Who's Happy So Far with Ryzen 3000 Series?

.....about the boost asthma:

The whole method is retarded, non-verifiable, non-controlled and false data could be inserted or used to manipulate it. Showing the significance of any variation is way more important - if that was done maybe he would realize how stupid this was. Using HWinfo at a high polling rate will indeed spawn more threads to point it will hinder boost (how many did not change polling rates). Anything in the background as in virus programs, web programs etc. can spawn threads which will - well reduce max boost clocks. Not to add poor cooling, cases, bad installation of coolers etc. Good grief what a waste YouTube video's have become and the high level of real journalism of previous years have virtually vanish.
 
The whole method is retarded, non-verifiable, non-controlled and false data could be inserted or used to manipulate it. Showing the significance of any variation is way more important - if that was done maybe he would realize how stupid this was. Using HWinfo at a high polling rate will indeed spawn more threads to point it will hinder boost (how many did not change polling rates). Anything in the background as in virus programs, web programs etc. can spawn threads which will - well reduce max boost clocks. Not to add poor cooling, cases, bad installation of coolers etc. Good grief what a waste YouTube video's have become and the high level of real journalism of previous years have virtually vanish.

That's a lot of excuses why a product isn't delivering as advertised performance across reports from users.

If a company can't deliver a product that performs as advertised at a premium price, then I won't buy it.
 
The whole method is retarded, non-verifiable, non-controlled and false data could be inserted or used to manipulate it. Showing the significance of any variation is way more important - if that was done maybe he would realize how stupid this was. Using HWinfo at a high polling rate will indeed spawn more threads to point it will hinder boost (how many did not change polling rates). Anything in the background as in virus programs, web programs etc. can spawn threads which will - well reduce max boost clocks. Not to add poor cooling, cases, bad installation of coolers etc. Good grief what a waste YouTube video's have become and the high level of real journalism of previous years have virtually vanish.

AMD disagrees with you as they are working on a new bios to address the issues in this “waste” of a YouTube video. So it appears that the biggest waste of bandwidth here is your opinion.
 
AMD disagrees with you as they are working on a new bios to address the issues in this “waste” of a YouTube video. So it appears that the biggest waste of bandwidth here is your opinion.

He will likely be proven right, You really have to go out of your way to isolate to 1 thread which is easy to do on a benching computer but a daily use computer will be another story. If they can make it boost as high on 2 cores then maybe that will make others happy, but I have my doubts that will happen. Likely you will see a single thread only boost properly then it will drop down rapidly with more cores, 2 cores might be 4.4, 3 cores will be 4.3 and 4 or more will be 4.3 or less as thermals permit. Thats my guess tho based on what I have seen, would prefer if they would allow me to have manual control over each boost stage based on how many cores are in use. But I am not sure if Windows is keen enough to use the right cores.
 
He will likely be proven right, You really have to go out of your way to isolate to 1 thread which is easy to do on a benching computer but a daily use computer will be another story. If they can make it boost as high on 2 cores then maybe that will make others happy, but I have my doubts that will happen. Likely you will see a single thread only boost properly then it will drop down rapidly with more cores, 2 cores might be 4.4, 3 cores will be 4.3 and 4 or more will be 4.3 or less as thermals permit. Thats my guess tho based on what I have seen, would prefer if they would allow me to have manual control over each boost stage based on how many cores are in use. But I am not sure if Windows is keen enough to use the right cores.

If he's right, AMD would have responded similarly to what you just said instead of spending time and resources fixing something that's operating perfectly normal.

The testing done isn't perfect but there's far too many accounts of boost issues and circumstantial evidence for it to be mere chance.
 
If he's right, AMD would have responded similarly to what you just said instead of spending time and resources fixing something that's operating perfectly normal.

The testing done isn't perfect but there's far too many accounts of boost issues and circumstantial evidence for it to be mere chance.
That is not the point at all, video was plain out retarded from a standpoint of validity, methodology or consensus and prevention of falsification. The results are meaningless in my view with that video.

Actual tests between the single thread correct boost to actual boost show such small deviations from a performance standpoint that it makes the significance trivial. Now I do expect AMD boost frequency advertise to be what a properly in spec configuration will have. Agree with Gideon on having a single thread Max Boost is so limited with today programs and usage that having two or more thread boost would actually be more meaningful.

The real question should be to AMD is will this max boost actually make any difference at all for gaming and other applications besides very limited benches and programs that are severely limited to one thread? To me it looks more to be a totally marketing driven spec that may have very limited to no impact to the user. Now to be clear, Boost is done by thread, threads and on down at lower frequencies each - multiple threads look to be the same between the older higher maxed boost SMU module to the newer slower maxed boost SMU module. 4.6ghz Maxed Boost is a single thread! Two threads it knocks it down to 4.5ghz. How useful in real applications, multi-threading going on, background processes that single thread boost is, well is another thing.

precision_boost-frequency-threads.png


The above is a graph from GamerNexus - Now what concerns me more is the 4.3ghz boost when having 14T - 24T, I am not seeing that myself and not sure how accurate the above graph represents what AMD has put out.

