• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

AMD Ryzen 9 3900X

Joined
Jul 18, 2016
Messages
506 (0.18/day)
System Name Gaming PC / I7 XEON
Processor I7 4790K @stock / XEON W3680 @ stock
Motherboard Asus Z97 MAXIMUS VII FORMULA / GIGABYTE X58 UD7
Cooling X61 Kraken / X61 Kraken
Memory 32gb Vengeance 2133 Mhz / 24b Corsair XMS3 1600 Mhz
Video Card(s) Gainward GLH 1080 / MSI Gaming X Radeon RX480 8 GB
Storage Samsung EVO 850 500gb ,3 tb seagate, 2 samsung 1tb in raid 0 / Kingdian 240 gb, megaraid SAS 9341-8
Display(s) 2 BENQ 27" GL2706PQ / Dell UP2716D LCD Monitor 27 "
Case Corsair Graphite Series 780T / Corsair Obsidian 750 D
Audio Device(s) ON BOARD / ON BOARD
Power Supply Sapphire Pure 950w / Corsair RMI 750w
Mouse Steelseries Sesnsei / Steelseries Sensei raw
Keyboard Razer BlackWidow Chroma / Razer BlackWidow Chroma
Software Windows 1064bit PRO / Windows 1064bit PRO
Set your cpu to stock
So in all your test u used the stock frequency without boost?

Anandtech shows the Ryzen 3000 under full load using less power than the Intel parts, Techpowerup shows it using more. Maybe Anandtech isn't loading it hard enough.
Anandtech does serious review. Also in other review ryzen use less power.
 
Joined
May 8, 2018
Messages
1,494 (0.69/day)
Location
London, UK
Ok I have the following boards and I am confident they will work.

Asus X470 Prime
As rock X470 Master
As ROck X470 Taichi
Gigabyte X470 Gaming 7

I am a bit confused about the chart stating that the 3950 will not work with some boards in the X470 lineup including a couple that I have mentioned that have up to 2 8 PIN CPU and 12 phases.


Work work, every am4 motherboard with a bios support will work with a 3900x, the question is how stable it will hold that. For those boards I would recommend only a 3700x, 65w will be fine, even with an overclock, it will get to 90w.



"We noticed that, interestingly, our R9 3900X and R7 3700X overclocked with much lower voltage requirements than our R5 3600. Our R5 3600 is a production sample from a third party, so we have two thoughts, here: (1) Most realistically, the R5 CPUs probably aren’t binned as aggressively as the 12-core part, which would need lower voltages for thermal reasons, and (2) the samples AMD shipped had paste left over on them and they were pretested. We don’t suspect they were binned by AMD for reviewers, though we did want to point it out. With a sample size of one each, we can’t draw conclusions -- maybe it’s luck or maybe R5 just runs with a higher voltage. The upside is that the limited R5 core count means it can take the higher voltages as there’s lower thermal density to dissipate. Our R9 3900X could do 4.3GHz all-core at 1.34V to 1.35V, the R7 3700X could hold similar voltages, but the R5 3600 required 1.43V for 4.3GHz all-core. We were able to push 4.4GHz on the 3900X with SMT disabled, shown in our 3900X review that’s still rendering at time of writing this, but we could not reach 4.4GHz on any chip under 1.46V. We stopped at 1.46V as we encountered issues with, predictably, thermals on a reasonable 280mm CLC. We’ll push harder with liquid nitrogen later in the week.

Finally, FCLK will be a big part of memory overclocking later on. We’ve done some infinity fabric FCLK tuning and found no meaningful change when memory is left to our standard test 3200MHz settings, at least not immediately, so we’ll need to table this for now and dig in more later. Keep in mind that we had to write and film 5 content pieces relating to AMD products in just a few days, so some sacrifices were made. Mostly to health, granted, but we did remove some tests from content for later benchmarking.

