With third Gen Ryzen it has become pretty clear that the Tech Media and Tech YouTubers are either too lazy - or too stupid - to be able to configure the CPU correctly.
I got Ashes of the Singularity: Escalation in the SteamSale and decided to see what difference actually being able to configure a Ryzen 3rd Gen CPU would do with regard to the built in benchmark (set to "Crazy" at 1080p) compared to what the numb-nuts of the Tech Media managed to fumble their way to bringing to print (or voice).
Let's look at what the likes of KitGuru or Tom's Hardware managed to produce with their FREEBIE 3950X compared to what I managed to achieve with my "sample" which I had to pay full retail for.
I really don't know (though I can suspect) that the "Reviewers" have put in a lot of work to configure the Intel CPUs and have gone out of their way to nerf AMD CPUs; I will leave it up to the reader of the following to make that judgement for themselves.
Here are the two comparison lists from KitGuru and Tom's Hardware:
Now let's see where the 3950X stacks up when in the hands (system) of someone who actually knows what they are doing, namely myself.
I would ask the reader to bear in mind that my ambient temperature is around 32 degrees Celsius and Third Gen Ryzen is sensitive to temperature.
Let's start off with running the benchmark, and I repeat, I am using exactly the same game settings as the two above, with SMT on running 16 Cores/32 Threads:
The value which is the one being tested is the second one listed, namely the "Average CPU Framerate (All Batches)" and as you can see the score is 60.7 FPS.
Not only is that score higher than the scores the Tech Media managed to achieve with their FREEBIE 3950X, but also higher than ANY INTEL CPU.
So let's see what happens when you turn off SMT and run the 3950X as a straight 16 Core/16 Thread CPU:
The score actually increased to 64.2 FPS and this is due to the fact that the CPU didn't get as hot as it did in the first run with 16C/32T.As you can see, my score is again higher than ANY INTEL CPU.
My friend, who also has a 3950X, which he paid for, and I configured, got a score of 64.7 FPS running it with 16C/32T but his ambient temperature is eight degrees Celsius lower than mine.
So what kind of result can you expect to get from the new 3800XT?
It won't be the score that the reviewers get, because, as I said, they are either too stupid or too lazy to configure the Third Gen Ryzen CPU correctly, and AMD will send those out as FREEBIES to those incompetents and of course AMD will look bad in comparison with "The Best Gaming CPU", namely Intel.
I know what kind of result I can get from a 3800XT because I effectively have one. To make my 3950X into a 3800XT all I have to do is turn off one CCD and hey presto, instant 8 Core/16 Thread 3800XT (which seems to have escaped the "geniuses" of the Tech Media and the Tech YouTubers who have been rampantly speculating as to what performance can be expected).
Of course I won't be sent a FREEBIE from AMD to confirm my prediction, so the public will be left with one more review cycle of a nerfed AMD CPU compared to Intel.
Without further ado, here is the result when running the benchmark with 8 Cores/16 Threads:
This time around the score dropped under 60 to 58.2 FPS; however even at that, it is still better than ANY INTEL CPU.
This result adds insult to injury with regard to the sorry showing of the Tech Media when it came to reviewing the 3950X and it is for this reason that I would strongly suggest that AMD should not send out review samples of their Fourth Gen Ryzen CPUS until these people actually learn how to configure the damn thing.
The results I achieved were with the following configuration:
16C/32T: All Core clockspeed 4.3 GHz with a maximum of 1.3 Volts
16C/32T: All Core clockspeed 4.4 GHz with a maximum of 1.3 Volts
8C/16T: All Core clockspeed 4.5 GHz with a maximum of 1.3 Volts
I hope that AMD will use this post to shame the Tech Media into finally doing their job and learning how to properly configure the Ryzen CPUs; but I fear that they won't and when the new generation of Ryzen CPUs are launched, the Tech Media and YouTubers will once again nerf them and then trundle out the old trope of "Intel still has the best Gaming CPU".