I have cores after hours of use, testing, benching that never go above 4.267ghz for max frequency obtain in HWinfo. Meaning the above graph is not representing what I am seeing with my 3900x. Even with lightly threaded, using most of the the CPU threads, 3 cores never go above 4.267ghz with PB. When all cores are in use depending upon application, I see 4.1-4.2ghz speeds, 300-400mhz Boost.
 
Last edited:
The whole method is retarded, non-verifiable, non-controlled and false data could be inserted or used to manipulate it. Showing the significance of any variation is way more important - if that was done maybe he would realize how stupid this was. Using HWinfo at a high polling rate will indeed spawn more threads to point it will hinder boost (how many did not change polling rates). Anything in the background as in virus programs, web programs etc. can spawn threads which will - well reduce max boost clocks. Not to add poor cooling, cases, bad installation of coolers etc. Good grief what a waste YouTube video's have become and the high level of real journalism of previous years have virtually vanish.

You must have paid attention in College like I did. I too made the same exact statement as you earlier.

Id much rather see a video about this from Dr. Ian Cutress of Anandtech.
 
Last edited:
  • Like
Reactions: noko
like this
Finally got time to do some testing tonight with my 3900x. Loaded the latest 1.0.0.4b beta bios for my asrock x470 taichi. This is with no overclocking, all stock settings and LLC set to Level 1 (lowest setting for asrock).

Untitled.png


Finally got to see above 4600MHz on a core!

wQ4iKxw.png
 

Attachments

  • 4p9Vki7.png
    4p9Vki7.png
    3.5 KB · Views: 14
Last edited:
Finally got time to do some testing tonight with my 3900x. Loaded the latest 1.0.0.4b beta bios for my asrock x470 taichi. This is with no overclocking, all stock settings and LLC set to Level 1 (lowest setting for asrock).

View attachment 208175

Finally got to see above 4600MHz on a core!

View attachment 208174

My 3960x hit 4.6 one day and its only designed to go to 4.5. Anomaly?

Grats on that killer 3900x. That thing is a monster 12 core.
 
My 3960x hit 4.6 one day and its only designed to go to 4.5. Anomaly?

Grats on that killer 3900x. That thing is a monster 12 core.

In HWInfo64 I will see 4.55-4.6 on most cores. If you run Ryzen Master and set it to keep the Peak frequency on the graph then I have never seen more that 4.4Ghz on any single core during testing. Now I have not sat and run Ryzen Master all day to see what it peaks to during use, but when running alongside the HWInfo they show much lower numbers. Maybe RM needs the core to stay at a speed for a certain amount of seconds instead of just a blip though to record the Peak.
 
Got a 3600x for $175, sold the old 1600 for $75. Feels great for a $100 upgrade! It doesn't do anything too special, 4.4 is the highest I see from it. But that's 400mhz over what the 1600 did along with the IPC improvements, definitely a snappier piece of hardware. Hoping the trend continues!
 
I bought my Ryzen 3600 last friday. It does not do anything special over my old 1600 (GTX1080 is the "bottleneck" most of the time) but it is an upgrade. Unfortunately my sample is partly bad, it does not do all-core 4.2 without going over safe voltages but on the other hand mine has really good infinity fabric. It does 1900mhz FCLOCK out of the box stable without any voltage adjustment. This is great news because even though I have slow 3000mhz ram that does not clock beyond 3333 and be stable I can run my decoupled way above it so I do get some good gains from it, even if the clocks are not in synch. Next step is getting the fastest ram that I can run on 1:1 memory clock and FCLOCK synchronized. And then new GPU after which I can start building a HTPC from my old 1600, RAM and GPU. (y)

*edit* Also what is the best way to test the singlecore boost clocks? Just to check if mine gets 4.2 at all. :D
 
Last edited:
I bought my Ryzen 3600 last friday. It does not do anything special over my old 1600 (GTX1080 is the "bottleneck" most of the time) but it is an upgrade. Unfortunately my sample is partly bad, it does not do all-core 4.2 without going over safe voltages but on the other hand mine has really good infinity fabric. It does 1900mhz FCLOCK out of the box stable without any voltage adjustment. This is great news because even though I have slow 3000mhz ram that does not clock beyond 3333 and be stable I can run my decoupled way above it so I do get some good gains from it, even if the clocks are not in synch. Next step is getting the fastest ram that I can run on 1:1 memory clock and FCLOCK synchronized. And then new GPU after which I can start building a HTPC from my old 1600, RAM and GPU. (y)

*edit* Also what is the best way to test the singlecore boost clocks? Just to check if mine gets 4.2 at all. :D
I thought pbo does highest single core possible? Maybe I'm wrong though.
 
