Support NeoGAF

[Dontero]

3.3 GHz base clock and 4.2 GHz boost clock is a far cry from 3.9/4.7, let alone 4.3/5.1, even for an engineering sample.

That is 500-600mhz difference. Not that much.
Secondly Adored source didn't really specify precise clocks they were extrapolated mostly from AMD cinebench show +some boost to clocks because that test was done on engineering sample.

Either way 16 cores at 4.2 for reasonable money is enough to completely lit the fucking fire under Intel, especially if AMD will reach more or less IPC parity.

 

[Leonidas]

Hoping they'll match Intel gaming performance this time, I'm personally holding off for the more efficient Zen3 2020

 

[wordslaughter]

Based off current Zen and Zen+, we can guesstimate the overclocking potential for Zen 2. Just add 200 - 350mhz on top of the boost clock.

 

[llien]

top

Truth be told, Intel gives process nodes more conservative ways. They are not wrong by comparing own 14nm to TSMC's 10nm etc.

 

[Dontero]

Based off current Zen and Zen+, we can guesstimate the overclocking potential for Zen 2. Just add 200 - 350mhz on top of the boost clock.

All previous Zens were made in GLoFlo.
this time they are using TSMCs 7nm.

Secondly there is a reason why it is called Zen 2 instead of Zen++

So we don't know how it will handle.

 

[llien]

Ok, I've re-checked the sources on that and the only thing I could find is VERY AdoredTV "leak" from 2018.
Meh.

This looks like "overhype hits AMD again" scenario all over.
16 core Ryzen is likely to stick with max 4.2Ghz (in line with what AMD staff openly talked about, mentioning that node bumps allow for more cores, but no tmuch clock improvements).

 

[Hellgardia]

Should be doable with good chiplet binning, tbh... Just need two 8 core chips that can hit those targets. I also know that I badly want this to be true.
I need a real upgrade but keeping a level head until I can actually buy one.

Something to note regarding the TUM_APISAK info: The 12nm TR 2950X has 16 cores and better specs than the unnamed engineering sample.
2950X has a base clock of 3.5Ghz and a max boost to 4.4Ghz.
The unnamed chip they mention is 3.3Ghz base, 4.2Ghz boost.
Cynic in me says problems with 7nm if they can't even match clocks. Optimist says it is just an engineering sample.

Don't forget the 2950x has a much larger package and a higher TDP (125w vs 180w if i'm not mistaken).
Even so, normally final products are a bit higher clocked than ES.

 

[shark sandwich]

Seems almost too good to be true. $99 for a 6-core CPU??

 

[CJY]

Almost too good to be true.

Seems almost too good to be true. $99 for a 6-core CPU??

AMD are targeting gross margins in the 40-45% region as far as I can recall when others in the industry have 60%+. This is to accelerate their gain of a critical customer mass base. They will increase their profits later. I believe they’re gonna pull through with insane value, setting AMD up for dramatic growth in the following decade.

I’d love to see Apple start using their APUs in their portable lineup. I’d buy a new MacBook Pro in a heartbeat AMD have a really incredible roadmap ahead.

 

[llien]

Uh, so, after re-reading toms post:

The actual (new) leak:


16 core, 3.3Ghz base, 4.2Ghz boost. (nowhere the 5Ghz from wherever that came)

Actual stuff shown by Lisa Su early this year was running at around 4Ghz, and yes, it was very cool on power consumption and yes, an engineering sample, but again nowhere the 5Ghz.

The rest is re-hash of 5 month old (!!!) AdoredTV leak that had already been wrong at least on that it would be announced at CES.

TLDR: tone down the expectations, you will likely get more cores and for cheap, boosting at around 4Ghz.

 

[Ascend]

That is 500-600mhz difference. Not that much.

It's close to 20% higher clocks... Not saying it's impossible. But the correlation between the engineering sample and the final specs, particularly the 4.3 base and 5.1 boost, seem too high.

