[updated] Ryzen 3000 CPU benchmark spotted on Sandra

[pas008]

9 minutes ago, GDRRiley said:

Why are most of you assuming it’s a ryzen 3. It could be an Athlon. 

Thats why to me it’s got lower clocks. 

If I can get 8 or 12 core at 4.2 or 4.5 OCed I’ll be supper happy. 

Goodbye my poor 6600k at 4.1

its es sample they always are clocked low from any company

also could have cores locked on purpose for many reasons

[GDRRiley]

6 minutes ago, pas008 said:

its es sample they always are clocked low from any company

also could have cores locked on purpose for many reasons

Some of the very late samples are only a few hundred under. 

[Taf the Ghost]

5 hours ago, CarlBar said:

Ok Looked up the original article and this might actually be some kind of Athlon part. The problem area is here:

 

2DS104BBM4GH2_38/34_N

 

Using the same decoder they link.

 

2 means it''s an ES (Early Sample) chip.

 

D indicates Desktop whilst S normally means server. So i have no idea what DS means. Thats a bit of a red flag but could be yet another tweak to the naming scheme.

 

104 should be in order: The revision, the chiplet count, and the core count. According to that this is a revision 1, 0 chiplet, 4 core processor.

 

This is either a fake, a new naming scheme, (which suggests it's probably not a normal 3000 processor), or a non-Ryzen Processor. Athlon makes the most obvious sense TBH. Whilst i kinda expected those to be a chiplet design too, they might be going with a monolithic for that. Though why they'd be sending out ES now on that i'm not sure... And why would you test that on a flagship X570 board, at least in this fashion, (obviously you want to check basic compatibility i should think).

With the way the I/O-based chiplet designs have worked for ES names, it's "S104" that's the associated part. We haven't seen a letter in that 3rd spot this cycle, thus this is another twist with these ES samples. 

[COTG]

3 hours ago, GDRRiley said:

Why are most of you assuming it’s a ryzen 3. It could be an Athlon. 

Thats why to me it’s got lower clocks. 

If I can get 8 or 12 core at 4.2 or 4.5 OCed I’ll be supper happy. 

Goodbye my poor 6600k at 4.1

I really doubt that amd will make an 8 core athlon part. Sure, they probably could, but that would absolutely cannibalize anyone who bought their first or second gen ryzen, and most likely rally up a large section of their customer base. If anything, its a ryzen 3

[GDRRiley]

21 minutes ago, COTG said:

I really doubt that amd will make an 8 core athlon part. Sure, they probably could, but that would absolutely cannibalize anyone who bought their first or second gen ryzen, and most likely rally up a large section of their customer base. If anything, its a ryzen 3

It’s a quad core. They could make a quad core 8 thread. 

[PianoPlayer88Key]

For those of you talking about some not wanting AMD to increase core count, or talking about generational progress...

 

Spoiler

 

 

 

9 hours ago, xtroria said:

I really don't want AMD to increase the core count again. If AMD increases their core count and then Intel also increase their core count, we who bought the 6 core processors are just gonna get screwed

 

9 hours ago, cj09beira said:

your cpu wont loose performance, it will only loose resale value 

 

9 hours ago, PacketMan said:

Better tech improvements than getting 2-3 generations of the same hardware (i7 4790K, i7 6700K, i7 7700K, remember?)

If no one improves the specs the ones screwed will be everyone, it's normal that the newer hardware is better, but the old hardware won't be obsolete because of newer and more powerful hardware

 

8 hours ago, Stefan Payne said:

Yeah, that's the price of competition and progess, advancement in technology.

 

You youngster don't seem to know that there was a time where you chould throw away your old shit every year or two because the new stuff is +50-100% better.

 

Just look at when the 5V Pentium was released, succeded by the 100MHz set, the 200MHz Version came and AMD K6-2 or Pentium PRO or 2. That's a timeframe of around 2-3 Years...

 

 

But yeah, its better than having 4 Core CPUs for more than 10 years - 12 to be precise (2007 was the release of the Dual Core 2 Duo aka Core 2 Quad) and 4 cores with SMT2 also for almost 10 years as well.

 

No, its good that the Core Counts increase and that there is progress again after a 10 year old Hiatus in CPU Technology.

