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

AMD Ryzen 9 5900X

Joined
Apr 12, 2013
Messages
6,728 (1.68/day)
Anands used 2933 on their 10900K, and 3200 on their Zen 3 CPUs. No sane person with a 10700K or above is going to run DDR4-2933.
They use the highest supported official memory, it is what we call stock vs stock. Not to mention tuned memory results would show better scaling & gains for AMD. They are just as valid as TPU numbers or any other reputable source out there.
 
Joined
Aug 4, 2020
Messages
1,567 (1.18/day)
Location
::1
Just gotta wait for the CPU to run into its power limit. For temps we're reporting the long term stable state, not the single highest temp. Does that make sense?
Are the Intel temps with PLs enforced? Would be interesting to have numbers with PLs on and off, because especially for the 10900 and 10850 theres a lot of difference both in thermals and performance.

Something akin to this:
 
Joined
Jul 9, 2009
Messages
26 (0.00/day)
Biggest hardware site (io-tech) in the Nordic region tested with 3600 memory on all platforms and 10900k got beaten easily. RTX 3090 GPU was used for gaming benchmarks.

Article is in finnish, but you can check out the graphs. There's also very interesting memory and IF scaling results.

Game results for e.g. SoTR and W3 should be wildly in favor of new Ryzen CPU's so something is definitely wrong with TPU's testing setup. Might be good idea to verify results with another motherboard for each system. Also please try RTX 3080/3090 GPU's if that turns the tables..

Btw. is TPU planning to add 1% lows or minimum fps at some point to benchmarks?
Shouldn't be too much extra work as pretty much any benchmarking tool will give you also that data.
 
Joined
Apr 12, 2013
Messages
6,728 (1.68/day)
Biggest hardware site (io-tech) in the Nordic region tested with 3600 memory on all platforms and 10900k got beaten easily.
Well people forget that faster memory on Intel will net them minimal to no gains beyong 3200MHz, except perhaps with much tighter timings. On AMD you get near perfect scaling with appropriate latency paired with higher (memory) clocks, basically going from 2933MHz (max rated officially for Intel) to DDR4 3600 or DDR4 4000 may net you 5~10% more performance but on AMD it can easily do 10~20% depending on games or application.
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
26,957 (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
Are the Intel temps with PLs enforced?
Yes of course. Also for all other data in the Ryzen reviews.

Like this?

Btw. is TPU planning to add 1% lows or minimum fps at some point to benchmarks?
Next round of rebench, early 2021, as soon as I have two or three weeks without launches, so I can figure out new tests, games, methodology, set up everything and retest 20 or 30 CPUs
 
Joined
Mar 21, 2020
Messages
77 (0.05/day)
naaah..
5000 ryzen might be good cpu,yes...hmm, but it cant offer anything better performance for gaming than 10900 ..(and another way i dont care ) like we seen test.
10900 is still lead over FHD result gaming.
also, i wonder alot, temperature is almost same with amd 7nm brand new ryzen cpu when verify 14nm intel 10900k cpus....i repeat 7nm ryzen, 14nm intel

we must remembe, amd 5000 series is 7nm tech maded,just release amd latest cpu, intel 10900 'old' 14nm tech maded

i mean, its so tricky to compare totally different tech cpus, aka amd 7nm cpu for intels 14nm cpu...handicap advance is so massive.
i should seen much more clear bigger difference...but no.


shortly more, if intel have even 10nm cpu against ryzen 7nm ,then we seen something for that question, is it ryzen 7nm cpus good.
we cant say it yet....well, sure 7nm ryzen get better scores for 14nm cpu,it shouls do, but.. i really not hurray it much.

well, incoming ,still 14nm cpu intel roclet lake is is direct answer amd 5000 series ryzen,and what for rumors leaks shows, not look good amd way.


finaly.
i think we get answer for question at least next year, june/2021,when intel 1st 10nm cpu coming,(sure now, btw),bcoz then we can compare little bit that question of amd ryzen quality.