16 core Ryzen is likely to stick with max 4.2Ghz (in line with what AMD staff openly talked about, mentioning that node bumps allow for more cores, but no tmuch clock improvements)

I don't think we can assume that. The 16 core will be made of two chiplets of 8 cores each. So the traditional idea where larger dies can't reach higher clock speeds is no longer true. It's simply about binning the best chiplets and putting them together. There is no reason to assume that the 8C/16T must have higher clocks than the 16C/32T in this case.

16 core, 3.3Ghz base, 4.2Ghz boost. (nowhere the 5Ghz from wherever that came)

But that's not any better than the 2700X, which is 3.7 GHz base and 4.3 GHz boost. That is why it's safe to assume that this is either not the best chip(let), or it's an engineering sample with lower clocks than the final product. Remember that even for 1st gen Ryzen, the slowest clocks are 3.1 GHz and 3.45 GHz, and the highest are 3.8 GHz and 4.0 GHz. And right now they're using 8 core chiplets, not a full single 16 core die.

Seems almost too good to be true. $99 for a 6-core CPU??

The chiplets will allow for great yields. The chances of them getting chiplets from the wafer with more than 2 faulty cores is extremely small. And since the i/o is on a separate chip, that doesn't reduce their yields either. I think that's the reason the bottom is 6 cores rather than 4 cores; Their produced chiplets don't really have less than 6 functional cores, so they release that as their lowest end. Maybe if they do have a few, they'll turn them into Athlons.

It will be interesting to see how the line-up will be. Since AMD is using 8 core chiplets, all of the line-up will consist of CPUs that contain either 6 activated cores, or 8 activated cores, provided what I said above is true. I doubt the 6C CPUs are two chiplets with 3 cores each. Yields would have to be atrocious for that to be the case. That would also mean that every 8 core is a full chiplet, rather than two chiplets of 4 cores. What will determine where the chiplet lands is its achievable clock speed.

What's also interesting is the APUs. The APUs will be getting the worst chiplets in terms of clock speed. The 3300G APU is slower than the slowest CPU, and the 3600G is equally as slow as the slowest of the CPU line-up, the 3300 (not APU). That's most likely to keep the power in check. But Navi must be quite efficient...

In any case...
This is the ranking in terms of chiplet quality, from best to worst, with why I think they were put there;

Ryzen 9 3850X 16 / 32 - 4.3 / 5.1 GHz (best 8 core chiplets)

Ryzen 7 3700X 12 / 24 - 4.2 / 5.0 GHz (best 6 core chiplets, can probably reach the speeds of the R9 3850X quite easily. They are basically clocked slightly lower to not incentivize buying the 3700X over the 3850X)

Ryzen 9 3800X 16 / 32 - 3.9 / 4.7 GHz (8 core chiplets that cannot reach 3850X speeds)

Ryzen 5 3600X 8 / 16 - 4.0 / 4.8 GHz (8 core chiplets that weren't good enough for being 3800X in terms of power consumption. They should be able to get close to 3850X speeds at a higher power draw.)

Ryzen 7 3700 12 / 24 - 3.8 / 4.6 GHz (6 core chiplets that can't reach 3700X clocks reliably.)

Ryzen 5 3600 8 / 16 - 3.6 / 4.4 GHz (8 core chiplet incapable of reaching 4.7 GHz reliably)

Ryzen 3 3300X 6 / 12 - 3.5 / 4.3 GHz (6 core chiplet incapable of reaching over 4.5 GHz reliably, or too power hungry to be an R3 3300)

Ryzen 3 3300 6 / 12 - 3.2 / 4.0 GHz (The most power efficient 6 core chiplets, but incapable of reaching R7 3700 speeds)

Ryzen 5 3600G 8 / 16 Navi (20 CU) 3.2 / 4.0 GHz (The slowest 8 core chiplets)

Ryzen 3 3300G 6 / 12 Navi (15 CU) 3.0 / 3.8 GHz (The slowest 6 core chiplets)

Note that there can be more power efficient and/or faster chiplets that will be used for Threadripper or EPYC.