People such as myself will have to wait, and then of course buy, the next Generation of AMD Ryzen CPUs and by the time I get around to being able to publish anything, the damage will have been done in the Media and on YouTube, and AMD will once again have lost out on sales it could have made by being declared in second place and inferior to Intel in the important Gaming Sector, unfairly, once again.
Are the Media and YouTubers out to harm AMD?
I think that one should apply Hanlon's Razor to this question which states:
"Never attribute to malice that which is adequately explained by stupidity".
You are comparing overclocking results to out of the box results.
Here is the spec.
https://www.amd.com/en/products/cpu/amd-ryzen-9-3950x
Of course you will get better results if you overclock.
Reviewers do not use overclocked numbers in benchmarks because overclocking is out of manufacturer specification and results vary from die to die.
The only problem with that argument is that I don't overclock.
I actually stay within spec.
What gives lie to your argument is the use of the terms "OC", "MCT" and "PBO" - all of these terms denote some form of overclocking - for a lot of the entries in the tables I referred to and compared my results to.
So I don't know where you are getting the idea from that I am comparing my results to "Out of the box" results.
OC = OverClocked
MCT= Multi Core Turbo
PBO = Precision Boost Overdrive
You said that you have done this:
8C/16T: All Core clockspeed 4.5 GHz with a maximum of 1.3 Volts
I think that means you are running all 8 core 16 thread at a Base Clock Speed of 4.5 GHz.
The specification for the processor is Base Clock Speed = 3.5GHz.
That means you are overclocking your CPU.
I have no idea what you thought you were trying to say with this post.
I have asked others, and they don't know either.
The only thing I can remotely think of is that you are trying to make a reference to TDP and have mixed up how Intel and AMD measure TDP.
Intel measures the TDP by the base clock of the CPU, AMD has a completely different and very weird (to the point of being fickle) way of measuring TDP part of which is dependant on the cooling solution being employed.
Needless to say I am not running the CPU outside of the TDP envelope when I have crippled my 3950X to 1 CCD.
RE: I have no idea what you thought you were trying to say with this post.
I have asked others, and they don't know either.
It is pretty easy - go look at your BIOS settings.
On Ryzen 2700X and ASUS ROG Crosshair Hero VII there is a "multiplier" figure.
I don't configure the CPU in the BIOS.
It is at "Auto" i.e. 35.00 in the BIOS.
The only thing I configure in the BIOS are values related to the motherboard (such as LLC etc.)
So are you using Ryzen Master to set all cores to run at 8C/16T: All Core clockspeed 4.5 GHz with a maximum of 1.3 Volts then?
Or maybe Asus AISUITE III or something else to set the cores to run at 4.5 GHz?
I am using the BIOS to limit the 3950X to one CCD
I am then using Ryzen Master to then configure that one CCD.
Ryzen 2700X.
54 FPS.
Pareto Principle.
About 20% of the cost gets at least 80% of the results:
I guess you shouldn't have bought a Ryzen 3950X just to play Ashes of the Singularity at 1080p.
You do know what the point of a benchmark is, don't you?
Now you are just trolling
RE:
You do know what the point of a benchmark is, don't you?
Now you are just trolling
What is wrong,
Are you upset that a Ryzen 2700X with 8 cores 16 threads scores 54, without me even trying, wheras you score 60 with a R9 3950X 16 core 32 thread using your "special knowledge"?
Shouldn't you be hitting at least 108 FPS with twice the number of cores and threads, 7nm instead of 12nm process and an IPC improvement?
You have claimed that all reviewers are either stupid or are deliberately trying to make AMD look bad.
Reviewers probably followed AMD own benchmarking recommendations when they get early hardware from AMD.
Perhaps it was AMD who requested the reviewers did not try to overclock the initial 3950X samples, but just relied on "PBO" for example.
There are so many factors that could affect performance at 1080p, including motherboard, motherboard BIOS version, motherboard BIOS settings, Windows 10 version, Windows 10 processes running in the background, Benchmark Version, GPU used, GPU Driver Version etc etc.
