AMD Confirms Zen 3's Performance Is Monstrous and Speculation Thread

I bought one of the newer X570 boards, the MSI X570 Tomahawk which people are raving about but I'm going to see if the X670 rumors are true and new boards are released (minus the chipset fan) alongside Zen 3...the fan is not a big deal but all things being equal I'd prefer a board without it

I keep thinking about getting an x570 or b550 board right now but I have a feeling an x670 or something is coming with Zen 3 and I know I will want the latest and greatest with Zen 3.
 
I keep thinking about getting an x570 or b550 board right now but I have a feeling an x670 or something is coming with Zen 3 and I know I will want the latest and greatest with Zen 3.

usually motherboard manufacturers have leaks ahead of time mentioning new boards incoming but there's been complete silence...so either AMD has everything locked down tight or there won't be any new X670 boards...an X670 is not really needed as X570 has everything...it really only makes sense to release a new chipset with AM5 and DDR5
 
You will most likely need to update your BIOS before installing, but yes, all x570's and b550's are supporting it, and SOME x470's and b450's will be as well (This will be manufacturer and model specific it seems).

I'm hoping my X370 does, but I am doubtful. But ASRock released some BIOS recently for my motherboard so maybe there is hope.
 
I'm hoping my X370 does, but I am doubtful. But ASRock released some BIOS recently for my motherboard so maybe there is hope.
hopefully AMD will surprise us and let us put their new Zen 3 beasts on the 300 series chipsets
 
usually motherboard manufacturers have leaks ahead of time mentioning new boards incoming but there's been complete silence...so either AMD has everything locked down tight or there won't be any new X670 boards...an X670 is not really needed as X570 has everything...it really only makes sense to release a new chipset with AM5 and DDR5
Because of the lack of leaks I’m worried that a Threadripper refresh won’t be happening on this announcement. Or it won’t have the feature set of if the TR Pro’s.
 
usually motherboard manufacturers have leaks ahead of time mentioning new boards incoming but there's been complete silence...so either AMD has everything locked down tight or there won't be any new X670 boards...an X670 is not really needed as X570 has everything...it really only makes sense to release a new chipset with AM5 and DDR5

I haven't been following closely in a while. For me right now other than my itch I can wait until October when AMD announces. If theres nothing new I'll probably grab an X570 ITX board.
 
Hopefully these processors will not be as hard to get compared to RTX 3080. I would love to buy something around black friday, but they might not even be in stores by that time.
 
Waiting patiently for October 8 AMD press release. I'm thinking they will be in stores 2-3 weeks after that. Sold most of my parts in anticipation for these CPU's. The 10700K for $350 is hard to beat, especially with another $20 off the motherboard. Exciting times to build a new rig that's for sure.
 
Hopefully these processors will not be as hard to get compared to RTX 3080. I would love to buy something around black friday, but they might not even be in stores by that time.
My Rep at CDW has been told by his vendors to not expect any meaningful numbers until early 2021 for any of the AMD nor NVidia parts. Lenovo has also mentioned to them that they aren’t going to meet their preorder demand for the Threadripper Pro based systems either.
 
My Rep at CDW has been told by his vendors to not expect any meaningful numbers until early 2021 for any of the AMD nor NVidia parts. Lenovo has also mentioned to them that they aren’t going to meet their preorder demand for the Threadripper Pro based systems either.


The wait continues! I will most definitely wait because the gains from the 3000 series (AMD) looks to be substantial. Maybe an all AMD build this time around! TY for the info.
 
The wait continues! I will most definitely wait because the gains from the 3000 series (AMD) looks to be substantial. Maybe an all AMD build this time around! TY for the info.
Granted I’m in Canada so the US should get better supply allocations.
 
Because of the lack of leaks I’m worried that a Threadripper refresh won’t be happening on this announcement. Or it won’t have the feature set of if the TR Pro’s.

it won’t have the TR Pro feature set. They’re more Epyc-Lite than TR. it’s a valid need, in a work context higher bandwidth with a higher clock speed has been a massive context. Can’t see them putting in 8 channel memory for anything prosumer. It’s too much “money on the table” for professionals.

Im holding out a tiny hope that they’ll launch everything together but it’s nothing more than a dream for me, I’m being told Q1 for Zen 3 TR. annoying as I’ve been meaning to do my new build for 5 months and am only just starting. I’m waiting till 10.08 as I’m still waiting for deliveries anyway, but Im certain that I’ll be buying the 3960 on the 9th.

On the upside, prices have dropped :)
 
usually motherboard manufacturers have leaks ahead of time mentioning new boards incoming but there's been complete silence...so either AMD has everything locked down tight or there won't be any new X670 boards...an X670 is not really needed as X570 has everything...it really only makes sense to release a new chipset with AM5 and DDR5

there's 3 options, either asmedia can't implement pcie 4.0 into their chipset architecture which i'm leaning on being the primary reason given b550 was basically a z390 knock off with a few minor changes. option 2, the 600 series will be the x570 chipset from AMD but on the original 7nm process from TSMC. option 3, they nix 600 series and board partners release updated versions of x570 using the newer features that were implemented in b550.

those are my theories because i agree, by now something would of leaked from board partners or partner test forums.