 

[llien]

But that's not any better than the 2700X, which is 3.7 GHz base and 4.3 GHz boost.

Well, yes , but that is the actual recent leak.

 

[PhoenixTank]

Don't forget the 2950x has a much larger package and a higher TDP (125w vs 180w if i'm not mistaken).
Even so, normally final products are a bit higher clocked than ES.

True enough, not like for like in every respect but still an interesting data point, I feel. The step down to the 12 core TR still has that same 180W TDP, so not a firm measurement.

 

[metareferential]

Ryzen 7 3700k seems a great value if you're itching or needing to upgrade.

I could jump in with a newer nvidia gpu as well.

Even though my 7700k+titan xp combo could last me for a loooong time.

 

[Hellgardia]

True enough, not like for like in every respect but still an interesting data point, I feel. The step down to the 12 core TR still has that same 180W TDP, so not a firm measurement.

It's not a firm measurement but was using it more like a guideline

The 12-core AM4 CPU will probably maintain the 125W tdp that the 16-core has (the 12-core will be higher clocked ofc).

I will probably get a 16-core Ryzen 9 since it would greatly accelerate my workloads. I wonder if my current X370 Taichi will handle it though.

 

[ResilientBanana]

Considering that the 9900K doesn’t perform drastically better (ok drastically better is a matter of opinion) than the 2700X, and considering that the 3000 demo at CES was running neck and neck with the 9900K, there is pretty much no way that the 9900K will perform significantly better.

Then throw in factors like cost then no way the 9900K will be the CPU to get.

But would I want to settle for almost better?

 

[V1LÆM]

nice gimme that 3850X pls

 

[Tygeezy]

That is 500-600mhz difference. Not that much.
Secondly Adored source didn't really specify precise clocks they were extrapolated mostly from AMD cinebench show +some boost to clocks because that test was done on engineering sample.

Either way 16 cores at 4.2 for reasonable money is enough to completely lit the fucking fire under Intel, especially if AMD will reach more or less IPC parity.

It's always best to look at percentage change rather than raw increase or decrease. Could you image if that statement was made during the 1 ghz race? No big deal, just a 120 % increase bro! Anyway, the base clock change is 18.18 % and the boost clock change is 12 % in the wrong direction of what we would like.

 

[JohnnyFootball]

But would I want to settle for almost better?

That's what I mean. The 9900K is likely to be a worse CPU in value and performance.

No reason to get unless Intel shaves $300 off the price.

 

[SonGoku]

Truth be told, Intel gives process nodes more conservative ways. They are not wrong by comparing own 14nm to TSMC's 10nm etc.

But that was because TSMC 10 nm had 14nm parts and their 14nm had some 28nm parts right?
I assumed the 7nm node is "pure" for both parties, tbf Intel is jumping straight to 7nm EUV

 

[ResilientBanana]

That's what I mean. The 9900K is likely to be a worse CPU in value and performance.

No reason to get unless Intel shaves $300 off the price.

So you're saying the 9900k should be ~$200? That makes no sense, it's better. I could see a few dollars being shaved off, but that's like saying a Ford Fusion and a Ferrari should be the same price because they both drive.

Show me the real world benchmarks and then maybe you will have a good argument. But my i7-8700k is most likely on par with the Ryzen 3600G, maybe even the Ryzen 3700. AMD cores aren't really apples to apples with Intel's cores.

EDIT: What's funny is that people are saying the 3700X (12C/24T - 4.2Ghz / 5.0Ghz - $329) is the 9900k (8C/16T -3.6Ghz / 5.0Ghz - $494 ) killer. The specs of the 3700X should litterally shit all over the 9900k but I have a feeling they could be rather evenly matched.

