensive game period. Not sure why it didn't become the Crysis of it's day.

I had a LOT of trouble running it when I had my Titan X Pascal even. This was on a 4K TV, but I had to drop to virtual 21:9 (1600P) and medium/high settings just to hit 60fps.

That game will melt your whole rig on ultra.

I never felt that game needed 60 FPS to be smooth. I think I played that on my 980Ti and 3770k @ 1440p without much of an issue.
 
1%Low of 47 FPS with Vsync on

Really shouldn't put V-Sync on for benchmarks, unless you're also testing with V-Sync off and discussing them side-by-side. V-Sync is going to throw off 1% and 0.1% readings quite a bit, and in a way that is going to mess with the applicability and repeatability of the benchmark.

[I say side-by-side because I understand playing with it on; it's a tradeoff between input lag and jitter, and tearing, and it depends on the viewer and the game which is worse...]
 
Vsync and under clock CPU 4250Mhz was on purpose to give the worst possible 1%LOW and it was still very good on the Ryzen.I could of ran no vsync and Max boost for my chip 4525Mhz and 1%LOW would not drop below 60FPS on a 20 minute run(I think but did not test 20 minute only 10 min)
 
Just put in a Ryzen 3700X, coming from a 2700X. Knew it wouldn't be a huge upgrade but did it anyways. A few things so far I don't like:

- At start up the 3700X (or the current BIOS for my motherboard) ramps up my CPU fan to full speed for 6-8 seconds before Windows loads. Very annoying; I recall some old motherboards/CPUs used to to that but I assumed those days were over. If anyone knows of a way to disable this I'd appreciate it. Specs below.

- Performance in games. Outside of games it does as expected, a little quicker for things like handbrake and 7zip. But in games, at the resolution I play at (2560x1440) it seems to be slower in half the games I tried. In Far Cry 5 I saw a small FPS bump. In Shadow of the Tomb Raider and Metro Exodus I get less frame rates. Ran all the benchmarks 3 times and even restarted Windows between the 2nd and 3rd one. Bit disappointing there.

Maybe there are some things I need to do in the BIOS as I know some settings were changed such as no more ASRock OC mode setting.

If anyone has any insight into issue one I'd appreciate it. Specs are the following:

ASRock X370 Killer SLI/AC with current 5.40 BIOS.
* CPU fan profiles in BIOS don't change this
16GB Corsair Vengeance LPX 3000 <-- running at 2933 MHZ.
Ryzen 3700X (previously used 1600 and 2700X which didn't have the fans ramp up to max at start up)
 
What were your actual numbers in the games before and after? Infinity Fabric works a bit different in Zen 2 and can be uncoupled from the memory speeds, but this can cause significant latency so make sure your IF/fclk is running 1:1 with your RAM speed.
 
Really shouldn't put V-Sync on for benchmarks, unless you're also testing with V-Sync off and discussing them side-by-side. V-Sync is going to throw off 1% and 0.1% readings quite a bit, and in a way that is going to mess with the applicability and repeatability of the benchmark.

[I say side-by-side because I understand playing with it on; it's a tradeoff between input lag and jitter, and tearing, and it depends on the viewer and the game which is worse...]

Ditto.
 
0.1% went as low as 1 FPS I think in the video on hacking the door.
1%Low of 47 FPS with Vsync on recording with shadow play is very good for DXMD game on 2560x1440 Ultra for 20 minute run.

Yeah, it’s the door. Just watched the video.

I got my numbers from DE:MD built-in benchmark, so that it’s easier to compare on any other hardware.

If you could re-run that, fully OC’d, no v-sync, I’d be grateful. It would be good to see what good Ryzen samples are capable of.
 
OK Built in benchmark it is.Another thing your may be hung up on max overclock which will not make a difference in some games on Ryzen like this one. I did not test Overclock on Intel system so I do not know if it makes difference on this game.
I tested 4500Mhz All Core OC and 4250Mhz All core Overclock.Both were the same results.

4250Mhz
Deus Ex Mankind Divided 1% LOW Test Vsync Off
2560x1440 Ultra
04-08-2019, 11:00:48 DXMD.exe benchmark completed, 6465 frames rendered in 79.844 s
Average framerate : 80.9 FPS
Minimum framerate : 64.9 FPS
Maximum framerate : 99.4 FPS
1% low framerate : 66.3 FPS
0.1% low framerate : 62.3 FPS

 
I like the performance a lot with my 3600, but don't like the way the processor behaves. Their new 1ms clock selection feature is pretty aggressive for a balanced power profile. Just moving around my mouse with nothing open, it ramps up my voltage and boost clock to max. When I'm just browsing around the web during very light workloads, it's constantly hitting max voltage and clocks. If I leave a web page open and don't move my mouse and just scroll, the CPU boosts to max. In turn, this just creates a lot of heat for nothing, which isn't good for summer time. Why do I need max clocks and voltages to scroll a small page of text? I can understand boost to render the page, but scrolling and moving my mouse around? It's kind ridiculous, as I'm hitting upwards of 65c just browsing the web.