 

7 hours ago, Sunako said:

no one is getting screwed if they increase core counts. if a person bought a 6 core at the time it is what they could afford and met their needs. Just because there is a large leap forward and people that bought something in the past and something Better comes  out does not mean that people are geting screwed or ripped off.

 

4 hours ago, 2Buck said:

That's a reeeeeeeeeeally bad way of thinking.

 

4 hours ago, VegetableStu said:

as a person who saw their X99 processor get matched by an R5, I say GOOD, PLEASE GO AHEAD

 

 

About massive performance improvements, I say, BRING IT ON!! 

 

I want my next desktop build's CPU (around 2021/22) to be massively better than my current i7-4790K at the same price I paid in January 2015 (~$330 IIRC) - for example able to render a movie based on the Cinebench scene at 30 or 60 fps, or able to do a Handbrake 4K H.265 render (one that's 0.5 fps on my current cpu) faster than my current CPU does Wave -> mp3 encoding.

 

I WANT my 4790K to be so hopelessly outclassed by then, that even someone who was legitimately in dire need of a CPU (like the tech equivalent of people on the street corners with "help needed, hungry, God bless, etc" signs) would throw my gift of a 4790K back in my face and scream "THROW THIS £μ¢κING SHIT AWAY! This is worse now than a Pentium 1 with the unpatched FDIV bug" (or whatever was the last CPU generation to be not able to run whichever 2nd or 3rd previous major version of Windows that had its extended support ended BEFORE Haswell's release) "was when your 4790K was new!"

 

I wish EACH generation was as big of a jump as we had in the 8086 -> 286 -> 386 -> 486 -> Pentium days.  I would really like to see the industry get to where the lowest-end mobile CPU from a new generation (like Celeron/Atom/Quark -Y / Sempron/Duron (<2-watt TDP)) would crush the top multi-socket previous-generation server CPU (like 8x Xeon Platinum/E7 / Epyc), like an RTX 2080 Ti beats Intel Graphics Media Accelerator in gaming.

 

 

 

Spoiler

 

2021 is about when I'd guess DDR5, PCI Express 5.0, and Socket AM5 should be coming out.  My Corsair AX760's warranty, I believe, should expire around January 2022.  I'm hoping to buy the parts for a new build around Black Friday 2021.

 

If that "2nd-last supported version of Windows ended before Haswell" was Windows 2000 or ME, then the CPU the quoted person compares the 4790K to wouldn't have even met the minimum requirements to run Windows 98 or 95.

 

 

 

 

I don't mind having to upgrade my CPU again after a few years, and yet again a few years after that.

Where I DO take issue though, is making me replace my motherboard so often.  Yes I'll have to do it this time, but going forward, I'd like to see physical CPU sockets, DIMM slots, etc. have as much longevity, forward/backward compatibility inter-brand interchangeability, etc, as PCI Express, USB type A, 4-pin Molex power, maybe even the QWERTY layout.

 

It's not so much the cost of a new motherboard I take issue with, it's the labor involved in swapping it out - you have to unplug EEEVEERRYYYTHIINNG from the old mobo, then after replacing it, plug it AALLLL back in, essentially.  If only it was as easy to swap a mobo as it is to swap a DIMM when you're using an AIO or push-pin stock heatsink, or swap a front-port-mounted USB flash drive...

 

[mr moose]

4 minutes ago, PianoPlayer88Key said:

 

I don't mind having to upgrade my CPU again after a few years, and yet again a few years after that.

Where I DO take issue though, is making me replace my motherboard.  Yes I'll have to do it this time, but going forward, I'd like to see physical CPU sockets, DIMM slots, etc. have as much longevity, forward/backward compatibility inter-brand interchangeability, etc, as PCI Express, USB type A, 4-pin Molex power, maybe even the QWERTY layout.

 

 

 

The issue is that the very nature of CPU's means that their layout changes are incredibly necessary in order to facilitate improvement (sometimes even minor improvement).   Imagine trying to replace the GPU chip in your graphics card but not the card itself, that is why expansions slots are so long time and versatile, like USB, etc.   The other problem with trying to maintain motherboards for any length of time is that when you start talking about improvements in the CPU you need the motherboard to be able to physically support said changes,  I.E when the cpu moves to supporting ddr5 or has more cores than the motherboard was designed to handle, even if you could force the CPU to work in the old motherboard you would not be able to use all the cores in the be CPU and you would be limited to slower ram.

