Ryzen 3000 release date leak

[Stefan Payne]

J

4 hours ago, VegetableStu said:

NDA? 可以吃的吗?

(i realise some of their staff aren't ethnic chinese, but bear with the joke for the moment ,_,)

Can you eat it?? (according to Google Translate)

And is it possible in japanese.

 

[Stefan Payne]

3 hours ago, GoldenLag said:

Probably just a white lie from the customer service people. while i would love for it to be released in April, the fact AMD is considering or has delayed Navi due to 7nm avavibility

, i dont think they will be able to build up stock in such a short ammount of time. and if they have, i fear it will be a paper launch. 

There is a solution to that problem:
Only release 12 and 16 cores first -> high priced stuff usually doesn't sell that well, so that gives you time to make other things.

AMD does that every time they don'T have good yields.

The best example I remember was the K6-3, wich was rediculously expensive compared to the K6-2 but also came with 256k integrated L2 in 0,25µ (250nm!) while the K6-2 didn't come with L2 Cache.

 

2 hours ago, GoldenLag said:

its a polaris replacement. current rumours point to changes in the CU/stream processor including the back end. 

Yeah, we do not know much about Navi at this time. Its said to be a Polaris replacement from the price range but its possible that they attack the 2070 or so with it.

With a very reasonable price...

2 hours ago, cj09beira said:

its both a major release and amd's 50th aniversary so there is a good chance they will want to do their own thing

Yeah, that's very reasonable to assume and that AMD might also mention navi there as it  is a huge event and they might want something to rock the boat.

 

1 hour ago, MeatFeastMan said:

Wow 4-5 weeks? Maybe they are worried that Intel will come from nowhere and release Ice Lake early.

Wouldn't be happening as they alredy messed up their 10nm stuff a couple of times with very bad yields and other shit.

 

I don't get why people are claiming that Intel will come with the answer around the corner when there is nothing to support that claim. They had to redo their 10nm stuff, so they stand there without any manufacturing.

And I doubt that TSMC would produce for Intel.

So there is nothing for them.

1 hour ago, MeatFeastMan said:

This was always going to be AMD's year and they would never want Intel to drop a bomb on them. You get a feeling Ice Lake is a major improvement over Coffee Lake, even bigger than the zen+ to zen 2 gap.

Yeah, like Coffeelake is an improvement of Kaby Lake...

Oh wait, it isn't.

ANd the efficiency of the Intel CPUs is rather bad right now.

For the same actual power consumption, there isn't really a performance difference. The 9900K is only better because it is allowed to pull 200W through the CPU and only then it is a decent thing.

With 95W Hard Limit the difference shrinks...

 

But hey, when Intel does a CPU that sucks 250W through the die, nobody cares.

But if someone else does it...

 

[Herman Mcpootis]

26 minutes ago, Stefan Payne said:

J

Can you eat it?? (according to Google Translate)

And is it possible in japanese.

 

Nah, just chinese for "can you eat it?"

[GoldenLag]

16 minutes ago, Stefan Payne said:

But hey, when Intel does a CPU that sucks 250W through the die, nobody cares.

But if someone else does it...

general rule in tech: People dont actually care about powerdraw. They just suddenly do when they need to shit on a company. 

 

Spoiler

sources: Just about every GPU launch with high powerdraw. 

 

[Blademaster91]

2 minutes ago, GoldenLag said:

general rule in tech: People dont actually care about powerdraw. They just suddenly do when they need to shit on a company. 

 

  Reveal hidden contents

sources: Just about every GPU launch with high powerdraw. 

 

The CPU power draw doesn't matter as much IMO if it has the performance to back up the higher power draw, in GPU's it does as not everyone wants a hot and loud card.

[GoldenLag]

1 minute ago, Blademaster91 said:

The CPU power draw doesn't matter as much IMO if it has the performance to back up the higher power draw, in GPU's it does as not everyone wants a hot and loud card.

again. they dont actually care about powerdraw. they care about the bi-products. not the powerdraw itself. and the biproducts can be mittigated. 

