AMD Ryzen 7 5800X

[efikkan]

The reason for the low RAM speed vs other reviewers is consistency so it can be compared to previous systems - wizz may need to go to some extra effort for high speed DDR4 testing on ryzen, as well as timing testing that he mentioned doing somewhere.

A reference comparison should put each contender in the best possible light, while remaining within specs.
That would mean stock memory speed (3200 MHz for Zen2/3, 2933 MHz for Comet Lake, preferably with JEDEC timings too), power limits enabled, etc.
Whether Ryzen supports 3200 MHz and others support something less is irrelevant. How would you otherwise compare between architectures supporting different types of memory, like e.g. Haswell(DDR3 1600 MHz) or upcoming DDR5 based CPUs?

Any faster memory than the official spec is overclocking, and it's not a problem to include both stock and overclocked results, as long as they are clearly marked as so.
But overclocking raises another concern, how far do you push it when OC is varying between each sample? Also there is also no telling how long you can sustain an OC over time, if you're buying based on that expectation.

 

[Mussels]

A reference comparison should put each contender in the best possible light, while remaining within specs.
That would mean stock memory speed (3200 MHz for Zen2/3, 2933 MHz for Comet Lake, preferably with JEDEC timings too), power limits enabled, etc.
Whether Ryzen supports 3200 MHz and others support something less is irrelevant. How would you otherwise compare between architectures supporting different types of memory, like e.g. Haswell(DDR3 1600 MHz) or upcoming DDR5 based CPUs?

Any faster memory than the official spec is overclocking, and it's not a problem to include both stock and overclocked results, as long as they are clearly marked as so.
But overclocking raises another concern, how far do you push it when OC is varying between each sample? Also there is also no telling how long you can sustain an OC over time, if you're buying based on that expectation.

That's just it, where do you stop? Does w1zz need overclocked CPU and RAM results for everything in the charts?

It gets too much, so it was totally reasonable to standardise things... but in these CPU reviews, it seems there may be a need for 4x high speed sticks of ram as i've seen claims of upto 10% boosts

 

[mainlate]

Maybe it´s OK for AMD that 5800x, 450-500€ processor is slower in memory write than Athlon 200GE, 50€ processor, but it´s not OK for me.

 

[efikkan]

That's just it, where do you stop? Does w1zz need overclocked CPU and RAM results for everything in the charts?

And due to sample variance he should probably test about 100 samples per CPU model too

It gets too much, so it was totally reasonable to standardise things... but in these CPU reviews, it seems there may be a need for 4x high speed sticks of ram as i've seen claims of upto 10% boosts

Yes, it does. The basis for most reviews should be stock configuration. And then perhaps for select relevant models there can be a separate overclocking focused test.

But since samples can vary so much in quality, overclocking results shouldn't be a basis for people's purchasing choices. And be aware that high memory speeds are not always achieveable. I've seen Kaby Lake CPUs struggle to achieve 3000-3200 MHz (with the "right" memory and motherboards), and unable to retain anything close over time, even though >3200 MHz were "expected" at the time.
You can also look across the forums and see the majority of issues with Zen(1)/Zen 2 builds are related to overclocked memory, and so many struggling with their first PC build due to XMP. If they had stuck to the fastest JEDEC speed of their CPU, it would have just worked.

 

[Octopuss]

Why are the temperatures so high?

 

[bug]

A reference comparison should put each contender in the best possible light, while remaining within specs.
That would mean stock memory speed (3200 MHz for Zen2/3, 2933 MHz for Comet Lake, preferably with JEDEC timings too), power limits enabled, etc.
Whether Ryzen supports 3200 MHz and others support something less is irrelevant. How would you otherwise compare between architectures supporting different types of memory, like e.g. Haswell(DDR3 1600 MHz) or upcoming DDR5 based CPUs?

Any faster memory than the official spec is overclocking, and it's not a problem to include both stock and overclocked results, as long as they are clearly marked as so.
But overclocking raises another concern, how far do you push it when OC is varying between each sample? Also there is also no telling how long you can sustain an OC over time, if you're buying based on that expectation.

That's just it, where do you stop? Does w1zz need overclocked CPU and RAM results for everything in the charts?

It gets too much, so it was totally reasonable to standardise things... but in these CPU reviews, it seems there may be a need for 4x high speed sticks of ram as i've seen claims of upto 10% boosts

Yes, the line should be drawn somewhere.
But instead of understanding where the line has been drawn and judge results starting from there, it is much easier for people to simply complain the line hasn't been drawn where they think it should have been drawn. Or even that it has been drawn at all.

 

[Xuper]

A reference comparison should put each contender in the best possible light, while remaining within specs.
That would mean stock memory speed (3200 MHz for Zen2/3, 2933 MHz for Comet Lake, preferably with JEDEC timings too), power limits enabled, etc.
Whether Ryzen supports 3200 MHz and others support something less is irrelevant. How would you otherwise compare between architectures supporting different types of memory, like e.g. Haswell(DDR3 1600 MHz) or upcoming DDR5 based CPUs?