so next target will be:
--------------------------------------------------------------------------------------------------------------
14nm intel rocket lake-s VS 7nm amd ryzen 5000 series, march(2021
10nm intel adler lake hybrid cpu VS 7nm amd ryzen 5000 series, june/2021 <-- this is it, total battle!
--------------------------------------------------------------------------------------------------------------
so, i say and im very very confident, we see it finaly next summer.


intel 7nm meteor lake-s is going 2022 or so...but truth is that after that we seen real test, which one make better cpus...
but,i think intel 10nm hydrib cpu must do it clear...and rocket lake-s start it.

d-day is for cpu world peformance and efficiency is clear next summer,and if ryzen can offer better performance and efficiency than adler lake-s, its 100% clear. not yet
 
Joined
Jul 19, 2016
Messages
476 (0.17/day)
Your gaming numbers are very different to all the other reviews I have seen out there i.e you have Ryzen 5000 slower than elsewhere. I think there may be an error in your testing akin to Anandtech's when they were blasted for their incorrect Ryzen 2000 benches.
 
Joined
Apr 12, 2013
Messages
6,728 (1.68/day)
we must remembe, amd 5000 series is 7nm tech maded,just release amd latest cpu, intel 10900 'old' 14nm tech maded
We must also remember it's 14nm++ not 14nm & TSMC's 7nm < Intel 10 nm, at least that's what Intel fans & executives tell us :rolleyes:

Beyond that an OCed 10900k @5.5GHz will obviously lead in gaming possibly even crushing 8c RKL, it also crushes everything else in power consumption though :nutkick:
 

Mike2Fr

New Member
Joined
Aug 14, 2020
Messages
10 (0.01/day)
Absolutely. Every reviewer has his own preferences and ways of doing things, and that's good, because it gives you more data to base your buying decision on
I don't think there is anything wrong in your review man.

To understand why you have differences with other reviews especially in gaming, I would try to compare your Max clock on the first core on different cpus (if you have the luxury to have more than one) I have seen some reviewers reaching 5ghz+ where you are at 4.85 ish.

Memory obviously, this is an enthusiast market. I understand your choice of memory (that is far from being as bad as some people reported here) but I want to know the top perf not what the average dude will produce. it is like willing to know the 0 to 60 mph of cars even though nobody drives like that all the time. (and I am not talking about LN2 here... It is meaningful for a handful of people on this planet. But I am pretty sure you got my point.) I think you are already working on memory scaling so thank you for that.

The bios version Cpu/Mb combo. Most of the test I read are based on Asus/Msi top Motherboard. There might be a small issue here as well.


Thank you for the hard work. I was a little bit surprised by your results but I am sure they are 100% honest. I am sure we will have a better understanding in few weeks.
 
Joined
Dec 4, 2013
Messages
1 (0.00/day)
Hello everyone! Can anyone confirm if 5000 series Ryzen support ECC RAM (combined with a compatible Motherboard) ? i have seen mixed info on that...
Thank you in advance!
 
Joined
Aug 4, 2020
Messages
1,567 (1.18/day)
Location
::1
Yes of course. Also for all other data in the Ryzen reviews.

Like this?
[ ... ]
Yes!
Must've missed/forgot about that for some reason, probably because you only link the stock values otherwise.

Hello everyone! Can anyone confirm if 5000 series Ryzen support ECC RAM (combined with a compatible Motherboard) ? i have seen mixed info on that...
Thank you in advance!
Since Matisse supports ECC and Vermeer is recycling the same old I/O die, I'd assume so ...
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
26,957 (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
Joined
Jun 23, 2014
Messages
1,374 (0.39/day)
Processor 3900X 4.425
Motherboard X570 Tomahawk
Cooling Galahad 360 push-pull
Memory 2x16gb Crucial Ballistix MAX 4400
Video Card(s) Asus Dual 3060 Ti OC
Storage Optane 280gb PCI-E
Display(s) PG348Q
Case Core X71
Audio Device(s) ATOLL DAC 100SE, Sony DN1070 - Dali Ikon 1 MK2, Presonus Studio 192, Line 6 POD HD rack, Audix VX10
Power Supply AX1500i
Mouse Pulsar Xlite wireless white
Keyboard Leopold FC980C 30g white
Software Win 10 Pro
I don't have pricing for the 16 GB kits anymore, but here are the ones I was looking at:

G.SKILL Trident Z Neo - DDR4 3200 - 14-14-14-34 (Samsung B-die) - $224.99

G.SKILL Trident Z Neo - DDR4 3600 - 16-19-19-39 (Hynix CJR/CFR) - $173.98

While not a big $50 gap, it is still a significant cost. This is taking into fact just sticking with stock/XMP/DOCP settings (for the average consumer). Obviously if one is experienced, it would be best to go with the 3200CL14 and overclock it.