 

Basically it's just not feasible for the motherboard to remain the same over any lengthy period.

[PianoPlayer88Key]

16 minutes ago, mr moose said:

 

The issue is that the very nature of CPU's means that their layout changes are incredibly necessary in order to facilitate improvement (sometimes even minor improvement).   Imagine trying to replace the GPU chip in your graphics card but not the card itself, that is why expansions slots are so long time and versatile, like USB, etc.   The other problem with trying to maintain motherboards for any length of time is that when you start talking about improvements in the CPU you need the motherboard to be able to physically support said changes,  I.E when the cpu moves to supporting ddr5 or has more cores than the motherboard was designed to handle, even if you could force the CPU to work in the old motherboard you would not be able to use all the cores in the be CPU and you would be limited to slower ram.

 

Basically it's just not feasible for the motherboard to remain the same over any lengthy period.

Hmm, then is there any way to have motherboards be easier to replace, more modular, etc?  If only...

 

 

That, combined with this and

Spoiler

 

 

 

 

 

 

 

 

 

[Stefan Payne]

53 minutes ago, PianoPlayer88Key said:

I want my next desktop build's CPU (around 2021/22)

(...)

I WANT my 4790K to be so hopelessly outclassed by then,

Yeah, that's 8 Years or so.

Back in 2006 when the Core 2 Duo was released, 8 years before that was 1998. And the best thing you could by in April 1998 was a 400MHz Pentium 2.

 

8 Years before that was 1990. The fastest thing you could buy in May 1990 was an Intel i80486 with a whopping 33MHz. And 27 Mips.

 

8 Years before that was 1982.

And the fastest thing you could get at the time was a 6MHz 80286,

 

 

And now you expect a CPU to be used 10 years or more?!
Why?!
 

In recent history, that was never the case that a 10 year old CPU was a good idea. In some cases it somewhat worked but everything ran like shit on it - if at all.

 

Just look at this link and look at a timeframe of a couple of years in the 80s, 90s or early-mid 2000s:
http://processortimeline.info

 

Never in the History of PC you could use Hardware as long as today - because there was no innovation, because there was a standstill and someone was more interested in increasing profit than to make a better product - for 10 years, not much happened in the CPU Field...

 

That an i7-920 (even if overclocked) is viable in 2019 is a joke because that shouldn't be the case...

 

an 80286 even with 16MHz was no fun in the early 90s, when I bought it for 700 Deutsche Mark with an awesome 14" Screen...

[Guest]

Looking at the fx-8150 AMD wasn't interested in either making money nor making a better product. You can't blame Intel for that. Wait nvm you'll find a way anyway.

[COTG]

46 minutes ago, Stefan Payne said:

And now you expect a CPU to be used 10 years or more?!
Why?!

Because in the past large performance increases were possible due to significant node shrinks and otherwise architecture improvements, due to the industry itself growing massively and gaining a large influx of cash.

Nowadays we are quite literally at/almost at the physical limitations of how small we can make our manufacturing process, as quantum mechanics will come into effect messing with everything.  
Of course Intel was purposefully keeping gen over gen improvements to a minimum to maximize profit, but even if they were not, we will not see performance improvements anywhere near to the old days anymore. This is part of why we are now purely increasing core count in order to obtain more performance.

But for an average consumer, a  current 4-6 core cpu will most likely  be perfectly viable for the next 10 or so years,.

[Stefan Payne]

1 minute ago, COTG said:

But for an average consumer, a  current 4-6 core cpu will most likely  be perfectly viable for the next 10 or so years,.

8 Core Desktop CPUs are how long available?? ~2 Years?
And how many games are there that benefit from those cores or want them?? A couple??

 

 

Conclusion:
If it took just 2 Years for Games that want more than 4 cores to show up, there is a demand for many core CPUs, even if you don't want to see it...

 

If you want to buy for the future, you need at least 8 Cores. 16 would be better because we will see an increase in core count...

 

Your claim that 4-6 Cores will be enough for the next 10 years is just outright false because there are a couple of games that already run like shit when on just 4 Cores. They demand at least 6, better 8!