Any faster memory than the official spec is overclocking, and it's not a problem to include both stock and overclocked results, as long as they are clearly marked as so.
But overclocking raises another concern, how far do you push it when OC is varying between each sample? Also there is also no telling how long you can sustain an OC over time, if you're buying based on that expectation.

and 3200Mhz can be either DK or SK.TPU tested with 2x8GB which is Single rank.while 2x16GB is dual rank ( according to buildzoid)

 

[xenocide]

Are we looking at the same chart?

Similar temps as last generation, much more perf

Sorry for the late response, yea I misread it. I was comparing it to the review of the R5 part which runs notably cooler, and the fact that the OC barely runs hotter, but given the way Boost works, that actually makes sense. You're right, entirely within expected limits.

Why are the temperatures so high?

As W1zz pointed out when I thought that, compare it to models of the past gen, not lower core count models within the same line. It's actually running about even (within 2c) of the 3800XT, meaning it's running barely hotter for substantially higher performance.

 

[Octopuss]

Why does it run 20°C hotter than 10900K, wich eats like 100W more though?

 

[Mussels]

Why does it run 20°C hotter than 10900K, wich eats like 100W more though?

1. heat density
2. different location of the chip

 

[W1zzard]

Why does it run 20°C hotter than 10900K, wich eats like 100W more though?

What @Mussels said, and 10900K runs into its power limit

 

[Emboldi]

This is test are wrong, techpowerup loses all credibility for me . !!!

I think they did adress this right. Also, W1zz is about the most credible source out there in my experience.

 

[Mussels]

just got mine today

no real benchies etc yet, got some air bubbles in the AIO that need to work themselves out before i get any real testing done

chips fast as hell with 64GB 3600 behind it

Wish i knew how w1zz got his temps down so low, mine pretty much boosts to max and sits at 90C, on a 240mm AIO :/

 

[xenocide]

just got mine today

no real benchies etc yet, got some air bubbles in the AIO that need to work themselves out before i get any real testing done

chips fast as hell with 64GB 3600 behind it

Wish i knew how w1zz got his temps down so low, mine pretty much boosts to max and sits at 90C, on a 240mm AIO :/

How is your AIO configured?

 

[Mussels]

How is your AIO configured?

pump at max speed, top mounted 240mm at 1200rpm - doesnt matter what i set, the chip just turbos higher and higher til it maxes at 90c

 

[xenocide]

pump at max speed, top mounted 240mm at 1200rpm - doesnt matter what i set, the chip just turbos higher and higher til it maxes at 90c

Hmm. As long as the hoses aren't twisted too much you should be getting optimal flow. I know a lot of people have bad results with AIO's when they have the radiators mounted in front or back with the hose outlets at the highest point due to air building up. Maybe run it for a bit and try new thermal paste down the line. Could just be unlucky.

 

[Mussels]

Hmm. As long as the hoses aren't twisted too much you should be getting optimal flow. I know a lot of people have bad results with AIO's when they have the radiators mounted in front or back with the hose outlets at the highest point due to air building up. Maybe run it for a bit and try new thermal paste down the line. Could just be unlucky.

reddit has dozens of people like me, regardless of cooler seeing flatline 90C in multi threaded load. there is talk about a BIOS issue being behind it, which is very possible.

Ahah! i had to rotate my EK AIO around so the hoses are on the same side as the RAM.

This should make no difference, but on the 5800x it does for some reason,now seeing 80-85C load instead of instant 90C

 

[harm9963]

Having fun with my new 5800x , just using PBO MAX 5050 ! and DOCP MAX 5050 !

 

[Mussels]

Happy times here too, she reads hot but damn... breaking 5GHz with this level of IPC is nuts

 

[magixx]

Is there a review covering memory timings somewhere? I searched this thread and @W1zzard mentioned there would be a post coming soon.
I'm curious between the difference of 3200 CL14, 3600 CL16, 4000 CL18.

 

[bug]

Is there a review covering memory timings somewhere? I searched this thread and @W1zzard mentioned there would be a post coming soon.
I'm curious between the difference of 3200 CL14, 3600 CL16, 4000 CL18.

AMD Zen 2 Memory Performance Scaling with Ryzen 9 3900X

We take a close look at memory scaling on AMD's new Zen 2 Ryzen 3900X, testing both application and gaming performance at seven different memory speed and timing combinations ranging from 2400 MHz all the way up to 4000 MHz.

www.techpowerup.com

Edit: oops, you were looking for Zen3 scaling, my bad.

 

[W1zzard]

Is there a review covering memory timings somewhere? I searched this thread and @W1zzard mentioned there would be a post coming soon.
I'm curious between the difference of 3200 CL14, 3600 CL16, 4000 CL18.

i got pwned by all the gpu reviews, hopefully i can do this testing around xmas time

 

[bug]

i got pwned by all the gpu reviews, hopefully i can do this testing around xmas time

If you were nice this year, you could have simply asked Santa to give you the benchmark results.

 

[W1zzard]

If you were nice this year, you could have simply asked Santa to give you the benchmark results.

I was really nice this year, so he gave me a ton of graphics cards

 

[Mussels]

I was really nice this year, so he gave me a ton of graphics cards

You shouldn't have wished for global warming to end, if you didnt want the entire planets stock of modern graphics cards