Also there is the fact that those who would be upgrading to any of the Ryzen gen. 4 CPUs (like from Coffee Lake or Zen 1/1.5) would most likely be using their previous RAM, which could be similar to the ones I mentioned above.

Why don't you have pricing for the 16 GB kits no more? It's not like 32 GB is a MUST with the CPU. It works completely fine and even gets to high loads easily with 16. I'm just going to PCPP for a bit because you definitely picked overpriced RAM.



4000 E-Die and 3600 B-Die, both better specced than yours, both give you better price / performance.
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
26,957 (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
Oh, absolutely. It just must've slipped my mind because of it. :)
No worries, often some nice improvements come out of feedback like this
 
Joined
Feb 26, 2016
Messages
546 (0.18/day)
Location
Texas
System Name O-Clock
Processor Intel Core i9-9900K @ 52x/49x 8c8t
Motherboard ASUS Maximus XI Gene
Cooling Corsair H170i Elite Cappelix w/ NF-A14 iPPC IP67 fans
Memory 2x16GB G.Skill TridentZ @3900 MHz CL16
Video Card(s) EVGA RTX 2080 Ti XC Black
Storage Samsung 983 ZET 960GB, 2x WD SN850X 4TB
Display(s) Asus VG259QM
Case Corsair 900D
Audio Device(s) beyerdynamic DT 990 600Ω, Asus SupremeFX Hi-Fi 5.25", Elgato Wave 1
Power Supply EVGA 1600 T2 w/ NF-A14 iPPC IP67 fan
Mouse Logitech G403 Wireless (PMW3366)
Keyboard Logitech G910 Stickerbombed
Software Windows 10 Pro 64 bit
Benchmark Scores https://hwbot.org/search/submissions/permalink?userId=92615&cpuId=5773
My questions:
Rendering
CB20 1T: How does the 8700K beat the 9900K? The 9900K has faster turbo speeds...

Software & Game Dev
UE4: How does a Ryzen 5 3600 beat both the X and XT versions?

Web Browser
WebXPRT: A 9900KS beats a 10900K??

Office & Productivity
Excel: Please tell me how the 10900 beats the 10900K. The whole benchmark takes less than a second, TDP cannot be an issue here.
Adobe Premiere: 9900KS beats a 10900K, even though it has 2 less cores and has lower frequencies?

Server & Workstation
VMWare: You are telling me that, a 5600X will about the same performance with the 5900X? The 5600X has half the cores and threads... why would that match a CPU that has the same architecture, and double the cores and threads?

Compression & Encryption
WinRAR 1.5GB: A 3700X beats a 3800XT?
VeraCrypt: 10900 beats the 10900K again..? And um, an 1800X beats a 3700X (not to mention also the 2700X and 5800X)... ok. There are so many anomalies in this one.

Media Encoding
H.265: 10600K beats an i7-10700, even though cores actually matter for this kind of benchmark... And how does a 9700K match a 10900? The 10900 has more than double the thread count..
MP3: I kind of expected the 9700K to beat the 8700K if it is a single threaded workload, as the 9700K doesn't have HT (having this enabled will result in slightly lower performance/thread), and it has 200 MHz higher 1c turbo, which cannot be considered "margin of error".