This seems to only happen with W10 1903, as this is the only build that supports this new feature. On build 1803, the CPU behaves much better for me, as it doesn't support the 1ms clock selection. Doesn't boost during mouse movement, scrolling, etc - and I get about 10-15c cooler operation during normal tasks, due to the CPU not boosting and maxing voltage constantly for something as mundane as moving the mouse around.
 
I like the performance a lot with my 3600, but don't like the way the processor behaves. Their new 1ms clock selection feature is pretty aggressive for a balanced power profile. Just moving around my mouse with nothing open, it ramps up my voltage and boost clock to max. When I'm just browsing around the web during very light workloads, it's constantly hitting max voltage and clocks. If I leave a web page open and don't move my mouse and just scroll, the CPU boosts to max. In turn, this just creates a lot of heat for nothing, which isn't good for summer time. Why do I need max clocks and voltages to scroll a small page of text? I can understand boost to render the page, but scrolling and moving my mouse around? It's kind ridiculous, as I'm hitting upwards of 65c just browsing the web.

This seems to only happen with W10 1903, as this is the only build that supports this new feature. On build 1803, the CPU behaves much better for me, as it doesn't support the 1ms clock selection. Doesn't boost during mouse movement, scrolling, etc - and I get about 10-15c cooler operation during normal tasks, due to the CPU not boosting and maxing voltage constantly for something as mundane as moving the mouse around.

Well if you want some help and would like to have no voltage spikes and low temperatures all you need to do is:
Install latest BIOS fr you X370/350/X470/450
Install Latest AMD Chip Set driver and use Balanced power plan also windows 10 1903

Last but not least if above was not to your liking ,do a proper controlled overclock keeping performance with low temperatures and voltage.

Use AMD overclocking to have your CPU down-clock to zero/sleep when using proper tool to monitor.Check spoiler for BIOS settings to use.

Desktop-Screenshot-2019-08-01-15-14-36-50.png


ADVCPUCONFIG.png

ADVCPUCONFIGPBODisable.png

AMDADVOC.png

AMDADVOC4325-MHZ.png

AMDADVOCPBODIS.png
 
Install Latest AMD Chip Set driver and use Balanced power plan also windows 10 1903

Is this for Ryzen balanced or Windows balanced? My understanding is, the Windows Balanced profile does not support the 1ms clock selection, even on 1903. They specifically say it's only supported for Ryzen balanced profile.

If it came down to it, I'd rather just revert back to 1803, as 1903 is bug ridden as usual with any new W10 build. I got bad performance hit in a few games with 1903, like Apex Legends, and the new color profile bug is really annoying.
 
Well Ryzen balanced ,I just offered some advise to you that would help.I do not play Apex legends and have no clue about a new color profile bug in 1903.Fuck maybe I should roll back to 1803 to avoid all the problems lol.

Do not look at the screenshot showing 33C and peek speed of 378Mhz and I forgot snappy as normal.J/K
 
Just put in a Ryzen 3700X, coming from a 2700X. Knew it wouldn't be a huge upgrade but did it anyways. A few things so far I don't like:

- At start up the 3700X (or the current BIOS for my motherboard) ramps up my CPU fan to full speed for 6-8 seconds before Windows loads. Very annoying; I recall some old motherboards/CPUs used to to that but I assumed those days were over. If anyone knows of a way to disable this I'd appreciate it. Specs below.

- Performance in games. Outside of games it does as expected, a little quicker for things like handbrake and 7zip. But in games, at the resolution I play at (2560x1440) it seems to be slower in half the games I tried. In Far Cry 5 I saw a small FPS bump. In Shadow of the Tomb Raider and Metro Exodus I get less frame rates. Ran all the benchmarks 3 times and even restarted Windows between the 2nd and 3rd one. Bit disappointing there.

Maybe there are some things I need to do in the BIOS as I know some settings were changed such as no more ASRock OC mode setting.

If anyone has any insight into issue one I'd appreciate it. Specs are the following:

ASRock X370 Killer SLI/AC with current 5.40 BIOS.
* CPU fan profiles in BIOS don't change this
16GB Corsair Vengeance LPX 3000 <-- running at 2933 MHZ.
Ryzen 3700X (previously used 1600 and 2700X which didn't have the fans ramp up to max at start up)

I have the exact same Mobo (well mine is a v1). Is yours a v2? Do you have LLC? If so then its a v1
 
Ill.come back and say with latest agesa and f5i bios from Gigabyte im like a fat kid in a candy store.