Finally, related to sacrificing sleep, please be advised that there are definitely going to be grammatical typos in this article. We simply don’t have infinite time, so we won’t comb through it for misspellings. Thank you for your understanding!"


I wonder what power consumption it would have with smt off on the 3900x and lowered voltage.

Ok I have the following boards and I am confident they will work.

Asus X470 Prime
As rock X470 Master
As ROck X470 Taichi
Gigabyte X470 Gaming 7

I am a bit confused about the chart stating that the 3950 will not work with some boards in the X470 lineup including a couple that I have mentioned that have up to 2 8 PIN CPU and 12 phases.

Work work, every am4 motherboard with a bios support will work with a 3900x, the question is how stable it will hold that. For those boards I would recommend only a 3700x, 65w will be fine, even with an overclock, it will get to 90


Look at the Anandtech article. i7 9700K and i9-9900k. i9 has all the advantages (slightly higher clock and L3 cache) and still in some benchmarks (Gimp, Webtests, Photoscan, Ice Storm) the i7 scores better than the i9 because of no static partitioning and the highly single threaded nature of the tests (and no effect of 100 MHz more clock/ 4MB more L3 cache). So it's a general SMT/HT thing in some benchmarks and not something specific to AMD. Just that with AMD you get SMT nearly everywhere and with Intel its only on the largest and smallest SKUs for a few generations on desktop. So you don't buy an i9 and disable HT, because then you could just buy the i7. And you don't buy a Pentium and disable HT, because that would just suck. :D Since you can't get a 6C/6T CPU (or 8/8) out of the box for AMD, it makes sense to test it for some scenarios.

AMD is not Intel, intel fixed the smt few years ago so smt off on intel cpus is useless, now about amd is different and that review on gamer nexus showed us that smt off helps the 3900x a lot.
 
Last edited:
Joined
Nov 1, 2018
Messages
583 (0.30/day)
I'll just drop this here...
126512


https://geizhals.eu/ , one of the primary price indexers in Europe. That Top 10 is simply nuts...

(Before you ask, it's based on searches and product clicks)
 
Last edited:
Joined
May 8, 2018
Messages
1,494 (0.69/day)
Location
London, UK
one of the primary price indexers in Europe. That Top 10 is simply nuts...

(Before you ask, it's based on searches and product clicks)

That x570 aorus elite is the best cost effective motherboard for the price. Also from what I have seeing from reviews all around, the gain in performance from going from b450, x470 to x570 is 1%.
 
Last edited:
Joined
Dec 31, 2009
Messages
19,366 (3.72/day)
Benchmark Scores Faster than yours... I'd bet on it. :)
I'll just drop this here...
View attachment 126512

https://geizhals.eu/ , one of the primary price indexers in Europe. That Top 10 is simply nuts...

(Before you ask, it's based on searches and product clicks)
Gee, who would have thought immediately after a product release that the top 10 searches contain the new stuff.... lol

the gain in performance from going from b450, x470 to x570 is 1%.
AKA, none. 1% is margin of error. THere shouldn't be any reason the newer boards perform better due to the chipset. If the board will let the CPU run stock, performance should always be in that 1-2% margin of error.
 

Kalyori

New Member
Joined
Jul 11, 2019
Messages
1 (0.00/day)
That chart with the power info is super helpful.

I wanted to ask what takes priority when looking at a board?
Is it the number of phases, the doubler, or the High Side / Low Side FET?

Also what's the difference between 6+6 and just 12?

EDIT: I'm considering the 3900X and the Gigabyte Aorus Elite.
 
