Very Excited About Upcoming 7nm Ryzen, but...

Yeah, it sounds like Zen 2 8 core is going to match the 9900k and embarrass the 9700k, so prices on that will have to fall significantly. I expect Zen 2 8-core to launch at the same price as 2700x is currenty, leaving plenty of room up-top for the more expensive 12-core, and possibly 16-core on launch day.
 
Last edited:
AMD said the AM4 socket was good to 2020...

TO 2020, not thru 2020...

So maybe sometime in 2020 we get an AM4+ / AM5 socket, along with quad-channel DDR5 memory (and maybe high-speed low-latency 32GB & 64GB sticks)...?!?
 
Boil said:
AMD said the AM4 socket was good to 2020...

TO 2020, not thru 2020...

So maybe sometime in 2020 we get an AM4+ / AM5 socket, along with quad-channel DDR5 memory (and maybe high-speed low-latency 32GB & 64GB sticks)...?!?
Click to expand...

Nope, AM5 wil likely l remain dual-channel DDR5. The secret to this is that DDR5 is using a dual bus for better utilization of each memory module, giving you about 35% actual bandwidth at the same bus transfer rates. And it's looking to START at 3200, and jump all the way up to 6400.

Unfortunately, they wil need a new socket for the new pin outs for DDR5, and will likely pair it with 5nm EUV Zen 3 in 2021/2022. Zen2+ will still be on DDR4, and will probably not touch much.

Right now DDR5 DIMMs will be available by the end of the year, and they are expected to show up in servers first, then high-end cell phones and mainstream desktops. AMD's independent I/O die means that you can use DDR5 on EPYC server refresh starting next year, while still using the same refresh chips on DDR4 mainstream systems.
 
Last edited:
defaultluser said:
Nope, AM5 wil likely l remain dual-channel DDR5. The secret to this is that DDR5 is using a dual bus for better utilization of each memory module, giving you about 35% actual bandwidth at the same bus transfer rates. And it's looking to START at 3200, and jump all the way up to 6400.

Unfortunately, they wil need a new socket for the new pin outs for DDR5, and will likely pair it with 5nm EUV Zen 3 in 2021/2022. Zen2+ will still be on DDR4, and will probably not touch much.

Right now DDR5 DIMMs will be available by the end of the year, and they are expected to show up in servers first, then high-end cell phones and mainstream desktops. AMD's independent I/O die means that you can use DDR5 on EPYC server refresh starting next year, while still using the same refresh chips on DDR4 mainstream systems.
Click to expand...

https://www.tomshardware.com/news/what-we-know-ddr5-ram,39079.html

DDR5-3200 RAM will see an increase of 1.36x in bandwidth compared to DDR4-3200. However, DRAM chips are expected to ship with a bandwidth of 4800MT/s, or 1.87x that of DDR4-3200 RAM. The official upper limit for the DDR5 RAM standard is 6400MT/s, but some designs may be able to push that further through overclocking.
Click to expand...

Sorry for the THG link :) . But it explains the numbers a bit better :) .
 
tangoseal said:
Zen 2 12 core is going to smoke 9900k
Click to expand...

In highly threaded encoding/rendering workloads, certainly.

The question is what happens in more typical workloads, when you are more likely to pin a single core, with much lower utilization on other cores.

Single core Cinebench scores are what I am going to be looking for as a better predictor of general purpose performance.
I'm excited to see where 7nm Ryzen lands in this regard.
 
Zarathustra[H] said:
In highly threaded encoding/rendering workloads, certainly.

The question is what happens in more typical workloads, when you are more likely to pin a single core, with much lower utilization on other cores.

Single core Cinebench scores are what I am going to be looking for as a better predictor of general purpose performance.
I'm excited to see where 7nm Ryzen lands in this regard.
Click to expand...


Single core benchmark as a better predictor of general performance???

You need to look at both of them to make a good decision.

I'd say they're equally important or else we'd be thinking unlocked dual cores are the way to go because their higher single threaded performance was a better indicator of general perf!

The people here looking to buy the higher core count CPUs are not going to be paying attention to single threaded perf AS much as MT. However I'm 100% certain that AMD will have a higher clocked CPU with less cores for the people that like ST to be high.
 
Rockenrooster said:
Single core benchmark as a better predictor of general performance???

You need to look at both of them to make a good decision.

I'd say they're equally important or else we'd be thinking unlocked dual cores are the way to go because their higher single threaded performance was a better indicator of general perf!