Because of the lack of leaks I’m worried that a Threadripper refresh won’t be happening on this announcement. Or it won’t have the feature set of if the TR Pro’s.

highly doubt they'd not release zen 3 TR. but given that TR typically releases 4-5 months after the first set of processors there probably won't be much said about them until after Christmas.

as far as TR pro's there will always be a difference between them and standard TR. so no i don't expect standard TR to ever work on the 8 channel boards.
 
Last edited:
there's 3 options, either asmedia can't implement pcie 4.0 into their chipset architecture which i'm leaning on being the primary reason given b550 was basically a z390 knock off with a few minor changes. option 2, the 600 series will be the x570 chipset from AMD but on the original 7nm process from TSMC. option 3, they nix 600 series and board partners release updated versions of x570 using the newer features that were implemented in b550.

those are my theories because i agree, by now something would of leaked from board partners or partner test forums.



highly doubt they'd not release zen 3 TR. but given that TR typically releases 4-5 months after the first set of processors there probably won't be much said about them until after Christmas.

as far as TR pro's there will always be a difference between them and standard TR. so no i don't expect standard TR to ever work on the 8 channel boards.


https://wccftech.com/amd-ryzen-9-59...hread-cpu-allegedly-up-to-5-ghz-150w-tdp/amp/
 
Let's get real. Gaming is pretty much the only reason people care about high end PCs, and games are going to be using 8 cores or less for the next decade.

Umm..no. The people who buy stronger CPU's use them mostly for encoding/rendering/folding etc where Intel gets trampled on. That statement of yours is so ignorant. Higher end CPU's make very little difference to gaming, its essentially for the usage cases as mentioned before where CPU performance actually matters.

Also, gaming performance at 2560+ resolution is within 2% when comparing the 3900XT to 10900K. But the Intel system will be anywhere from 15-35% slower at most other things.
 
Last edited:
Umm..no. The people who buy stronger CPU's use them mostly for encoding/rendering/folding etc where Intel gets trampled on. That statement of yours is so ignorant. Higher end CPU's make very little difference to gaming, its essentially for the usage cases as mentioned before where CPU performance actually matters.

Also, gaming performance at 2560+ resolution is within 2% when comparing the 3900XT to 10900K. But the Intel system will be anywhere from 15-35% slower at most other things.

Well...not quite. At 4K yes, at 1440p and lower not so much.

https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-amd-3900-xt-vs-intel-10900k/27.html
 
Umm..no. The people who buy stronger CPU's use them mostly for encoding/rendering/folding etc where Intel gets trampled on. That statement of yours is so ignorant. Higher end CPU's make very little difference to gaming, its essentially for the usage cases as mentioned before where CPU performance actually matters.

Also, gaming performance at 2560+ resolution is within 2% when comparing the 3900XT to 10900K. But the Intel system will be anywhere from 15-35% slower at most other things.

Encoding/rendering is all done in hardware now. No one's doing it in software. Your statement is the ignorant one.

The gaming market is the large chunk of high end CPUs. Intel CPUs have had a 20-30% advantage in games over AMD, and you're saying it makes "very little difference."

Video game emulators are basically all CPU constraints. You have no idea what you're talking about.
 
Hexus.net posted the leaked benchmarks of the supposed 5900X, comparing it to the 3700X.

They also mentioned what the old i7-2600K results are, so I made a little comparison.

I am using purely speculative guesses for the 5700X, based on using the exact same single-thread performance as the 5900X (which won't be accurate because of likely clock speed differences out of the box) and applying the 27.75% gain to the 3700X multi-thread result to get the 5700X multi-thread result, since they both are/should be 8C/16T.

Screenshot_20201001-163530.png
 
Last edited:
Nexus.net posted the leaked benchmarks of the supposed 5900X, comparing it to the 3700X.

They also mentioned what the old i7-2600K results are, so I made a little comparison.

I am using purely speculative guesses for the 5700X, based on using the exact same single-thread performance as the 5900X (which won't be accurate because of likely clock speed differences out of the box) and applying the 27.75% gain to the 3700X multi-thread result to get the 5700X multi-thread result, since they both are/should be 8C/16T.

View attachment 284667


Yaasss! Time to upgrade my 2700k. Let's extrapolate at least a 50% gain in ST 🤘
 
Encoding/rendering is all done in hardware now. No one's doing it in software. Your statement is the ignorant one.

The gaming market is the large chunk of high end CPUs. Intel CPUs have had a 20-30% advantage in games over AMD, and you're saying it makes "very little difference."

Video game emulators are basically all CPU constraints. You have no idea what you're talking about.

Software encoding has a quality advantage over hardware, at the cost of additional time. So there absolutely a use for faster CPUs when doing media storage vs streaming.

In addition, you've repeatedly used that 20-30% point, but dont include sources. Posting them would help by giving everyone else a look at your data. I've included a couple to counter your claims.

https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-amd-3900-xt-vs-intel-10900k/27.html

https://www.techspot.com/amp/review/2084-amd-or-intel-for-gaming-benchmarking/
 
Software encoding has a quality advantage over hardware, at the cost of additional time. So there absolutely a use for faster CPUs when doing media storage vs streaming.