As it stands the 2970WX is pretty evenly matched to the 9900k already.

UserBenchmark: AMD Ryzen TR 2970WX vs Intel Core i9-9900K

 

[llien]

But that was because TSMC 10 nm had 14nm parts and their 14nm had some 28nm parts right?

Yes on first part, no on second, basically, when Intel says "xnm" it's the size of the BIGGEST element.

I assumed the 7nm node is "pure" for both parties

No, it isn't, to my knowledge, silicon (smallest possible piece) will be 7nm, but metal part 10nm or even bigger.

 

[SonGoku]

Yes on first part, no on second, basically, when Intel says "xnm" it's the size of the BIGGEST element.


No, it isn't, to my knowledge, silicon (smallest possible piece) will be 7nm, but metal part 10nm or even bigger.

AdoredTV said it was going to be 7nm all

 

[PhoenixTank]

It's not a firm measurement but was using it more like a guideline

The 12-core AM4 CPU will probably maintain the 125W tdp that the 16-core has (the 12-core will be higher clocked ofc).

I will probably get a 16-core Ryzen 9 since it would greatly accelerate my workloads. I wonder if my current X370 Taichi will handle it though.

Aye, just can be an unreliable guideline unfortunately
Leak pegs the 12 core at 95/105W depending on clocks. We just don't know yet.

I think the current rumour is that the new CPUs won't have the headroom to hit full performance on the older boards. Plenty of trash rumours out there, though.

Show me the real world benchmarks and then maybe you will have a good argument. But my i7-8700k is most likely on par with the Ryzen 3600G, maybe even the Ryzen 3700. AMD cores aren't really apples to apples with Intel's cores.

EDIT: What's funny is that people are saying the 3700X (12C/24T - 4.2Ghz / 5.0Ghz - $329) is the 9900k (8C/16T -3.6Ghz / 5.0Ghz - $494 ) killer. The specs of the 3700X should litterally shit all over the 9900k but I have a feeling they could be rather evenly matched.

Cinebench is real world, no? AMD's 8 core chip matching Intel's 8 core 9900k at lower power draw is still pretty damn impressive. The rumours are that IPC and clocks are getting healthy bumps, so the argument of AMD cores being weaker may not be the case anymore. If you are looking for gaming benches... there is little we can offer yet - the products aren't released!

Evenly matched how? Gaming? Possibly. I doubt 12 cores will be needed in games for quite some time, but we've just told you that the lower specced 8 core 3600/3600X matches the 9900k in Cinebench.

 

[ResilientBanana]

Aye, just can be an unreliable guideline unfortunately
Leak pegs the 12 core at 95/105W depending on clocks. We just don't know yet.

I think the current rumour is that the new CPUs won't have the headroom to hit full performance on the older boards. Plenty of trash rumours out there, though.

Cinebench is real world, no? AMD's 8 core chip matching Intel's 8 core 9900k at lower power draw is still pretty damn impressive. The rumours are that IPC and clocks are getting healthy bumps, so the argument of AMD cores being weaker may not be the case anymore. If you are looking for gaming benches... there is little we can offer yet - the products aren't released!

Evenly matched how? Gaming? Possibly. I doubt 12 cores will be needed in games for quite some time, but we've just told you that the lower specced 8 core 3600/3600X matches the 9900k in Cinebench.

Interseting, I'll be keeping an eye out then. I used to run AMD during the Phenom days, I'm not opposed to them, but every PC I've built using AMD, including a few Ryzen builds, have been lack luster in real world applications. So I'd be pleasently surprised so see how they fair against the 9900k. However, the 10nm 10th gen Intel processors are coming. Seeing as how AMD's Zen 2 are a 7nm chip I'd like to see how they fair against the 10nm intel as that would be closer to an apples to apples comparison.