Joined
Apr 30, 2011
Messages
2,648 (0.56/day)
Location
Greece
Processor AMD Ryzen 5 5600@80W
Motherboard MSI B550 Tomahawk
Cooling ZALMAN CNPS9X OPTIMA
Memory 2*8GB PATRIOT PVS416G400C9K@3733MT_C16
Video Card(s) Sapphire Radeon RX 6750 XT Pulse 12GB
Storage Sandisk SSD 128GB, Kingston A2000 NVMe 1TB, Samsung F1 1TB, WD Black 10TB
Display(s) AOC 27G2U/BK IPS 144Hz
Case SHARKOON M25-W 7.1 BLACK
Audio Device(s) Realtek 7.1 onboard
Power Supply Seasonic Core GC 500W
Mouse Sharkoon SHARK Force Black
Keyboard Trust GXT280
Software Win 7 Ultimate 64bit/Win 10 pro 64bit/Manjaro Linux
For anyone who bought any of the Zen2 CPUs and cannot clock @ their official boost clocks check below (hint: it's caused by a BIOS default setting if the latest chipset driver is installed)

126753
 
Joined
Apr 30, 2011
Messages
2,648 (0.56/day)
Location
Greece
Processor AMD Ryzen 5 5600@80W
Motherboard MSI B550 Tomahawk
Cooling ZALMAN CNPS9X OPTIMA
Memory 2*8GB PATRIOT PVS416G400C9K@3733MT_C16
Video Card(s) Sapphire Radeon RX 6750 XT Pulse 12GB
Storage Sandisk SSD 128GB, Kingston A2000 NVMe 1TB, Samsung F1 1TB, WD Black 10TB
Display(s) AOC 27G2U/BK IPS 144Hz
Case SHARKOON M25-W 7.1 BLACK
Audio Device(s) Realtek 7.1 onboard
Power Supply Seasonic Core GC 500W
Mouse Sharkoon SHARK Force Black
Keyboard Trust GXT280
Software Win 7 Ultimate 64bit/Win 10 pro 64bit/Manjaro Linux
A great video for oc seperate ccx in Ryzen 3900X that applies to all Zen2 CPUs and other interesting things

 
Joined
Dec 31, 2009
Messages
19,366 (3.72/day)
Benchmark Scores Faster than yours... I'd bet on it. :)
For anyone who bought any of the Zen2 CPUs and cannot clock @ their official boost clocks check below (hint: it's caused by a BIOS default setting if the latest chipset driver is installed)

View attachment 126753
Wonder when we will see that chipset driver in the wild... Wasn't there yesterday. Anyone seen it today anywhere?

And he has a great point. . Why isn't that default???????????


EDIT: Latest one for the board I am working on is...... [19.10.16]

What is that he's listing?

Amazon.com best sellers in processors is just as brutal viewing for Intel
Yeah, brutal... brand spanking new inexpensive processors are out versus Intel who's platform has been out for a year... way to think it through.
 
Last edited:
Joined
Apr 30, 2011
Messages
2,648 (0.56/day)
Location
Greece
Processor AMD Ryzen 5 5600@80W
Motherboard MSI B550 Tomahawk
Cooling ZALMAN CNPS9X OPTIMA
Memory 2*8GB PATRIOT PVS416G400C9K@3733MT_C16
Video Card(s) Sapphire Radeon RX 6750 XT Pulse 12GB
Storage Sandisk SSD 128GB, Kingston A2000 NVMe 1TB, Samsung F1 1TB, WD Black 10TB
Display(s) AOC 27G2U/BK IPS 144Hz
Case SHARKOON M25-W 7.1 BLACK
Audio Device(s) Realtek 7.1 onboard
Power Supply Seasonic Core GC 500W
Mouse Sharkoon SHARK Force Black
Keyboard Trust GXT280
Software Win 7 Ultimate 64bit/Win 10 pro 64bit/Manjaro Linux
Wonder when we will see that chipset driver in the wild... Wasn't there yesterday. Anyone seen it today anywhere?

And he has a great point. . Why isn't that default???????????


EDIT: Latest one for the board I am working on is...... [19.10.16]

What is that he's listing?

