Discussion Ryzen 3000 series benchmark thread ** Open **

[Dave3000]

I noticed that this review is showing that the 3700x has much slower memory writes than the 3900x. https://hexus.net/tech/reviews/cpu/132374-amd-ryzen-9-3900x-ryzen-7-3700x/?page=4

Is this a BIOS issue or is it a hardware limitation of the 3700x due to having only one chiplet? Can someone here do the same memory write benchmark on their 3700x with the same memory speed that is using a different motherboard than the one in this review and tell me if they get around the same memory write score.

 

[Kenmitch]

I noticed that this review is showing that the 3700x has much slower memory writes than the 3900x. https://hexus.net/tech/reviews/cpu/132374-amd-ryzen-9-3900x-ryzen-7-3700x/?page=4

Is this a BIOS issue or is it a hardware limitation of the 3700x due to having only one chiplet? Can someone here do the same memory write benchmark on their 3700x with the same memory speed that is using a different motherboard than the one in this review and tell me if they get around the same memory write score.

Yes it's a 1 chiplet thing. Design choice made by AMD.

 

[IEC]

I noticed that this review is showing that the 3700x has much slower memory writes than the 3900x. https://hexus.net/tech/reviews/cpu/132374-amd-ryzen-9-3900x-ryzen-7-3700x/?page=4

Is this a BIOS issue or is it a hardware limitation of the 3700x due to having only one chiplet? Can someone here do the same memory write benchmark on their 3700x with the same memory speed that is using a different motherboard than the one in this review and tell me if they get around the same memory write score.

It's hardware. 2 chiplets (e.g. 3900X) gives double the write bandwidth versus 1:

 

[Dave3000]

Does X-Plane 11 make heavy use of memory writes, especially with HD Mesh installed? Also does memory write performance have an effect in load times in games?

 

[BTRY B 529th FA BN]

Anyone keeping track of the CPU stepping on these chips?

 

[Markfw]

Anyone keeping track of the CPU stepping on these chips?

Well I am on stepping zero

 

[BTRY B 529th FA BN]

Well I am on stepping zero

Is it indicated by B0?

 

[IEC]

Is it indicated by B0?

CPU-Z:


I haven't seen any other stepping in the retail channel (yet).

 

[BTRY B 529th FA BN]

What is the AMD recommended absolute limit to these chips? I know my i7 970 was 1.375v while the uncore was 1.35v

 

[Markfw]

What is the AMD recommended absolute limit to these chips? I know my i7 970 was 1.375v while the uncore was 1.35v

As we speak, with all 24 threads@100%, I am running 4 ghz@ 1.275 vcore (approx, keeps changing) 1.243 a lot.

 

[IEC]

What is the AMD recommended absolute limit to these chips? I know my i7 970 was 1.375v while the uncore was 1.35v

1.325V all-core. But you will be limited by cooling unless on custom water. So it's better to do the maximum you can achieve at 1.25V per CCX for comfortable temps and better than stock MT performance. If you rely on ST performance, stock boost works better unless you get a golden sample capable of 4.4GHz+ at <1.3V

 

[BTRY B 529th FA BN]

Sorry, yes, I meant AMD recommend safe voltage limit.

Thanks

 

[Kedas]

Looks like we are missing an important parameter here for choosing the better memory: (same clock and latency but diff performance)

CMK16GX4M2B3200C16

 

[Fanatical Meat]

Looks like we are missing an important parameter here for choosing the better memory: (same clock and latency but diff performance)

CMK16GX4M2B3200C16

Can you please give a summary

 

[Kedas]

Can you please give a summary

It's between the brackets there is one older kit that acted faster than expected no reason yet why, and obvious get higher than 3000Mhz for your Zen2 but that you already knew.

 

[beginner99]

Looks like we are missing an important parameter here for choosing the better memory: (same clock and latency but diff performance)

sub timings can matter a lot.

 

[Kedas]

Sure but do you really think he didn't check that?
or do you mean some subtimings that can't be set in the bios?

 

[IEC]

Sure but do you really think he didn't check that?
or do you mean some subtimings that can't be set in the bios?

He mentioned in the video that the memory kit with worse performance required 3 boot-retrain cycles before it booted up. That means each time it used successively looser subtimings before it succeeded. That's why it was slower versus the set that worked at the same timings in 1 boot. It had tighter default subtimings.

 

[Kedas]

Is the moral of the story if you have to choose between CL16 or CL14 choose the one that is approved for Ryzen? ....

if he would have changed subtimings in the bios the same as the other module you think it would not boot?

in some test the 3200CL16 did beat the 3600CL16...

 

[beginner99]

Is the moral of the story if you have to choose between CL16 or CL14 choose the one that is approved for Ryzen? ....

if he would have changed subtimings in the bios the same as the other module you think it would not boot?

in some test the 3200CL16 did beat the 3600CL16...

Oh i think if he had done that it might have booted and got the same performance. We don't know but the point Cl and mhz aren't the only 2 things that matter.

Also as has been shown higher RAM clocks can affect thermal headroom of the chip which actually reduces performance due to chip boosting less.

 

[coercitiv]

higher RAM clocks can affect thermal headroom of the chip which actually reduces performance due to chip boosting less.

Looks like AMD is on the verge of offering us the ultimate CPU RPG experience. People complained overclocking is dead so they gave us the ultimate tweaking machine. By the time we're done the next Zen expansion will be out.

 

[therealmongo]

Looks like AMD is on the verge of offering us the ultimate CPU RPG experience. People complained overclocking is dead so they gave us the ultimate tweaking machine. By the time we're done the next Zen expansion will be out.

I like this

 

[B-Riz]

Anyone keeping track of the CPU stepping on these chips?

There should be no more steppings, the next "stepping" will be with a new chip, Zen3. AMD does not work like Intel, final silicon should be B0.

I would link if I could remember where I read that, there was a comment from an ex-AMD employee that explained their silicon steppings / process and why.

Also, B0 was finalized late last year / early this year, so, they have been producing raw chips for half the year, abouts; demand has outstripped supply, there are no supply issues, FWIW.

 

[Asterox]

https://www.techpowerup.com/review/new-amd-chipset-driver-performance-test-ryzen-9-3900x/3.html

"The biggest benefactors of these drivers are tests that are less parallelized, such as web rendering, in which all three benchmarks post over 5% performance gains. Microsoft Office tests also post 3-4 percent performance gains, as does Adobe Photoshop

Overall, to gain performance improvements of any kind through a "chipset" driver update is a pleasant surprise, and it's good to see AMD delivering lifecycle-quality updates mere weeks into the product launch, especially when it's troubleshooting not one, but two distinct product lines—Ryzen and Radeon. If you have one of these processors, we highly recommend you update your drivers."

 

[BTRY B 529th FA BN]

So for the rest of us who don't care about the idle power savings and want a 1ms polling rate, we are now screwed cause people were complaining about a software layered observing app?