6 minutes ago, COTG said:

Because in the past large performance increases were possible due to significant node shrinks and otherwise architecture improvements,

 Nothing has really changed.

A Core i7-920 was around 300mm² and sold under 200€.

A Core i7-7700 is just 122mm²...

That is less than half the Bloomfield CPU and 

The smaller Ivy Bridge Die was also only 133mm² small.

 

And now you claim seriously that its not possible?!
That's just not true. It was at least since 22nm viable to put 6 Cores into Desktop. The Ivy-E6 was only 257mm small, heck even the Evy E-10 was only 341mm² (10 Cores).

 

It was more than viable at the time to introduce 6 cores for the average user and bring the cost of 4 Cores down.

 

So no, the Lack of Innovation was not a technical reason at all.

 

 

[Stefan Payne]

54 minutes ago, Medicate said:

Looking at the fx-8150 AMD wasn't interested in either making money nor making a better product. You can't blame Intel for that. Wait nvm you'll find a way anyway.

You know, 1 060 000 000 €uro??

http://europa.eu/rapid/press-release_IP-09-745_en.htm

 

That influenced the development of Bulldozer dramatically...

Because Bulldozer wasn't a bad architecture, AMD just didn't have enough money to develop the right front end for that and a decent Northbridge because when the clock of the Northbridge (=Memory Controller + L3 Cache and other stuff as well) is equal to the CPU Clock, you know that here is a Problem...

Because people who did some Benchmarking and overclocking of Vishera CPUs claimed that 100MHz higher Nothrbridge Clockrate was in some games equal (or even better) than +100MHz Core Clock.

 

Also Planet3dNow Did a good review once:

https://www.planet3dnow.de/cms/18564-amd-piledriver-vs-steamroller-vs-excavator-leistungsvergleich-der-architekturen/

 

That shows that in some situations the "IPC" at the same clockrate increased by 30% and in some cases even +100%...

 

SO yeah, it was lack of Money that caused Bulldozer...

[COTG]

4 minutes ago, Stefan Payne said:

8 Core Desktop CPUs are how long available?? ~2 Years?
And how many games are there that benefit from those cores or want them?? A couple??

 

 

Conclusion:
If it took just 2 Years for Games that want more than 4 cores to show up, there is a demand for many core CPUs, even if you don't want to see it...

 

If you want to buy for the future, you need at least 8 Cores. 16 would be better because we will see an increase in core count...

 

Your claim that 4-6 Cores will be enough for the next 10 years is just outright false because there are a couple of games that already run like shit when on just 4 Cores. They demand at least 6, better 8!

 Nothing has really changed.

A Core i7-920 was around 300mm² and sold under 200€.

A Core i7-7700 is just 122mm²...

That is less than half the Bloomfield CPU and 

The smaller Ivy Bridge Die was also only 133mm² small.

 

And now you claim seriously that its not possible?!
That's just not true. It was at least since 22nm viable to put 6 Cores into Desktop. The Ivy-E6 was only 257mm small, heck even the Evy E-10 was only 341mm² (10 Cores).

 

It was more than viable at the time to introduce 6 cores for the average user and bring the cost of 4 Cores down.

 

So no, the Lack of Innovation was not a technical reason at all.

 

 

i think you are confusing a multiple of things here. Average consumer does not mean gamer. It means the 90% of people who don't need their computer for anything besides browsing and sometimes light compute. 

Secondly, when I say shrink in manufacturing process, I dont mean die area. Sure, die area might have not changed/gotten smaller, but the lithography process. i7-920 was build on 45nm, whereas the i7-7700 was on 14nm. Every further shrink in the process gets exponentially harder, and the performance gain is not linear which is the result of the smaller and smaller performance increases we are seeing.

[MadmanRB]

1 hour ago, Stefan Payne said:

Yeah, that's 8 Years or so.

Back in 2006 when the Core 2 Duo was released, 8 years before that was 1998. And the best thing you could by in April 1998 was a 400MHz Pentium 2.

 

8 Years before that was 1990. The fastest thing you could buy in May 1990 was an Intel i80486 with a whopping 33MHz. And 27 Mips.

