• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

Associate
Joined
19 Jun 2019
Posts
86
Location
Leicester, UK
I havent heard of intel wanting cinebench gone, note that geekbench still shows AMD making huge gains, so what I posted is nothing to do with AMD vs intel.

I know that, but I'm less inclined to believe those numbers as they represent far too many tasks into one overall number. It would be better to go through the individual numbers for the tasks you might want to compare. I don't pay attention to the overall figures. If one sub-score is way out, it throws everything off.
 
Last edited:
Soldato
Joined
27 Feb 2015
Posts
12,596
Some more notes regarding the testing I did on my 2 rigs.

On lightning returns the code is not just bad due to the threading but also how assets are loaded.

The game is ported from the ps3 and xbox360 which both have a tiny amount of memory to work with. So the game code basically unloads assets as soon as they out of view and has to keep reloading them again as you move around, it cannot even keep the map in memory.

When all this garbage collection happens and assets get reloaded you get stuttering and framerate drops, and the yasnaan area is really bad for this. What I discovered is memory timings and bandwidth massively reduce the effect.

Its worth pointing out at the time I did my 2600X testing I didnt optimise the ram timings. I didnt either for 8600k but the 8600k is higher spec'd its 3200CL14 vs 3000CL16 ram in the AMD system. So that was very likely a factor in my results as well. I might retest this, as now the ram in the AMD system is running at 3000CL14, and more optimised ram refresh timings.
 
Soldato
Joined
11 Jun 2003
Posts
5,056
Location
Sheffield, UK
Didn't think this would happen, was so hot for the 3700x but now with the supposed rubbish UK/EU pricing I'm tempted to go for 9700k.

9700K - £354
Motherboards.
£45 for a H310M
£100 for a cheap Z390
£250ish for something (pretty good) flagship

3700K - £320
Motherboards.
£52.49 for a cheap B450
£136 for a cheap X570
£250ish for something flagship (there's WAY more extreme options ofc. £350/450/700 you CAN keep going)

So...
9700K 8c/8t: £400-605
3700X 8c/16t: £373-570 (with options to go MUCH more mental).

If you think SMT will be quite so useless for your use case, sure. Otherwise... it's... more or less the same money for SMT.
Back when it was i5 vs i7 we all know which was preferred.
 
Permabanned
Joined
15 Oct 2011
Posts
6,311
Location
Nottingham Carlton
https://www.pcgamesn.com/amd/ryzen-3700x-overclocking-performance

“You know, with our boost algorithms, we eke out just about everything you can get. So maybe a couple hundred megahertz. With the 65W parts you’ll get a lot more because their specs are run with a lower power. So you can overclock the thing, get all the power of it, and, obviously, you get more headroom out of it.”


As expected 3800/3900 will be 2700x situatuion aka "overclockers dream" ;). I think 12x4.5 is max i will be able to get on my custom water with 230wTDP
 
Soldato
Joined
16 Sep 2018
Posts
12,638
They use the same dies, however Pinnacle Ridge is the codename for the AM4 Zen+ processors. Threadripper is known as Colfax.
Sorry, my mistake, i don't often use the code names so couldn't remember if TR used Colfax or PR, apologies.
That article you linked was written before it was 'unlocked' for AM4 processors via an AGESA update. Whether or not it has been removed again, I don't currently know the answer to that, but it looks that way.
I guess it comes down to what we consider officially supported to be, personally I'd say not launching with it, adding it and then removing it, isn't what would be considered official support.
That is actually the article I wanted to grab for my response, so thank you for that. XFR is the technology that analyzes the system, and Precision Boost reacts to those findings. They work together.

PBO is one step further than that by identifying the limitations of the VRM configuration for the board in use. It raises those limits imposed by standard Precision Boost, but it still has its own limits.
Yes and no, PBO doesn't identify the limitations of the VRM configuration for the board in use, it simply increase the limits imposed on PB, it's similar to increasing the power and thermal targets on their GPUs.
Also, I would like to go back and clarify something;

- Precision Boost 1.0 is akin to Turbo Boost 2.0.
- XFR1 is akin to Turbo Boost Max 3.0.
- There is no equivalent from Intel for Precision Boost 2.0 and XFR2.
Honestly I've not paid much attention to Intel in recent years as for me personally they got a bit boring, if i had to guess though i think you maybe drawing some equivalences in the wrong places, PB1 and PB2 are basically the same thing just with algorithm changes (PB2 is finer grained) and the changes from XFR1 to 2 basically extended how many of the cores to include in the data set (XFR1 used 1-2 cores, XFR2 uses all of them).
 
Associate
Joined
6 Nov 2013
Posts
41
https://www.pcgamesn.com/amd/ryzen-3700x-overclocking-performance

“You know, with our boost algorithms, we eke out just about everything you can get. So maybe a couple hundred megahertz. With the 65W parts you’ll get a lot more because their specs are run with a lower power. So you can overclock the thing, get all the power of it, and, obviously, you get more headroom out of it.”


As expected 3800/3900 will be 2700x situatuion aka "overclockers dream" ;). I think 12x4.5 is max i will be able to get on my custom water with 230wTDP

But that means you'd have to be able to set ~+300mhz to ~+500mhz before matching the higher tier chips, then more to actually make it worthwhile. And you haven't got the extra physical resources like extra L3 if you do get there?

I would have assumed the extra TDP would afford you a better starting point, or at least scales with the expected base clock. Assuming the base clocks are indicative of an abusive full thread load, it seems to make sense that the TDP of the 3700x @ 3.6 is 65W and 3800x @ 3.9 is 105W?

attachment.php


So I'd assume, by the time you can add enough of an offset to match the 3800x all core, your TDP is likely to be identical?

and likewise, for a 16 thread load, the 3900x and 3950x are likely to be making the same TDP, though obviously for higher thread loads, they'll begin dropping to lower clocks to stay within envelope?

Have no experience with previous Ryzen - when you use PE/PBO does it truly unlock the algorithm? i.e. leaving you to balance thermal/voltage management only and no longer having to concern yourself with TDP/current limits?

that actually makes me think the 3900x looks quite good - a 100mhz trade off for an additional 8 threads loaded...
 
Last edited:
Back
Top Bottom