720p
AC Odyssey: 10700 beats 10700K? 10900 beats 10900K? 10600K beats 9900K? 3600X beats 3600XT? 2700X beats 3600? 3700X beats 3800XT? 3900XT beats 5900X? 5600X beats 5800X?
Battlefield V: 10600K beats 9900K?
Far Cry 5: 3900X beats 3900XT? 10700 beats 10700K?
Metro Exodus: 3600 beats 3900X and 3900XT? 10600K beats 10900 and 9900KS?
Rage 2: 10600K beats 10900K? 3600XT loses to both 3600 and 3600X? 2700X beats 3700X? 3900X beats 3900XT?
Sekiro: 10700 beats 9900K?
The Witcher 3: 1800X beats 2700X? 3600X beats 3600XT? 3900X beats 3900XT? 10600K beats 9900K?
Wolfenstein II: 3900X beats 3900XT?

1080p
AC Odyssey: 3800XT loses to 3600, 2700X, and 3700X? 3900X beats 3900XT? 3600X beats 5600X and 5800X? 10900K loses to 8700K, 9900KS, 10900, 9900K, 10700K, and 10700?? 10700 beats 10700K?
Battlefield V: 3600 beats 3600X? 3700X beats 3800XT? 5600X beats 5800X? 10700 and 10600K both beat 10700K, 10900K, 9900K, and 9900KS??
Civilization 6: 10700 beats 10700K and 10900K? 10900 beats 10900K?
Far Cry 5: 3900X beats 3900XT? 10700 beats 10700K?
Metro Exodus: 3600XT beats 3800XT? 8700K beats 9900K and 9900KS? 10600K beats 10900, 9900KS, and 9900K? 10700 beats 10700K,10900, 9900KS, and 9900K?
Rage 2: 3600X beats 3600XT? 3600 beats 3600X and 3600XT? 10600K beats 10700K and 9900K? 8700K beats 9900K? 10700 beats 10700K, 10900, 9900KS, and 9900K? 5600X beats 5800X? 3800XT beats 5800X?
Sekiro: 10600K beats 9900K?
Shadow of the Tomb Raider: 10700 and 10600K both beat 10700K, 10900K, 9900K, and 9900KS? 3600 and 3600X both beat 3700X and 3800XT? 3600XT beats 5600X and 5800X?
The Witcher 3: 3700X and 3800XT both lose to 3600, 3600X, and 3600XT? 5600X beats 5800X? 10700 and 10600K both beat 10900, 10700K, 9900K, and 9900KS?
Wolfenstein II: 3600 beats 3600X? 3700X and 3600XT both beat 3800XT? 5600X beats 5800X? 10700 beats 10600K, 10700K, 9900K, and 9900KS?

1440p
AC Odyssey: 1800X beats 2700X? 3600 beats 3600X and 3600XT? 3600X beats 3600XT? 3800XT and 5600X both beat 5800X? 8700K beats 10900, 10600K, and 10700K? 9900KS beats 10900K? 10700 beats 10700, 10700K, and 9900K?
Battlefield V: 3600 beats 3600X? 3900X beats 3900XT? 10700 beats 9900K, 9900KS, 10700K and 10900K? 8700K beats 9900K, 9900KS, 10900K?? 9900K beats 10900K and 9900KS?

ok im getting a bit tired after typing this out for over 2 hours, you get the point.

TLDR, I guarantee you your testing methodology is flawed. Did you force the GPU to be at a constant frequency in all tests? That alone could fix the disparities between most of the tests and get rid of the "margin of error" argument that I always get in response. Why did I spend 2+ hours typing this up? Because y'all spent days, if not weeks testing all of these processors, and it would not be very fair of me to just say "these results don't make sense" without any explanation, so I have shown a lot of instances where there are anomalies. Clock speed is performance, and a CPU on the same architecture with a higher clock speed should outperform a lower clocked CPU from the same generation, as long as core counts are the same. It makes zero sense for a 6c12t CPU at 3.5 GHz to beat a 6c12t CPU at 4.5 GHz provided they are the same chip. But yes, we do get the point, Ryzen 5000 doesn't actually live up to the hype.