LOVE IT
 
It’s not RAM timings, it’s CPU and more specifically a single thread performance. Try other games, like DE:MD or DirectX11 games, like GR:W.

For me and the games I play, 3700x without an overclock was a downgrade from 4790k @4.8GHz. Yes, extra cores helped, but smoothness they bring get shattered perceptibly when FPS tanks briefly in certain game locations.

If you're a hardcore gamer, Intel is still the way to go. I owned both 3800X@ stock and 9900K@5Ghz with the same setup except the cpu's and motherboards. They trade blows. Since I'm not a hardcore gamer, 10~15FPS advantage isn't an advantage anymore when I'm already hitting ~120FPS on games.
When it comes to all other non-gaming apps, 3800X@stock easily outpaces 9900K@5Ghz. Too bad I can't oc 3800X to 5Ghz.
 
OK Built in benchmark it is.Another thing your may be hung up on max overclock which will not make a difference in some games on Ryzen like this one. I did not test Overclock on Intel system so I do not know if it makes difference on this game.
I tested 4500Mhz All Core OC and 4250Mhz All core Overclock.Both were the same results.

4250Mhz
Deus Ex Mankind Divided 1% LOW Test Vsync Off
2560x1440 Ultra
04-08-2019, 11:00:48 DXMD.exe benchmark completed, 6465 frames rendered in 79.844 s
Average framerate : 80.9 FPS
Minimum framerate : 64.9 FPS
Maximum framerate : 99.4 FPS
1% low framerate : 66.3 FPS
0.1% low framerate : 62.3 FPS



Perhaps, the game depends more on the graphic card than the cpu at that resolution? Try the benchmark at 1080P instead.
 
If you're a hardcore gamer, Intel is still the way to go. I owned both 3800X@ stock and 9900K@5Ghz with the same setup except the cpu's and motherboards. They trade blows. Since I'm not a hardcore gamer, 10~15FPS advantage isn't an advantage anymore when I'm already hitting ~120FPS on games.
When it comes to all other non-gaming apps, 3800X@stock easily outpaces 9900K@5Ghz. Too bad I can't oc 3800X to 5Ghz.

Your comment sounds contradictory possibly? Confusing?

You say Intel is the way to go, but then they trade blows, and AMD easily outpaces?

Not sure what your actual stance is on the chips.

I would argue that Single thread performance of the 3900x and possibly the 3700x is identical with the appropriate ram and timings as Intel at 5ghz. This is quantifiable by common benchmarks. But this is a hot debate because we are letting emotion get in way of science (not claiming that about you in the least bit). That was a generalist statement.
 
Your comment sounds contradictory possibly? Confusing?

You say Intel is the way to go, but then they trade blows, and AMD easily outpaces?

Not sure what your actual stance is on the chips.

I would argue that Single thread performance of the 3900x and possibly the 3700x is identical with the appropriate ram and timings as Intel at 5ghz. This is quantifiable by common benchmarks. But this is a hot debate because we are letting emotion get in way of science (not claiming that about you in the least bit). That was a generalist statement.

I didn't find what he said that hard to follow, especially having seen benchmarks of games and non-gaming loads. They do trade blows in games, but overall, Intel comes out on top more often than not. He also said in non-gaming apps, AMD easily outpaces.
 
Perhaps, the game depends more on the graphic card than the cpu at that resolution? Try the benchmark at 1080P instead.

Come on dude get real RTX 2080/Ti are not for playing games at 1080p for 95% of people.
I play 4k with RTX 2080/Ryzen 3600X and guess what it works the same as Intel 9900K/RTX 2080 and I am also not worried that someone may get 5 FPS more than me in a game playing 720/1080/1440/2160.
What I do care about is showing people you do not need to spend more money than what is needed to get the same gaming experienced.I am not brainwashed by brand power.I know the difference.

No offense,I happy with Ryzen 3600X
 
Your comment sounds contradictory possibly? Confusing?

You say Intel is the way to go, but then they trade blows, and AMD easily outpaces?

Not sure what your actual stance is on the chips.

I would argue that Single thread performance of the 3900x and possibly the 3700x is identical with the appropriate ram and timings as Intel at 5ghz. This is quantifiable by common benchmarks. But this is a hot debate because we are letting emotion get in way of science (not claiming that about you in the least bit). That was a generalist statement.

For gaming or signle core apps 9900K@5ghz is the winner.
For all multi-thread non-gaming, 3800X stock is the winner.