[Stefan Payne]

18 minutes ago, Blademaster91 said:

, in GPU's it does as not everyone wants a hot and loud card.

That's total horse shit, because you can mitigate it with a good cooling solution.

 

Source:
My experience with the Sapphire VEGA64 Nitro+ and the Power Color RX480 Red Devil.

According to your claim the Power Color should be quieter because lower Power Consumption - WRONG, the Sapphire is quieter under load - and cooler as well...

 

The Power Color is somewhere around 80°C under full load, the Sapphire at 60 or so...

 

Quote

general rule in tech: People dont actually care about powerdraw. They just suddenly do when they need to shit on a company. 

Exactly!!
Hardly anyone cares about Powerdraw.

They care about Arguments for their team and against the other team.

[Blademaster91]

18 minutes ago, GoldenLag said:

again. they dont actually care about powerdraw. they care about the bi-products. not the powerdraw itself. and the biproducts can be mittigated. 

Most gamers won't, but I think it's ironic to be sh#tting on a CPU that uses 200w when it was designed for overclocking and performance, and the targeted consumer isn't going to care about power draw or the heat from it.

[GoldenLag]

8 minutes ago, Blademaster91 said:

but I think it's ironic to be sh#tting on a CPU that uses 200w when it was designed for overclocking and performance

people dont shit on the CPU, people shit on the spec given, and the fact 6/7 mobo vendors completely ignore the spec on their out of the box config. and then you have the temporary violation of TDP spec that boosts do. (which people generally dont care about, because the cooler can handle it)

8 minutes ago, Blademaster91 said:

and the targeted consumer isn't going to care about power draw or the heat from it.

true, but it further inforces the point that peopl dont care about powerdraw. they might not like the biproducts of it. people care when they are given something that doesnt follow its own spec, or when the included parts dont work well enough with the Part they bought. (looking at you stock intel cooler on the 8700)

8 minutes ago, Blademaster91 said:

Most gamers won't,

there are offcourse exceptions, but i can tell we both agree on that. 

[Herman Mcpootis]

33 minutes ago, GoldenLag said:

general rule in tech: People dont actually care about powerdraw. They just suddenly do when they need to shit on a company. 

 

12 minutes ago, Stefan Payne said:

Exactly!!
Hardly anyone cares about Powerdraw.

They care about Arguments for their team and against the other team.

TBF they also worry if they have a low wattage or crappy PSU, but then you should be replacing a crappy PSU anyways. wattage thing ends up depending more on the situation.

[GoldenLag]

1 minute ago, Herman Mcpootis said:

 

TBF they also worry if they have a low wattage or crappy PSU, but then you should be replacing a crappy PSU anyways. wattage thing ends up depending more on the situation.

the only place where wattage is truly a concern is very low profile systems that may have limited PSU capability or ventialation. 

 

and if your PSU can handle a higher powerdraw GPU, you should be swapping it anyway. just generally avoid pairing a 2080ti or vega 64 with a cx450 (2017). 

[Herman Mcpootis]

1 minute ago, GoldenLag said:

the only place where wattage is truly a concern is very low profile systems that may have limited PSU capability or ventialation. 

 

you might get things like a 300w PSU with a standard height pci-e slot prebuilt but with proprietary connectors or form factor, so you end up supposedly being able to fit most GPUs but the PSU limits what you can use. it's not that common but it can happen.

[Doobeedoo]

Well, that's looking quite close, though would be cool to see it as sooner than thought. I take it they're more or less ready by now possibly. 

[porina]

1 hour ago, GoldenLag said:

in terms of dekstop we will have to see. expect 10-20% IPC upgrade from Intel as they widen their architecture with minor or no coreclock upgrade but much better thermals. 