"Correction - "PBO"
The purpose of a benchmark is to check the efficacy and stability of the settings one has configured.
I got Ashes of the Singularity: Escalation because it was on the current SteamSale and I was interested to see what result the benchmark would produce.
It uses more cores than a game normally does and I was interested to see if running it with 16 Cores/32 Threads would run better than 16 Cores/16 Threads for instance.
I was also interested to see what the difference would be running it with 16 Cores/16 Threads and then running with only one CCD with 8 Cores/16 Threads.
It's pretty bloody silly to suggest that I bought a 3950X to run AotS; however you can see the generational uplift in performance from Ryzen 2nd Gen to Ryzen 3rd Gen from my 8 Core/16 Thread test.
I ran the test at 1080p because that is the resolution that everyone else ran their tests at.
If I just wanted to game I would have stuck with the 3600X and the X470 Motherboard I originally bought last June. For the games I play, I don't need a high refresh monitor and 60 Hz does me just fine.
The thing is that "Reviewers" go out of their way to nerf their Ryzen 3rd Gen CPU tests. One of the favourite ones is to run their tests with 3200 MHz CL14 RAM (on Intel there is not a lot of performance uplift from 3200 CL14 to 3600 CL16, but of course with the Infinity Fabric of 3rd Gen Ryzen being so vital to the performance of the CPU it is very much a stealth nerf when you cripple the performance of a vital component of the CPU by at least 12.5% and try to put that forward as a "fair" comparison).
Unlike the reviewers however I run my CPU within spec, and because AMD is very vague about the maximum voltage to run the CPU at I stick with the TSMC spec for their 7nm node of 1.3 Volts as a maximum.
I could bump my clockspeed a lot higher if I were willing to shove in the voltage that is applied when for instance PBO is activated.
The thing is, with regard to your other mewling, that my friend who also has a 3950X which I have configured was running a lot more tasks than I when running the AotS benchmark, including Discord which he was using to chat with me, and his CPU score was HIGHER than mine, which is due to the fact that his ambient temperature is a lot lower than mine (at least eight degrees Celsius).
He has a different motherboard, different RAM and a different graphics card. The only common denominator is that I have configured both of our systems.
Here is his benchmark score:
I have to keep my ambient temperature a lot higher than normal because I have had two spine operations and have spinal arthritis.
Are there any other excuses you would care to make?
RE:
The thing is that "Reviewers" go out of their way to nerf their Ryzen 3rd Gen CPU tests. One of the favourite ones is to run their tests with 3200 MHz CL14 RAM (on Intel there is not a lot of performance uplift from 3200 CL14 to 3600 CL16, but of course with the Infinity Fabric of 3rd Gen Ryzen being so vital to the performance of the CPU it is very much a stealth nerf when you cripple the performance of a vital component of the CPU by at least 12.5% and try to put that forward as a "fair" comparison).
Ryzen 3950X Specification: https://www.amd.com/en/products/cpu/amd-ryzen-9-3950x
System Memory System Memory Specification 3200MHz DDR4
RE: Are there any other excuses you would care to make?
You are the one who needs to make exuses for such poor overclocking result on an next gen CPU with twice the core count.
3200 (or 1600 Data Rate) is the highest JEDEC specification for DDR4RAM, that's why AMD states 3200.
AMD does however recommend 3600 (or 1800 Data Rate) RAM.
That's a little bit of info you failed to mention.
Once again, as I have stated numerous times, I do not overclock my CPU, I stay within the specs.
If you care to put up your CineBench R20 result that you achieve with your 2700X then we will see just how poorly my score of 10,170 is in comparison.
I would say you get maybe 4,100?
So you are using Ryzen Master to overclock.
This: - is overclocking as a starter for 10.
"i) Click in the first field beside "C 01" and change the clockspeed. You should have absolutely no problems setting it to "4250". When you have done the rest of the configuration then test it and increase it (in my case it is set to 4300 and I have no problems). When you set one field, because the Green X is activated, all the other values will change to what you set."