 

8 Years before that was 1982.

And the fastest thing you could get at the time was a 6MHz 80286,

 

 

And now you expect a CPU to be used 10 years or more?!
Why?!
 

In recent history, that was never the case that a 10 year old CPU was a good idea. In some cases it somewhat worked but everything ran like shit on it - if at all.

 

Just look at this link and look at a timeframe of a couple of years in the 80s, 90s or early-mid 2000s:
http://processortimeline.info

 

Never in the History of PC you could use Hardware as long as today - because there was no innovation, because there was a standstill and someone was more interested in increasing profit than to make a better product - for 10 years, not much happened in the CPU Field...

 

That an i7-920 (even if overclocked) is viable in 2019 is a joke because that shouldn't be the case...

 

an 80286 even with 16MHz was no fun in the early 90s, when I bought it for 700 Deutsche Mark with an awesome 14" Screen...

Thing is all things considered the core 2 duo still holds up for modern tasks like web browsing and multitasking,

The thing you seemed to forget is we are at the end of Moores law, so massive performance jumps are not as common.

And yes I say that an i7-920 is still good even by modern standards though if you are using one to game its certainly time for an upgrade.

Moores law has come to its logical conclusion and the only way forward is to try something new.

Hard to do really but both AMD and Intel no doubt have plans.

[Guest]

Plot-Twist: This is the new AMD Athlon

[Derangel]

10 hours ago, VegetableStu said:

as a person who saw their X99 processor get matched by an R5, I say GOOD, PLEASE GO AHEAD

Very much this. I have been using a X99 system since 2016, I am more than ready for Zen 2.

[mr moose]

9 hours ago, PianoPlayer88Key said:

Hmm, then is there any way to have motherboards be easier to replace, more modular, etc?  If only...

 

I don't want to say no, because the idea certainly is interesting, but I feel it is probably unlikely.

[williamcll]

Didn't the previous rumours say the ryzen 3 3000 would start as a hexa core chip?

[CarlBar]

17 hours ago, Taf the Ghost said:

With the way the I/O-based chiplet designs have worked for ES names, it's "S104" that's the associated part. We haven't seen a letter in that 3rd spot this cycle, thus this is another twist with these ES samples. 

 

That does make more sense, but raises even more questions than it answers.

 

3 hours ago, VegetableStu said:

mmmmnnnn... if the entire R3 stack for 3000 are all 4 cores only, I wouldn't put away the possibility that the dual-die R9 would cap out at 12 cores (in turn, a 16 core TR 3000 part might make a bit more sense, although the price jump still wouldn't make it preferable to anything on the AM4 desktop stack (e.g. 1800X vs 1900X, or 2700X vs 2950X for completion's sake))

 

so something like

R3: 4c8T (as usual, instead of 6c12T)

R5: 6c12T (as usual, instead of 8c16T)

R7: 8c16T (as usual, instead of 12c24T)

R9: 12c24T (instead of 16c18T)

TR: 16c32T onwards (instead of possibly clashing with a 16c R9, and having to start more expensive at 24c?) (although 16 cores how? 4x4c or 2x8c?)

 

EDIT: heck this, while I'm here

 

TR is going to be interesting but i don't think we'll see a 16 core part for TR 3000. 16 core has clearly been an AM4 consideration this gen and whilst they may not launch it out the gate, i don't think there's really any question that it's going to arrive.

 

4 hours ago, williamcll said:

Didn't the previous rumours say the ryzen 3 3000 would start as a hexa core chip?

 

As discussed further up we don't actually know for sure this isn't an athlon, and even if it is the R3 it's possibble marketing decisions surrounding the possibble 16 core skew mean they decided to reshuffle the product stack. Naming and pricing schemes are the things that can change at the last minute the most.

[Derangel]

11 hours ago, mr moose said:

I don't want to say no, because the idea certainly is interesting, but I feel it is probably unlikely.

Pretty sure its not feasible to do. At least not within the confines of the various current motherboard standards and not without extreme cost to the end user.

[Taf the Ghost]

10 hours ago, williamcll said:

Didn't the previous rumours say the ryzen 3 3000 would start as a hexa core chip?

Yes, but there will be 4c variants of the APU chips. There's 2c variants of the current APU chips, they're just called Athlons.