Where do numbers like this come from? I'd severely doubt it'll be more than low single digit % on average. Maybe if you look at a very specific test condition there are scenarios where there are much bigger jumps. Note I've said similar about some of the other claims of Zen2 performance over Zen+. We know the FPU is getting a significant upgrade, but outside of that it will be a much smaller difference.

 

Quote

the only intel has is Timing. by the time they have something on the market, people will be waiting for Zen 3. another question is if they will wait late 2020 for DDR5 plattform as they are seemingly extending their coffeelake design to "long-boy" 10 core. 

Have AMD actually said Zen 3 will immediately follow Zen 2? For example, is there going to be a Zen 2+ first? That would make more sense as they can refresh more easily in DDR4 without needing a new platform beyond AM4 (AM5?) for DDR5.

[boey]

33 minutes ago, porina said:

Note I've said similar about some of the other claims of Zen2 performance over Zen+.

 

Have AMD actually said Zen 3 will immediately follow Zen 2? 

-Didn’t they say something like 20+% increase in IPC for Epyc during the Next Horizon event? Rome uses the same cores as Ryzen so I’d assume it’d be similar

 

-Yeah, they did

 

 

[porina]

15 minutes ago, _d0nut said:

-Didn’t they say something like 20+% increase in IPC for Epyc during the Next Horizon event? Rome uses the same cores as Ryzen so I’d assume it’d be similar

They also carefully stated it was in a particular compute condition, not a general expectation. I'd guess, it was one that uses a fair amount of the much faster FPU in Zen 2.

15 minutes ago, _d0nut said:

-Yeah, they did

Thanks.

[GoldenLag]

49 minutes ago, porina said:

Have AMD actually said Zen 3 will immediately follow Zen 2? For example, is there going to be a Zen 2+ first? That would make more sense as they can refresh more easily in DDR4 without needing a new platform beyond AM4 (AM5?) for DDR5

AMD's roadmap points to Zen 3 after Zen 2. AM4 is seemingly for the most part sticking to Ddr4. Idk what custom solutions they may come up with.

 

51 minutes ago, porina said:

Where do numbers like this come from? I'd severely doubt it'll be more than low single digit % on average. Maybe if you look at a very specific test condition there are scenarios where there are much bigger jumps.

Iirc there has been mention of a widening of Intel's architecture with icelake. No specific numbers, but they usually ballpark around 20%. 

 

Intel might have mentioned something when discussing their roadmap and future products. I dont rememher clearly.

[cj09beira]

3 hours ago, Stefan Payne said:

 

But hey, when Intel does a CPU that sucks 250W through the die, nobody cares.

But if someone else does it...

 

MOB:we Ready!!!

 

2 hours ago, porina said:

Where do numbers like this come from? I'd severely doubt it'll be more than low single digit % on average. Maybe if you look at a very specific test condition there are scenarios where there are much bigger jumps. Note I've said similar about some of the other claims of Zen2 performance over Zen+. We know the FPU is getting a significant upgrade, but outside of that it will be a much smaller difference.

 

Have AMD actually said Zen 3 will immediately follow Zen 2? For example, is there going to be a Zen 2+ first? That would make more sense as they can refresh more easily in DDR4 without needing a new platform beyond AM4 (AM5?) for DDR5.

zen+ is from a architecture standpoint the exact same design, with tightened L1 and L2 caches, so they had a very long time to improve on the design, IF should also now be more than 2 times as fast, the instruction cache was reworked too, and some say its one of zen's problems (no clue if it is or not)

 

if i am not mistaken we know very little about it, one thing we do know is that it will have a epyc version of it still on sp3 (so probably still ddr4) and should be on 7nm euv 

 

[mxk]

8 hours ago, dgsddfgdfhgs said:

oh .....4-5 weeks from now is the April's fool.......

Ohh shit you're right

[justpoet]

I, for one, care about power draw, and more importantly how much work per watt), since I run towers as servers and then use my desktop pretty close to 24/7 as well.  I'm fine with max output power being higher (such as during a video/3d render), but an average or normal draw needs to be reasonably low (same with graphics card) while I'm doing the usual tasks and surfing/researching various stocks/industries/companies/solutions.