RE:"
AMD does however recommend 3600 (or 1800 Data Rate) RAM.
That's a little bit of info you failed to mention."
Where does it say that on the specification for the CPU? It says 3200MHz.
Sure there are some AMD motherboards that claim to allow running Ryzen CPUs at higher RAM Speeds.
However RAM Compatibility can be a nightmare, and hitting those speeds often provides marginal benefit.
I know that Ryzen 2700X memory controller had lots of problems and the practical limit is 3200MHz with 4 sticks of 16GB RAM to give 64GB Ram in total.
If I run with a pair of 16GB sticks I can hit slightly higher RAM speed.
Ryzen 3600 memory controller seems better w.r.t. Ram Compatibility, based on builds I have done., and maybe that is the case on 3950X as well.
How much performance improvement do you expect to see running Ashes on 3600MHz Ram versus 3200MHz?
Maybe 1% improvement in FPS?
In that case perhaps you would be interested in seeing AMD launch a 3950XT then?
Which part of "JEDEC specification for DDR4 RAM" do you not understand?
For the 3950X AMD states System Memory Specification: 3200 MHz
For your 2700X AMD states System Memory Specification: 2933 MHz
What do you think the word "Specification" refers to?
You are getting into The Verge PC Build region of cognitively redacted now, or, to paraphrase from the movie Pulp Fiction, "Techie M**********r, do you speak it?"
AMD will not be launching a 3950XT and if you had actually comprehended my original post then you would know why.
You need to make up your mind about if you want to use XMP profiles to overclock your RAM or not.
You claimed you are not overclocking your 3950X, even though you are overclocking it using Ryzen Master.
Since you lose the discussion you start to throw insults.
I pointed out to you for the 3950X AMD states System Memory Specification: 3200 MHz.
In other words, anything higher than that is an overclock.
In response you said earlier:
"3200 (or 1600 Data Rate) is the highest JEDEC specification for DDR4RAM, that's why AMD states 3200.
AMD does however recommend 3600 (or 1800 Data Rate) RAM.
That's a little bit of info you failed to mention."
So make up your mind, are you overclocking your system or not?
I know what the base recommended memory speed is for Ryzen 2700X. I own one.
I also know how difficult it is to get stable memory overclock using XMP Profiles with Ryzen 2700X.
I have a stable system running Ryzen 2700X at 4.3GHz all cores and 64GB of RAM running at 3200MHz and it passes MemTest86 and Karhu RamTest in Windows 10.
If you know anything, you will know that is pretty much the limit for Ryzen 2700X RAM performance.
I managed that by tweaking motherboard BIOS settings and sending the data to ASUS for BIOS updates.
RE: AMD will not be launching a 3950XT and if you had actually comprehended my original post then you would know why.
I comprehend you do not know what overclocking means.
It is not overclocking when I am running the CPU within spec and optimising the CPU whilst staying with that spec. In spec meaning staying within the voltage and power envelope, which I have already stated and you obviously STILL have not comprehended.
The CPU spec states a base all core clockspeed and a maximum single core boost - that's it. Whatever all core clockspeed I can squeeze out of the voltage and power limits is, by definition, NOT overclocking.
I had to get a Ryzen 1st Gen 1200 for my X470 board to upgrade it to a BIOS which would recognise the 3600X and it ran very happily at 3600 CL16 (4 x 8GB).
If you really need to run dual rank RAM then you are a lot more limited in choice.
You are getting to the stage where you are just wasting my time with your ignorance.
Yawn.
Looks like this is the Tom#s Hardware Review:
AMD Ryzen 9 3950X Review: 16 Cores Muscles Into the Mainstream | Tom's Hardware
There have been a couple of updates to Ashes of the Singularity since that review.
Have you looked at the Patch Notes?
Of course you haven't, if you had, then you would not have posted something so inane.
Yes I have looked at the patch notes for Ashes of the Singularity.
There have been 2 updates since that review.