Yeah, brutal... brand spanking new inexpensive processors are out versus Intel who's platform has been out for a year... way to think it through.
1563384423627.png
 
Joined
Oct 15, 2010
Messages
951 (0.19/day)
System Name Little Boy / New Guy
Processor AMD Ryzen 9 5900X / Intel Core I5 10400F
Motherboard Asrock X470 Taichi Ultimate / Asus H410M Prime
Cooling ARCTIC Liquid Freezer II 280 A-RGB / ARCTIC Freezer 34 eSports DUO
Memory TeamGroup Zeus 2x16GB 3200Mhz CL16 / Teamgroup 1x16GB 3000Mhz CL18
Video Card(s) Asrock Phantom RX 6800 XT 16GB / Asus RTX 3060 Ti 8GB DUAL Mini V2
Storage Patriot Viper VPN100 Nvme 1TB / OCZ Vertex 4 256GB Sata / Ultrastar 2TB / IronWolf 4TB / WD Red 8TB
Display(s) Compumax MF32C 144Hz QHD / ViewSonic OMNI 27 144Hz QHD
Case Phanteks Eclipse P400A / Montech X3 Mesh
Power Supply Aresgame 850W 80+ Gold / Aerocool 850W Plus bronze
Mouse Gigabyte Force M7 Thor
Keyboard Gigabyte Aivia K8100
Software Windows 10 Pro 64 Bits
Joined
Apr 30, 2011
Messages
2,648 (0.56/day)
Location
Greece
Processor AMD Ryzen 5 5600@80W
Motherboard MSI B550 Tomahawk
Cooling ZALMAN CNPS9X OPTIMA
Memory 2*8GB PATRIOT PVS416G400C9K@3733MT_C16
Video Card(s) Sapphire Radeon RX 6750 XT Pulse 12GB
Storage Sandisk SSD 128GB, Kingston A2000 NVMe 1TB, Samsung F1 1TB, WD Black 10TB
Display(s) AOC 27G2U/BK IPS 144Hz
Case SHARKOON M25-W 7.1 BLACK
Audio Device(s) Realtek 7.1 onboard
Power Supply Seasonic Core GC 500W
Mouse Sharkoon SHARK Force Black
Keyboard Trust GXT280
Software Win 7 Ultimate 64bit/Win 10 pro 64bit/Manjaro Linux
Joined
Feb 5, 2017
Messages
44 (0.02/day)
Location
North Carolina, USA
System Name ASUS Z97-PRO Gamer
Processor Intel Core i7-4790K (Delidded) with Rockit 88 100% Copper IHS
Motherboard ASUS Z97-PRO Gamer
Cooling Noctua NH-D15
Memory 32GB (8GB x4) GSkill TridentX DDR3 2400 (XMP Mode #1) Timings: 10-12-12-31
Video Card(s) EVGA 1070 FTW + ACX 3.0 (8GB GDDR5)
Storage 512GB Samsung 970 PRO (NVMe), 2TB WD Gold WD2005FBYZ-01YCBB1 SATA-3 Data drive
Display(s) LG 27UL500-W 27-Inch UHD (3840 x 2160) IPS Monitor with Radeon Freesync Technology and HDR10, White
Case Fractal Design Define R5 (Black, no Window)
Audio Device(s) Onboard Optical (TOSLINK) from MB to VIZIO soundbar
Power Supply EVGA Supernova G2 650
Mouse Logitech M215
Keyboard Logitech K330
Software Windows 10 Pro Secured by Emsisoft Internet Security & Malwarebytes Premium Version 3
Benchmark Scores http://www.userbenchmark.com/UserRun/2157372
At the conclusion of the review, the only con was no onboard graphics.

Some time back, onboard graphics was on the back of the MB OEM's. They took the lead of Intel at providing graphics with 1st gen 'i' series & didn't give us as much of a cent in savings & as a result of, reaped the profits for producing a lesser product at higher pricing. Sure, some has the ports for onboard (CPU) graphics, yet that's all. Audio is different, near the best ever, just no new Realtek driver since 2.82 & these were Vista certified. Basically, Realtek engineers have been sitting on their cans for a long time with audio drivers, leaving it to the modders (some on this very site) for us to get added features, such as Dolby Digital Live & the DTS variants. Or maybe the MB OEM, still they're confined by what they have to work with.