 

It honestly makes a HUGE difference in electric bill each month.

 

I'm also surprised we don't see more of the ICE car concepts coming into CPUs and GPUs more often, such as when a v8 only uses more than 4 cylinders when you mash the throttle, then goes back to isolating 4 of them to make it more efficient again.  Unlike with car engines, under normal light workloads there's really no penalty for shutting off a core until sustained load says you REALLY need it (which I'm pretty sure is already done fairly commonly in the phone space with ARM architectures), and of course a flag you can set to keep them all always on or only use specific ones if you're going for single thread OCs would then be pretty easy to add as well.  Then everybody can have their cake and eat it too.

[porina]

2 hours ago, cj09beira said:

zen+ is from a architecture standpoint the exact same design, with tightened L1 and L2 caches, so they had a very long time to improve on the design, IF should also now be more than 2 times as fast, the instruction cache was reworked too, and some say its one of zen's problems (no clue if it is or not)

I have no doubt that the updates will make the CPU better in IPC, but my caution is that changes only affect things that are sensitive to them. 

 

Taking a current example, say you have Ryzen and compare single and dual channel ram with the same speed, so that IF speed is not a variable, only the ram bandwidth. If you run an APU, then the ram bandwidth makes a good difference. If you run Cinebench R15, it'll make hardly any difference at all.

[justpoet]

38 minutes ago, VegetableStu said:

yeah i wonder what would it take for an ARM+CSIC hybrid system where it could run pretty simple apps on the ARM core, then when it comes to workstation programs the system would boot the intel/AMD CPU and take over o_o

 

first beneficiaries would be laptops, but yeah

It doesn't even need to be a hybrid system.  Just undervolt, underclock, and turn off half the cores until they're really needed.  Most power management in CPUs already does at least part of this, so going the next step would seem like a natural progression.

[cj09beira]

55 minutes ago, porina said:

I have no doubt that the updates will make the CPU better in IPC, but my caution is that changes only affect things that are sensitive to them. 

 

Taking a current example, say you have Ryzen and compare single and dual channel ram with the same speed, so that IF speed is not a variable, only the ram bandwidth. If you run an APU, then the ram bandwidth makes a good difference. If you run Cinebench R15, it'll make hardly any difference at all.

thats mostly because cinebench fits all inside the caches, it doesn't even need to be cached in ram, which can deduced as their is no difference in performance from memory bandwidth or latency,

except for avx most other improvements should be quite relevant for our daily use, 2.3x IF bandwidth should mean better ccx to ccx communication (important as windows loves to move threads), op and instruction cache changes should also be quite useful for most applications, 

[porina]

1 hour ago, cj09beira said:

except for avx most other improvements should be quite relevant for our daily use, 2.3x IF bandwidth should mean better ccx to ccx communication (important as windows loves to move threads), op and instruction cache changes should also be quite useful for most applications, 

My point remains, these changes may significantly increase specific areas, but overall you're still going to be doing much of the same stuff as before. I'd be amazed if we get a wide ranging non-AVX speed IPC increase 10%+. Some very specific scenarios may exceed that, but I really can't see it as a general improvement.

[MetEishYa]

21 hours ago, _d0nut said:

These shops usually don't have the balls to lie to customers, they want as many sales as possible so as to get profit. They usually violate NDAs to release products sooner than the competition to make money too. Sounds really fishy but that's how it is here. I've seen people get 'scammed' often at these shops, one guy walked in asking for a motherboard replacement for his i5-8400(he was mostly clueless and asked for the cheapest motherboard) and walked away with an ROG STRIX Z370, thinking it was a regular low end board. He was slightly overcharged for it too.

Not surprised they sell before the release date considering they in Sim Lim and there are plenty of stores all trying to sell the same products. They know that once a customer leaves the store the chances of him buy there are much less, so try up sell and haggle.