 

[PhoenixTank]

Interseting, I'll be keeping an eye out then. I used to run AMD during the Phenom days, I'm not opposed to them, but every PC I've built using AMD, including a few Ryzen builds, have been lack luster in real world applications. So I'd be pleasently surprised so see how they fair against the 9900k. However, the 10nm 10th gen Intel processors are coming. Seeing as how AMD's Zen 2 are a 7nm chip I'd like to see how they fair against the 10nm intel as that would be closer to an apples to apples comparison.

Open mind is important - time will tell whether this is overhyped or not, but I do feel like Intel has been sandbagging since Sandy Bridge.
Don't hold your breath on 10nm Intel for the comparison. By the time they have good yields for reasonably sized (6+ core) 10nm chips AMD/TSMC will be on 7nm+.

 

[Ascend]

very PC I've built using AMD, including a few Ryzen builds, have been lack luster in real world applications.

Which applications by which metric? If you use software that has been optimized for Intel for ages, obviously Ryzen is not going to be great at those. Every CPU has its strengths and weaknesses too, obviously.

 

[llien]

(adoredTV again, confirming higher clocks on 12C)

 

[Dontero]

AdoredTV said it was going to be 7nm all

No adored was one of the few that said his source said this will be chiplet design with both 7nm and 10/14nm dies. Every single other professional hardware site said it would be one die on 7nm aka like previous ryzen cpus.

 

[wordslaughter]

We will see come Computex but I've always believed Adored TV. It seems like much of what he has predicted is coming true.

Sure not EVERY price point and EVERY clock speed will be bang on, but I bet most of it will be true. True enough to demonstrate he wasn't just making things up.

 

[SonGoku]

No adored was one of the few that said his source said this will be chiplet design with both 7nm and 10/14nm dies. Every single other professional hardware site said it would be one die on 7nm aka like previous ryzen cpus.

Watch from 6:00 to 6:47

(start at 5:00 for full context)

I think it heavily implies pure 7nm, or am i mixing things?

 

[Dontero]

I think it heavily implies pure 7nm, or am i mixing things?

That is only talk about 7nm node itself.
New Ryzens are going with 7nm cores chips and 14nm I/O chip

And it wasn't rocket science because AMD already have shown before Rome CPU for servers with 1 i/o die and 4 chips with cores before showing R3000.

Press said it will be monolithic while he and very few other said they have sources telling him that it will be chiplet. Once you understand chiplet on Rome you know that it would be mixed die node with 14nm i/o die and 7nm cores chip/s.

 

[Ar¢tos]

Why is the I/O chip 14nm and not also 7nm? Is it a case of not being worth designing the I/O at 7nm?

 

[llien]

Why is the I/O chip 14nm and not also 7nm? Is it a case of not being worth designing the I/O at 7nm?

Because GloFo sucks, yet AMD has obligations to buy from them.

 

[Ar¢tos]

Because GloFo sucks, yet AMD has obligations to buy from them.

That sucks. Looking at the EPYC chip, they could easily get 80 cores there instead of 64 if the I/O was made using 7nm. That would put AMD in an even better position for supplying chips for servers.

 

[shark sandwich]

Because GloFo sucks, yet AMD has obligations to buy from them.

Do they still? I thought that was over with. Everything I’ve heard is that it’s a cost saving measure.

 

[McHuj]

Because GloFo sucks, yet AMD has obligations to buy from them.

Has absolutely nothing to do with that.

There is no need to make the IO die 7nm as the IO dies is just a bridge between the system and the cpu cores. It’s limited by the physical pins so it can’t really be shrunk. There’s no benefit to making it 7nm.

 

[Pagusas]

So much crazy expectations and AMD fanboys coming out of the wood work. Calm it down guys, AMD has been notorious on hyping shit up and then failing to deliver. Wait for real world performance numbers and cost, then if it proves to be true get excited and buy the hell out of it.

We all want a return to the AMD64/X2 days.

 

[PhoenixTank]

Do they still? I thought that was over with. Everything I’ve heard is that it’s a cost saving measure.