Seriously, anyone considering this type of CPU will most certainly be running a discrete card, even if it did have onboard GPU chip, which takes away from the raw CPU power. If only for the time being, one laying around, or current used in existing system while saving for a PCIe 4.0 model. NVIDIA jumped the gun in releasing the RTX 2000 lineup, knowing full well that PCIe 4.0 was coming & will be a far less successful line versus the GTX 1000 series. Am sure NVIDIA will come up with something, yet bargain hunters who won't be diving into PCIe 4.0 will be getting steep discounts on the RTX 2000 cards in a few months.

It's simply time to focus on CPU & GPU's as different components, even 4K (& 8K) TV OEM's does this, with the better models having one CPU & OS for the TV, another more powerful to render the picture. In an ideal world, would be best to have both for testing/setup. However this should be on the backs of the MB OEM's, as even the best of graphics combined with a CPU are inferior to cards released in 2012-13, as long as these are rated for 4K. Only advantage being HDMI 2.0 compliant for some onboard GPU chips, most still uses old school Displayport 1.2, hopefully the X570 MB's, or some, will offer DP 1.4 for the few Ryzen 3 chips with onboard graphics.

I'll be looking forward to the release of the Ryzen 9 3950X, by then, there'll be more PCIe 4.0 MB's & surely the Samsung 980 NVMe variants will be the same. Then we'll see the real speed of the PCIe 4.0 NVMe SSD. Maybe in hindsight, this is why I was able to snag a 970 PRO for half of the original price, Samsung knew what was coming & have been working on the real next gen NVMe.

Two-thirds of my X570 build cash is in the bank, thankfully this & other reviews has led me to totally ditch Intel when it comes to performance per dollar.:rockout:

That value goes into nothing if you buy a expensive 570x

In one respect yes, when it comes to the CPU upgrade alone.:)

However, to have the best PCIe options for everything connected, the X570 will be a must have.

Cat
 
Joined
Feb 4, 2013
Messages
16 (0.00/day)
Location
Australia
Hi W1zzard,

Just wondering if you could clarify why Rainbow Six Siege was excluded from the CPU gaming tests, when it seems to be used for the GPU gaming tests? The reason I ask is because there appears to be some significant performance regressions for the Ryzen 9 3900X with regards to minimum FPS within Rainbow Six Siege and possibly other gaming titles.

For example, when upgrading from the Ryzen 7 1800X, with all other factors (i.e. hardware and software) being equal, the Ryzen 9 3900X performed 13% worse. This was at 1080P with all visual settings set to low or off (aside from anti-aliasing, which was set to TAA 2X, and anisotropic filtering, which was set to 16X). AdoredTV experienced a similar outcome in his review, but the performance delta was smaller. Additionally, the 99th percentile results produced by Linus indicate that something is wrong.

Ultimately, I'm a bit disappointed that an $800 AUD processor is seemingly performing worse than a two year old processor (that wasn't a particularly good gaming processor to begin with) in gaming titles I frequent, and that no tech journalists/reviewers have even made mention of it. In the reviews by AdoredTV and Linus, even when the graphs indicated a problem, they skipped right over addressing it.

Any chance you could investigate?
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
26,956 (3.71/day)
Processor Ryzen 7 5700X
Memory 48 GB
Video Card(s) RTX 4080
Storage 2x HDD RAID 1, 3x M.2 NVMe
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
why Rainbow Six Siege was excluded from the CPU gaming tests, when it seems to be used for the GPU gaming tests?
No specific reason, I had to pick around 10 game tests to keep workload reasonable and that's what I ended up with
 