I think this is the KitGuru review you mention:
https://www.kitguru.net/components/cpu/luke-hill/amd-ryzen-9-3950x-cpu-review/7/
They were very positive about the 3950X in the review, as they should be.
If that is the correct review, then according to this, they set a manual overclock that would pass the most stringent test:
https://www.kitguru.net/components/cpu/luke-hill/amd-ryzen-9-3950x-cpu-review/2/
And of course it shows their incompetence, because to achieve that OC they punted in 1.35 Volts and were shoving 228 Watts into their 3950X with a temperature of 96.26 degrees Celsius
I on the other hand achieve a stable 4.3 GHz with 1.3 Volts and only have 158 Watts going into my 3950X with the CPU temperature even with my higher ambient temperature, just going over 80 degrees Celsius. My friend running CineBench R20 at 4.3 GHz and 1.3 Volts has a maximum temperature of low to mid 70's.
I am also pretty damn sure that KitGuru didn't manage to beat my CineBench R20 score of 10,170 with their overclock, because their CPU will have been throttling.
The link that you provided only emphasises my point that reviewers are either too lazy, or too stupid, to properly configure a third Gen Ryzen CPU.
In future it might help if you provide links to the original articles you are talking about.
Maybe you just won the silicon lottery with your particular 3950X.
How did you overclock it again?
I wrote down every step I made in configuring my 3950X here:
https://community.amd.com/thread/249957
My friend, whose 3950X I configured on a different motherboard with different RAM has very similar results to my own, so no, I didn't get a golden sample or win the silicon lottery.
I posted the relevant portion of the articles. Although it was my bad that I didn't link to the actual articles themselves.
I really get what jou are saying... nu won results often beat results from reviews.
In soms cases of might be the reviewer, most of the time of will be of het stuff.
For example reviewers often get rev1.0 of a motherboard or even a beta board. Also they get a BIOS which often focuses on stability. In turn memory support might not be great and sometimes even required running on JEDEC specs, instead of rated timmings.
Is your BIOS newer? Most likely! Is the agesa version more recent?
Is your Windows up to date? Reviewers mostly use an image as a standardized Windows install.
The videocard drivers can still influence the CPU score...
Most reviewers do not retest when a new piece of hardware gets tested. So see review numbers as a comparision between components at releasetime. They give an indication, but not realworld exerience. There has been a lot of ryzen optimisations to a lot of softwaretitles recently, which also adds up.
You always have to review the reviewer for yourself, but a good reviewer will not go full opitimal-setup. They should strive to use a setup which enables them to isolate a component and quantify their numbers. They require these tests to be the same (as much as possible) everytime, in order to compare it with previously tested hardware ( & software).
Some reviewers do revisits, or update their tests when it can't be avoided. They will mention it, as system and software specs.
Still, Nice numbers @Op
Tldr; read those whole article to get context on those numbers and validity for yourself, dont just look at the numbers.
I might concur with your argument if it were the 3600 or any of the originally launched CPUs.
I might even let you stretch that to the launch of the 3900X
I am NOT letting you get away with stretching it to the 3950X where AGESA 1.0.0.4 B was a pre-requisite.
Thus your entire argument, with regard to my original article is moot.
If I use a Swiss Army Knife to remove a couple of those RAM thingys and increase the 3200MHz RAM (which is only rated at 3200MHz) frequency to 3233MHz (which isn't overclocking) hit this:
Ryzen 2700X, 8 cores 16 threads.
55.6 FPS.
I guess I should stick a hoover in the AIO radiator and remove some of the fluff and see if that helps.
Well if you are running RAM specified for 3200 at 3233 then that would be overclocking.
If I run RAM specified for 3600 at 3600, then no, that's not overclocking.
You're not quite there yet, but if you keep banging the rocks together then I see the promise of fire in your future.
Where are the specifications for your PC in this post?
Motherboard, CPU, RAM, PSU etc.
Normally supposed to be given at the start of the post when you are asking for help.
I haven't needed to ask for help; and from reading what you have posted, I don't think you would be in a position to give it.
You need help alright.