So they trade blow.

However, it's a different story if both CPU's run at stock or equal mhz.
 
Using 3600 and productivity apps. Have made a nice increase.depending on the game there is either a good gain at the top end or not.Now minimums have boosted quite a bit and things have become smoother.Overall quite happy while using a asus b450 board with bios update.

Only bitch is the bios is a pain to figure out as documents with mb or on line are nonexistent.
 
I been getting the upgrade itch for AMD in past week. Probably cause I have been having problems with my x299 7820x bit that has been sorted with a new MB. I don't need this and would be a side grade at best. But me wants....
 
I been getting the upgrade itch for AMD in past week. Probably cause I have been having problems with my x299 7820x bit that has been sorted with a new MB. I don't need this and would be a side grade at best. But me wants....

Maybe buy/sell/trade section and ask if someone wants to trade.It sounds silly but hey you never know.
 
Bumped up from a 3470 to a 3600 last week. Paired with a MSI 1060 6GB. Goes well. No issues. FC5 at maxed setting without a hitch. I'm satisfied.
 
I have an MSI Gaming Pro mATX 350b thats running a 1600 now and I plan to update all the chipsets and drivers as dropping another Gen 3 under the roof . As the First Gen cheap boards ain't cheap no more and the board is worth more paired with a smooth running 3600 now that I have had the time with my first Ryzen Gen 3 / 350b .. the MSI board is setting in a Red STYX, RAIJINTEK's with an RX 580 8Gb that I was plan to use for test mule for people interested in a build to spend time with before you buy .
 
Is this for Ryzen balanced or Windows balanced? My understanding is, the Windows Balanced profile does not support the 1ms clock selection, even on 1903. They specifically say it's only supported for Ryzen balanced profile.

If it came down to it, I'd rather just revert back to 1803, as 1903 is bug ridden as usual with any new W10 build. I got bad performance hit in a few games with 1903, like Apex Legends, and the new color profile bug is really annoying.

I saw no difference between windows balanced and ryzen balanced when it comes to boost clocks. But windows balanced did reduce voltage further than ryzen balanced. But to check voltages with ryzen balanced you need to have ryzen master open. CPUZ for me didn't monitor the voltages quickly enough but with new chipset drivers it was much more relaxed for me compared to original profile.
 
I'm mostly happy with my new rig (3700x + Aorus Pro Wifi + 32GB Hynix CJR). It's twice as fast as my old 4790k at everything (except gaming of course). Have had no crashing or driver issues but I do have some gripes. I get good enough performance, though of course when I do high refresh gaming - there is a bit of remorse and wondering if I should have grabbed a 9700k/9900k, but I'm planning on upgrading this to a 16 core (hopefully 4950x) down the line, so it felt like the better investment. In the short term (minus power consumption) the 9900K with a cheaper board would have come out to about the same price and be the better performer in nearly everything.

My gripe, like mostly everyone, is the boost clocks / nothing happening with PBO. I was excited when I saw that Robert Hallock (link) video advertising the ability to +200 boost these CPUs. Out of the box with PBO on, I get 4250 on all 8 cores during gaming (4175ish under AIDA64 stress load) so the all core performance is very nice. However, My max boost is only 4315 and that comes very rarely in a small enough flash to make it useless. So I effectively have a 4.25ghz all core CPU that doesn't boost at all, and I've never seen the box rated 4400 (chip is under water too). I'd like to see a fix, but not expecting much as it's hard enough to even get a single ccx above 4300mhz on the 3700x. This chip is completely maxed out, and clearly the lowest of the low binned full 8 core CCDs.

The gimped memory write speed also annoyed me a bit (didn't do my due diligence before buying), I understand it's still within spec and wont hurt games and most mainstream applications, but, it's still literally a 50% drop in write performance and the "to save power" reason they gave makes no sense, unless the 3800x has dual CCDs, that chip should also have 28000MB/s~ write speeds at the same 105W as the 3900x, so that sounds like a lie to me. The truth is it's just a side effect of the design, but one that won't be noticed by the average consumer, so I'm just annoyed by the explanation primarily.

so, knowing what I know now. I would do a 3600 for $199 or go all out on a 3900x (or soon 3950x). 3700x is kind of the monkey in the middle, too expensive for what it is, too cheap to be the top of the line. Same should apply to 3800x.

I'm thinking to upgrade from a 4770k @ stock. Is the 3700x really slower @ gaming? Ubuntu 19.04 with GTx 1070 but I may pick up a 3700x or 3900/3950x after the 3950x is out. May also pick up a 5700XT - undecided.
 
Last edited:
Back
Top