Joined
Feb 4, 2013
Messages
16 (0.00/day)
Location
Australia
No specific reason, I had to pick around 10 game tests to keep workload reasonable and that's what I ended up with.
Thank you for replying. Do you still have access to Rainbow Six Siege and the Ryzen 9 3900X, and if so, would you be able to verify if the performance metrics I've mentioned above are repeatable? My concern is that the poor performance within Rainbow Six Siege on low settings may be indicative of minimum FPS issues in other titles as well. Unfortunately, reviewers seemingly aren't using game settings other than high/ultra in their reviews (whether at AMD's request or because they believe it to be the most appropriate use case), so verification is difficult. I would have thought the Ryzen 9 3900X being outperformed by the Ryzen 7 1800X would be cause for concern, and that playing games at 1080P on low settings would be a common use case for gamers with high refresh rate monitors.
 
Joined
Feb 4, 2013
Messages
16 (0.00/day)
Location
Australia
The inaction regarding the above seems to indicate that TPU intends to turn a blind eye on any issues that may potentially reflect negatively on AMD. That's very disappointing, and I question if there is any integrity left in tech journalism. I'm curious, what is AMD's tactic that has you toeing the line? Incentives or threats? There is definitely an issue with the Ryzen 9 3900X regarding minimum FPS when it is outshined by the two generation old Ryzen 7 1800X, which wasn't a good gaming processor to begin with. You've lost a reader, as the regurgitation of press slides with no critical analysis and no further investigation beyond the initial review (such as examining unobtainable boost frequencies featured in misleading marketing materials and specification sheets) can be had literally everywhere else.
 
Last edited:

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
26,956 (3.71/day)
Processor Ryzen 7 5700X
Memory 48 GB
Video Card(s) RTX 4080
Storage 2x HDD RAID 1, 3x M.2 NVMe
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
The inaction regarding the above seems to indicate that TPU intends to turn a blind eye on any issues that may potentially reflect negatively on AMD. That's very disappointing, and I question if there is any integrity left in tech journalism. I'm curious, what is AMD's tactic that has you toeing the line? Incentives or threats? There is definitely an issue with the Ryzen 9 3900X regarding minimum FPS when it is outshined by the two generation old Ryzen 7 1800X, which wasn't a good gaming processor to begin with. You've lost a reader, as the regurgitation of press slides with no critical analysis and no further investigation beyond the initial review (such as examining unobtainable boost frequencies featured in misleading marketing materials and specification sheets) can be had literally everywhere else.
I've been working 18 hour days every single day since Zen 2 launch, there's just too much going on to follow every request/suggestion
 
Joined
Aug 27, 2019
Messages
1 (0.00/day)
Location
Camarillo, CA
Processor 3900x
Motherboard Asus Prime x570 Pro
Cooling Stock
Memory G.Skill 3200C14 Samsung B-Die
Video Card(s) Fury X
Storage Samsung Evo 860 1TB
Display(s) LG 2560 x 1080
Case HAF 912
Power Supply SeaSonic Focus 850
I've been working 18 hour days every single day since Zen 2 launch, there's just too much going on to follow every request/suggestion
Wizard,
Thank you for your review of the 3900x. I hope you are able to spend less time at work now.
Myself and others are trying to get our 3900x systems to Max Boost to 4.6Mhz on any Core with various tests. We haven't had much luck getting to those Mhz. Could you please let me know what test and what conditions were used with your results of 4.571. Also, what were you using to view the Mhz on each core?

If you are interested. I've started a topic at https://community.amd.com/thread/242812 if you would like to check out our results and/or contribute.

Thank you.
 

bad_sign

New Member
Joined
Oct 12, 2019
Messages
1 (0.00/day)
Guys you should update your vera crypt benchmarks. I get about 20GB/s AES en-decoding
veracrypt 1.23 with 1003 ABBA, GB X370 Gaming 5, 32GB 3733CL16, 3900X
Picture with internal benchmark
 

Attachments

  • Veracrypt 1.23.JPG
    Veracrypt 1.23.JPG
    75.5 KB · Views: 220
Top