In addition, you've repeatedly used that 20-30% point, but dont include sources. Posting them would help by giving everyone else a look at your data. I've included a couple to counter your claims.

https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-amd-3900-xt-vs-intel-10900k/27.html

https://www.techspot.com/amp/review/2084-amd-or-intel-for-gaming-benchmarking/
I would like someone to do a real analysis of software vs hardware quality in today's era. I think the advantage of software encoding was more apparent at lower resolutions (1080p), but I think when going to high res above 4K, hardware might be just as good since it has more information to work with.
 
https://hothardware.com/reviews/amd-ryzen-9-3900x-vs-core-i9-9900k-ipc-shootout?page=3

In emulators the difference can be even bigger (and it probably matters more in them than native PC games, really).

Look at the Dolphin benchmarks.

https://www.extremetech.com/computi...-amd-solidifies-its-control-of-the-cpu-market

Ok, I see two games tested, one that's 10% slower and one where the 12 core is drastically slower, but when you drop it down to 8 cores it's 11% slower.

Not a single game tested in the second link is 20% slower, unless I'm misreading.


Yep, I reread and the only test that tells the story you're claiming is Dolphin. A single case does not a trend make. You can't extrapolate from that single data point a general statement, especially when the rest of the data points contradict.
 
Last edited:
I would like someone to do a real analysis of software vs hardware quality in today's era. I think the advantage of software encoding was more apparent at lower resolutions (1080p), but I think when going to high res above 4K, hardware might be just as good since it has more information to work with.

Would be good to see updated testing, so many variables and quality can be subjective as well. Definitely makes it a daunting task.
 
Would be good to see updated testing, so many variables and quality can be subjective as well. Definitely makes it a daunting task.
I'm not volunteering! I basically use hardware encoding for anything personal use, and software for any paid projects. I'm not trying to look for differences, and I'm OK with hardware encoding quality, but clients get highest quality output.
 
Ok, I see two games tested, one that's 10% slower and one where the 12 core is drastically slower, but when you drop it down to 8 cores it's 11% slower.

Not a single game tested in the second link is 20% slower, unless I'm misreading.


Yep, I reread and the only test that tells the story you're claiming is Dolphin. A single case does not a trend make. You can't extrapolate from that single data point a general statement, especially when the rest of the data points contradict.

It's not just Dolphin. There are other emulators and games where the difference is just as extreme.
 
It's not just Dolphin. There are other emulators and games where the difference is just as extreme.

It's well known that intel has a performance advantage in Dolphin, if that's your use case, then your current cpu choice should be clear.

It's your extrapolation of that dolphin data to make a blanket statement that I'm questioning. The data that I've supplied, and frankly that you've supplied, doesn't show a 20% performance delta except in that limited case.

Can you link to some data? If there are other games with a 20-30% disadvantage, I'd love to see the info.
 
It's not just Dolphin. There are other emulators and games where the difference is just as extreme.

So don’t get AMD for emulators. You seem to think your use case is the most important and AMD should be making the CPU just for you. Buy Intel if you only care about emulators but maybe stop making blanket statements such as “20% worse at gaming” considering most of us play “regular” games and the difference there is much smaller.
 
So don’t get AMD for emulators. You seem to think your use case is the most important and AMD should be making the CPU just for you. Buy Intel if you only care about emulators but maybe stop making blanket statements such as “20% worse at gaming” considering most of us play “regular” games and the difference there is much smaller.

Hmm i've only just started using dolphin on my 3900x and it seems to work fine emulating some games, same with my haswell quad in my laptop. I haven't compared performance though.
 
sad,
so you absolutely need that 10900K then? :(

Well, if the rumors ring true that Zen3 is approx 27% faster than Zen2, then that would mean a completion time of 4m17s for a Zen3 equivalent of the 3700X 8C/16T.

Of course, that may not scale linearly in Dolphin, so the 10900K may still take the lead.

In this usage case, the 10900K is the safe bet because one knows exactly what their money will be getting instead of an unknown by waiting for Zen3 and potentially being disappointed.
 
  • Like
Reactions: erek
like this
Well, if the rumors ring true that Zen3 is approx 27% faster than Zen2, then that would mean a completion time of 4m17s for a Zen3 equivalent of the 3700X 8C/16T.

Of course, that may not scale linearly in Dolphin, so the 10900K may still take the lead.

In this usage case, the 10900K is the safe bet because one knows exactly what their money will be getting instead of an unknown by waiting for Zen3 and potentially being disappointed.


 
Encoding/rendering is all done in hardware now. No one's doing it in software. Your statement is the ignorant one.

The gaming market is the large chunk of high end CPUs. Intel CPUs have had a 20-30% advantage in games over AMD, and you're saying it makes "very little difference."

Video game emulators are basically all CPU constraints. You have no idea what you're talking about.

Am I missing something ?
How do you use hardware without software ?
20 - 30 % which games are we talking about here ?
 
Back
Top