Yes, but also no. A bit of clarification needed on this topic.
The wafer supply agreement between AMD & GloFlo keeps changing. As it stands as of Jan 2019 AMD no longer need to pay GloFlo anything when making chips on 7nm node or better from others i.e. TSMC/Samsung.

IO silicon apparently doesn't scale as well with node changes, so less of a need for 7nm (as mentioned above). Without the IO on the same chip it means that the 7nm processor chiplets are smaller, in turn meaning better yields. AMD, in return for minimum purchase commitments to GloFlo, get a (presumably) very nice rate on 12nm silicon which helps to make this IO die & chiplets approach more economical.

They're leveraging all the resources they have to innovate. If they pull this off technically it'll make the hardware geek in me very happy.

 

[Ascend]

So much crazy expectations and AMD fanboys coming out of the wood work. Calm it down guys, AMD has been notorious on hyping shit up and then failing to deliver. Wait for real world performance numbers and cost, then if it proves to be true get excited and buy the hell out of it.

We all want a return to the AMD64/X2 days.

Remember before Ryzen came out where everyone was saying that there's no possible way AMD can catch up to Intel?

 

[SonGoku]

it would be mixed die node with 14nm i/o die and 7nm cores chip/s.

Whats the benefit of this hybrid approach vs pure 7nm

Has absolutely nothing to do with that.

There is no need to make the IO die 7nm as the IO dies is just a bridge between the system and the cpu cores. It’s limited by the physical pins so it can’t really be shrunk. There’s no benefit to making it 7nm.

what do you mean it cant be shrunk? It will remain 14nm forever?

 

[wordslaughter]

The reason for the IO die being 14nm is it's part of leveraging Ryzen's chiplet design advantages.

By separating the IO die and then by breaking down cores into 8 core chiplets ( all connected by infinity fabric ) you no longer have to make a monolithic die.

This increases yields substantially ( which reduces costs ). You also now have the freedom to actually DO something like have different parts at different processes, one part at 14nm while others are 7nm ( again to reduce costs ). Intel can't do this with their current designs.

But the chiplet design is the future for both companies. Look at Intel's Foveros, clearly chiplet based where they are stacked ontop of eachother rather than side by like what AMD is doing now. But AMD will be stacking their chiplets in the coming years too. AMD may even have a stacked chiplet product out before Intel.

 

[thelastword]

Should be doable with good chiplet binning, tbh... Just need two 8 core chips that can hit those targets. I also know that I badly want this to be true.
I need a real upgrade but keeping a level head until I can actually buy one.

Something to note regarding the TUM_APISAK info: The 12nm TR 2950X has 16 cores and better specs than the unnamed engineering sample.
2950X has a base clock of 3.5Ghz and a max boost to 4.4Ghz.
The unnamed chip they mention is 3.3Ghz base, 4.2Ghz boost.
Cynic in me says problems with 7nm if they can't even match clocks. Optimist says it is just an engineering sample.

Good post, I was going to say just that........ High core AMD chips already have higher base and boost speeds at 12nm, there's no way AMD won't reach higher speeds on 7nm, especially on Ryzen desktop processors......

If 12 core Ryzen is doing 4.3-5.1 Ghz, 16 core Ryzen could do 3.9-4.7 Ghz on final retail hardware at the very least......

 

[Dontero]

Another problem with 3D stacking is heat dissipation.
2D design is really good because you have more surface area to cool.

3D stacking though might have its own advantages. Even lower latency than normal monolithic die.

 

[Coulomb_Barrier]

Ryzen 9 3850X

16 / 32

-

4.3 / 5.1 GHz

AdoredTV redeemed Coulomb_Barrier ?

Lol no no no. These are just a copy paste of his fake chart.

Actual numbers should be revealed in 2 weeks. I'm guessing 85% of this will be wrong (considering names, clocks, prices)at least, therefore it was all a guess. Let's see.