I enjoy my 3700x, it tops out around 4250, so if anything I wish the 3000 series could clock higher. It rips through everything I toss at it... I don't really pay attention to FPS much, and even though I think the 9900k might give me better higher lows, it's a solid CPU. I'm looking forward to the new year, see what Comet Lake brings with the 10c/20t part, and of course Ryzen 4000 series late in the fall 2020.
 
I am not happy with the 3000 series, because I do not own one yet. :D ;) Only thing worth getting to me is the 3900X, since I already own the 1600, 1700 and 2600.
 
I thought pbo does highest single core possible? Maybe I'm wrong though.

The way I understand it the PB2 does the max boost for single core already. Maybe even more if cooling allows. PBO may increase the clockspeed when more cores are utilised but in practice it really does nothing but increase voltage and heat with barely any performance gain.
 
The way I understand it the PB2 does the max boost for single core already. Maybe even more if cooling allows. PBO may increase the clockspeed when more cores are utilised but in practice it really does nothing but increase voltage and heat with barely any performance gain.
Ah, fair enough. I got my best single core speed with pbo and running 3dmark. I got overkill cooling anyways, so I don't mind the heat :D
 
Ah, fair enough. I got my best single core speed with pbo and running 3dmark. I got overkill cooling anyways, so I don't mind the heat :D

Yes the Max clock speed is on a single core. On my 3900X I get around 4525Mhz when doing a 3DMark Stress Test or 4550Mhz when doing Cinebench R20. Never quite hit the 4.6Ghz mark as of yet. Playing around I all core clocked it at 4.4Ghz but it runs a little warm at that speed.
 
Just jumping in. I put together a system recently with a 3900x and X570 Aorus master. Been very happy with it so far. Almost have all the i/o in use as well - all 3 m.2 slots (2 NVMe and 1 sata) 2 4tb hdds, 2 sata ssds, 1 gpu, 2-port i350 nic, x4 riser card for another small NVMe and m.2 sata ssd. All that’s left is the second x16 (x8) slot which may yet get another gpu for passthrough once Ubuntu 20.04 is out (and presumably works a little better with the 5700xt). 64gb ram. H150i pro aio for cooling. O11 case with 5 fans in it right now, few more ordered just to fill it up (all ML120 pro with red led).

I had actually bought most of the parts to do a 3700x mATX build with a b450m mortar max but got a great deal here on the 3900x and board.

I planned to put together a system I’d eventually use for gaming but for the first few months would be running virtual machines for work. Was going to use Linux kvm to do gpu passthrough but had just enough trouble with my 5700xt that I decided against it. I ended up going full on esxi install.

have esxi on a usb flash drive and windows 10 installed on one of the NVMe drives. Easy enough to reboot to flip between the two. Esxi is usually pretty wonky with consumer stuff - won’t recognize Realtek nics or the intel/amd onboard raid for instance - but once I disabled some of the “auto overclock” functions in the bios it has been happy.

I honestly though I’d need threadripper to put something like this together, just kind of boondoggled that after almost a decade of “4 cores” being the high end for mainstream platforms I’ve actually put together a full on gaming workstation with regular desktop (albeit almost top end) parts.

the first two generations of Ryzen really shook things up, but I’d call third gen Ryzen of the biggest advancements in end user computing since multi core went mainstream with core 2 duo.

now I just have to tell myself there’s no need for 4 more cores with a 3950x and wait to see what zen2+ or zen3 can bring to am4.
 
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
Replaced my Corsair H150i with a Noctua NH-D15. Pump was starting to make a weird noise and I wanted to get more airflow over the VRM's and memory. Replaced the case fans that came with my Fractal Design R6.

Before / After temps using AIDA64 stress test:

CPU (Tctl/tdie) - 88.5C / 80.3C
CPU Die (Average) - 81.2C / 77.5C
CCD1 - 89C / 80.3
CCD2 - 83.3C - 80.8C
Motherboard - 44C / 38C
CPU - 88.9C / 80.4C
VRM - 54C / 45C
Voltage Regulator Loop 1 - 61C / 51C
Voltage Regulator Loop 2 - 54C / 48C
DIM 0 - 50.5C / 42C
DIM 1 - 48.8C / 41.8C
NVME SSD (Idle) 41C / 35C
2080 Ti - 67C / 62C


IMG_20200111_082713.jpg
IMG_20200111_120025.jpg
IMG_20200111_124905.jpg
 
Last edited:
pandora's box thats really impressive temps! How many case fans did you have in there before?