Enough about the negatives, here are some things I like about these reviews:
  1. You don't just compare power consumption, you also compare the total amount of energy consumed for a benchmark, which I think is EXTREMELY practical.
  2. You actually show us the results in an easy to read (and organized) format.
  3. The relative performance is included
  4. You tested 720p, 1080p, 1440p, AND 4K. Not a lot of [well-known] reviewers do more than 2 resolutions.
Also, instead of putting the charts as images and showing the data that way, is it possible to embed the data in the post? I am asking this because I want to press ctrl+f and find a CPU that way, but I cannot do that since it is an image, and you can't "grab text" from an image like that.

Anyways, thank you for taking the time to make this review, I appreciate all the time and effort y'all put into making these reviews!
 
Joined
Oct 10, 2018
Messages
943 (0.47/day)
So at the end what is the optimum memory speed and timings for ryzen 5xxx series? not asking for the best but optimum with price in mind.
 
Joined
Dec 31, 2009
Messages
19,366 (3.72/day)
Benchmark Scores Faster than yours... I'd bet on it. :)
So at the end what is the optimum memory speed and timings for ryzen 5xxx series? not asking for the best but optimum with price in mind.
Nothing has changed it seems... same IO die, so 3600/3733.
 
Joined
Oct 30, 2020
Messages
48 (0.04/day)
Well people forget that faster memory on Intel will net them minimal to no gains beyong 3200MHz, except perhaps with much tighter timings. On AMD you get near perfect scaling with appropriate latency paired with higher (memory) clocks, basically going from 2933MHz (max rated officially for Intel) to DDR4 3600 or DDR4 4000 may net you 5~10% more performance but on AMD it can easily do 10~20% depending on games or application.

And tuning via DRAM calculator yields massive gains as well.

Nothing has changed it seems... same IO die, so 3600/3733.

Most reports are that 4000 is doable 1:1 with the new ryzens. Same I/O die but due to other changes the IF seems to scale 1:1 much better. Would be interesting to see tuned 4000 1:1 performance.
 
Joined
Dec 31, 2009
Messages
19,366 (3.72/day)
Benchmark Scores Faster than yours... I'd bet on it. :)
Most reports are that 4000 is doable 1:1 with the new ryzens. Same I/O die but due to other changes the IF seems to scale 1:1 much better. Would be interesting to see tuned 4000 1:1 performance.
Right... but the dude asked about optimum speed and timings with price in mind. ;)
 
Joined
Jun 23, 2014
Messages
1,374 (0.39/day)
Processor 3900X 4.425
Motherboard X570 Tomahawk
Cooling Galahad 360 push-pull
Memory 2x16gb Crucial Ballistix MAX 4400
Video Card(s) Asus Dual 3060 Ti OC
Storage Optane 280gb PCI-E
Display(s) PG348Q
Case Core X71
Audio Device(s) ATOLL DAC 100SE, Sony DN1070 - Dali Ikon 1 MK2, Presonus Studio 192, Line 6 POD HD rack, Audix VX10
Power Supply AX1500i
Mouse Pulsar Xlite wireless white
Keyboard Leopold FC980C 30g white
Software Win 10 Pro
If you're buying 16, 4000 to 3600 is not all that big of a problem since the 3600 Elites cost 80 and 100 for the 4000 ones. Things look worse on the 32 end though.
 
Joined
Jan 8, 2009
Messages
548 (0.10/day)
System Name AMD RyZen PC
Processor AMD RyZen 5950x
Motherboard ASUS Crosshair VIII Hero 570x WIFI
Cooling Custom Loop
Memory 64GB G.Skill Trident Z DDR4 3200 MHz 14C x4
Video Card(s) Evga 3080 TI
Storage Seagate 8TB + 3TB + 4TB + 2TB external + 512 Samsung 980
Display(s) LG 4K 144Hz 27GN950-B
Case Thermaltake CA-1F8-00M1WN-02 Core X71 Tempered Glass Edition Black
Audio Device(s) XI-FI 8.1
Power Supply EVGA 700W
Mouse Microsoft
Keyboard Microsoft
Software Windows 10 x64 Pro
AnandTech data is not matching Tec powerup.. any idea why ?


 
Joined
Mar 18, 2015
Messages
2,960 (0.90/day)
Location
Long Island
For ya next Gaming Build what's the value proposition ? ...... the $174 10400F .... or ...... $550 5900 XT

 
Top