 

[Coulomb_Barrier]

JohnnyFootball Can we edit the OP? That table is literally the AdoredTV leak and the article only includes it for reference.
Unfortunately TUM_APISAK is saying that the 16 core chip he has info on is slower than AdoredTV's leak. Engineering samples can change, but still, extra hype not needed.

There's slower and then there's a whopping 1ghz lower base clock than Adored's fake numbers for the 16-core, which he had at 4.3. So from engineering sample to retail we can at best expect +400Mhz on eng. sample numbers, but that's only if this 16-core is releasing within 5 months.

 

[PhoenixTank]

There's slower and then there's a whopping 1ghz lower base clock than Adored's fake numbers for the 16-core, which he had at 4.3. So from engineering sample to retail we can at best expect +400Mhz on eng. sample numbers, but that's only if this 16-core is releasing within 5 months.

Steady. Fake is a bit much given the things that have been right but the leak as a whole is otherwise still unconfirmed. The OP made it seem like this was a separate confirmation which is too far in the other direction, hence my post to try to keep things real.

Alongside the other info I've posted, worth considering the Zen 1 Engineering Samples.
The 8 core 95W TDP chip (among others) surfaced July 2016. Base clock 2.8GHz, boost to 3.2GHz.
Final 1800X chips (8 cores 95W) had 3.6GHz base, 4GHz boost. Flat 800MHz improvement there.
Works out as 28% uplift on base, 25% on boost from ES to retail.
Applied to that tasty 16 core would result in 4.2GHz base, 5.2 GHz boost which is amusingly close to a purported 3850X

Hold your objections - I know this isn't the same exact situation. To my knowledge, frequencies and the like don't scale linearly. There is also a longer period between that Zen 1 ES leak and retail release (July -> February) compared to TUM_APISAK's leak and the expected release. Different process. Different engineering challenges: brand new arch vs evolution of existing arch with new packaging strategy.
Entirely possible that 16 cores may be held back from initial release in a couple of months - there have been rumblings to that effect thus far.
Not PROOF™ but I personally still find it notable and interesting, even with the caveats.
A mere 400MHz improvement upon that sample just seems way too much of a low ball.

 

[JohnnyFootball]

Oh boy oh boy oh boy. X570 specs have leaked courtesy of Biostar

X570 Chipset Specs for AMD Ryzen 3000 Series Revealed by Biostar Racing GT8 Motherboard Listing

Biostar accidentally lists the X570 Racing GT8 motherboard, whipping the covers off the X570 chipset.

www.tomshardware.com

 

[Xyphie]

24 PCIe 4.0? Would match Ryzen 1000/2000's 24x PCIe 3.0. Hard to guess how many PCIe lanes they have since there's no mention of PCI generation on the 16x slots.

PCI-e:
First: 16x 4.0 (8x with second slot in use?)
Second: 8x 4.0 Shared with first slot?
3rd slot: 4x 4.0 Shared via PCH
1x slots: Shared

M.2:
First: 4.0 4x: from CPU
Second: Shared via PCH
Third: Shared via PCH

PCH Shared:
4.0 4x?

16x+4x+4x?

 

[JohnnyFootball]

24 PCIe 4.0? Would match Ryzen 1000/2000's 24x PCIe 3.0. Hard to guess how many PCIe lanes they have since there's no mention of PCI generation on the 16x slots.

PCI-e:
First: 16x 4.0 (8x with second slot in use?)
Second: 8x 4.0 Shared with first slot?
3rd slot: 4x 4.0 Shared via PCH
1x slots: Shared

M.2:
First: 4.0 4x: from CPU
Second: Shared via PCH
Third: Shared via PCH

PCH Shared:
4.0 4x?

16x+4x+4x?

All I want is for my GPU to run at 16X (although my 2080 Ti wont take full advantage of 4.0 bandwidth) and my two M2 drives to run at 4X