I like my h150i pro and prefer the cleaner look it gives my setup (8x ML 120 pro red led, PC-O11 dynamic case), I think these dual tower designs are a bit much, but it’s hard to argue with results like that.
 
I've found no complaints about my 3600X. It's surprising because the last time I built a desktop was when the Phenom II was AMD's flagship and they did not compete well with Intel's HyperThreaded processors.
 
Ryzen 3800x / GTX980Ti @1440 / 32GB Ballistix 3200 @ 3400 (gonna try and hit 3600 when i have time) / ASUS TUF X570

plays Red Dead Redemption 2 smooth as butter on mixture of ultra to mostly high @ 1600x1200 85hz on my NEC FE 1250+ CRT. so i'm chillin right now. can't wait for cyberpunk 2077. kinda mad it got delayed because was suppose to be coming out on my birthday weekend so was gonna take a day off work, but if they need more time, they need more time. Was also wondering why they were releasing "Predator: Hunting Grounds" at the same time but they must have gotten word that cyberpunk was gonna be delayed.
 
I had a 6850k that runs games at 4k pretty much the same. But now I can stream and play with all of these extra cores on my 3900x. They overclock almost the same as well. I lost 2 fps in timespy switching from intel but obviously get a much better (like double) cpu score. Because of the memory fabric I can overclock my memory which I couldnt before. I also now have pcie 4 nvme and 3.2 usb. All in all its pretty good. It's alot better than my last amd the 9370 bulldozer. :)
 
I upgraded from a 1700 to a 3900x 3 weeks ago.

And even though my motherboard is not a top model, it has been completely painless. It restarted a few times on the first boot, before starting Windows, and since then it has just worked.

I installed all the newest drivers after installing the new CPU. The system is rock solid stable, I've not had a single crash, except for when I have been messing with memory speed.

Only disappointment is memory speed, that I have not been able to get above 3466 yet (was hoping for 3600), but I've read this can be troublesome with some b350/b450 boards.
Perhaps I'm also too aggressive on the timings.
Edit: I managed to make it run @3600, had to run max voltage for memory (1.48V) as recommended in Ryzen dram calculator.
Edit 2: Got it to 3666 with 1.495V. Mem temps are getting close to 65c with stress tests (sits at about 45c with normal usage), so I'll leave it there. I'm having a hard time booting to Windows @ 3733, I think the Infinity Fabric is getting to its limits on this chip/board. I am very satisfied with these results.

The CPU runs just shy of 4.2Ghz all cores when running benchmarks, and I have had cores boost to 4625Mhz. Worst cores boost to 4325Mhz

All in all a very big improvement over the 1700 it replaced.
 
Last edited:
Zero regrets going 3600. My X58 Dell T3500 was running fine but it was time.. I have not gamed lately, but it's fast for everything, and benches about 100% faster than the X58 system
 
DX50 rules! I had a friend with one back in the day and he was running Linux on it, back when it was new. Sweet ass system I tell you.
DX33 was the first system I built myself. DX50 was not yet available The DX33 was on a 16 slot 9 bit (with parity) FPM RAM. All ports were on ISA extension cards, including graphics. I had a paradise SVGA graphics card at that time with 85GB Seagate hard drive full heigth but still 3.5''. I used MS DOS 5 at first with Autocad 10, then sicthed also to Windows 3.0 and later to Windows 3.1 enhanced by PC Tools Desktop UI. I still don't use that good a desktop UI as that one. I laso used Linux 0.99 first and later some Linux 1.2, every time from Slackware. I had multiple boots and several setups on MSDOS/Windows. Went finally with Windows 95 with that one which it was still able to handle. Upgraded the graphics first to Cirrus Logic and next to S3 3D Virge (still on ISA bus), RAM to 16GB, and had drive. Could play Tomb raider with no problem (however 3D Virge had buggy 3D hardware accel and best was to use software 3D). Could also play Quake but with poor FPS and lower graphics. Last hard drive was a several Gigabyte Bigfoot with extender. After that I switched as my main personal PC to a Pentium 133 64GB, ATI graphics 128, and NT4. After I switched to a Pentium II 266 on slot, 512GB, still on NT4 at first, and an Abit BX6 (1.0) MB and also an Intel i740 (replaced later by an ATI 9500). and that one last very long, finished using it around 2010 on Windows 2000 and a Celeron 1200 over slotket, used for secondary services. Also my screen went from poorsman 14'' SGVA to 22'' Trinitron by Mitsubishi as the last non LCD.

So all this assembling PC thing is a continuation for more than 30 years now.
 
Back
Top