The people here looking to buy the higher core count CPUs are not going to be paying attention to single threaded perf AS much as MT. However I'm 100% certain that AMD will have a higher clocked CPU with less cores for the people that like ST to be high.
Click to expand...
Honestly, if the 12c 24t processor lags just a little behind in single threaded performance, I will still jump all over it. 300% core count increase over my 7700k... Yes please!
 
Rockenrooster said:
Single core benchmark as a better predictor of general performance???

You need to look at both of them to make a good decision.

I'd say they're equally important or else we'd be thinking unlocked dual cores are the way to go because their higher single threaded performance was a better indicator of general perf!

The people here looking to buy the higher core count CPUs are not going to be paying attention to single threaded perf AS much as MT. However I'm 100% certain that AMD will have a higher clocked CPU with less cores for the people that like ST to be high.
Click to expand...


Not what I intended to say.


My point is that you know how many cores you have. You don't need to test to verify this. Scaling with added cores isn't all that different from arcitecture to architecture.

If you need more cores, choose the model with more cores, if you need fewer cores, choose the model with fewer cores.

Where the key difference comes in is how capable each core is. "IPC"*clock speed.

Very rarely have I had a workload that is limited by my number of cores. Typically when I am CPU limited (which is rare) it is because I have a bunch of cores that are either idle or loaded no more than ~50% and one core that is pinned at 100% holding everything back.

This situation is improved by adding more per core performance, NOT by adding additional cores.
 
Last edited:
Zarathustra[H] said:
Not what I intended to say.


My point is that you know how many cores you have. You don't need to test to verify this. Scaling with added cores isnt all that different from argcitecture to architecture.

If you need more cores, choose the model with more cores, if you need fulewer cores, choose the model with fewer cores.

Where the key difference comes in is how capable each core is. "IPC"*clock speed.

Very rarely have I had a workload that is limited by my number of cores. Typically when I am CPU limited (which is rare) it is because I have a bunch of cores that are either idle or loaded no more than ~50% and one core that is pinned at 100% holding everything back.

This situation is improved by adding more per core performance, NOT by adding additional cores.
Click to expand...

That makes more sense.
 
Zarathustra[H] said:
Not what I intended to say.


My point is that you know how many cores you have. You don't need to test to verify this. Scaling with added cores isn't all that different from arcitecture to architecture.

If you need more cores, choose the model with more cores, if you need fewer cores, choose the model with fewer cores.

Where the key difference comes in is how capable each core is. "IPC"*clock speed.

Very rarely have I had a workload that is limited by my number of cores. Typically when I am CPU limited (which is rare) it is because I have a bunch of cores that are either idle or loaded no more than ~50% and one core that is pinned at 100% holding everything back.

This situation is improved by adding more per core performance, NOT by adding additional cores.
Click to expand...

These are concepts taught in baseline virtualization domains that are further reinforced when you get closer to being the lead architect.

Being wrong in workload core density scaling vs core frequency can be incredibly costly.
 
Personally I'm at disappointment at Ryzen 3000 series specs so far, because I was expecting leaked specs(yes I'm a fool for believing leaks from known AMD fanboys) or at least similar..

According to leaked specs, we were suppose to get four CPU's(8 cores, 12 cores & 2x 16 cores) running @ 4.7-5.1GHz. and 3 of em(12 cores & 2x 16 cores) were clearly better than i9 9900K at lower price.
jZNBO3P.jpg

But we didn't get ANY 16 core CPUs, and on average Ryzen 3000 series are 400MHz slower than leaked specs.
We only get one 12 core CPU @ 4.6 GHz. Rest are 6-8 core CPUs @ 4.2-4.5GHz.. sigh

AMD Leaked Spec Formula = X(likely spec) + 400MHz + 4 cores
 
gintama6888 said:
Personally I'm at disappointment at Ryzen 3000 series specs so far, because I was expecting leaked specs(yes I'm a fool for believing leaks from known AMD fanboys) or at least similar..

According to leaked specs, we were suppose to get four CPU's(8 cores, 12 cores & 2x 16 cores) running @ 4.7-5.1GHz. and 3 of em(12 cores & 2x 16 cores) were clearly better than i9 9900K at lower price.
View attachment 164516
But we didn't get ANY 16 core CPUs, and on average Ryzen 3000 series are 400MHz slower than leaked specs.
We only get one 12 core CPU @ 4.6 GHz. Rest are 6-8 core CPUs @ 4.2-4.5GHz.. sigh

AMD Leaked Spec Formula = X(likely spec) + 400MHz + 4 cores
Click to expand...
Then tbh it’s really your own fault to be disappointed. The leaker even said to take it with a grain of salt. Sounds you took it as the truth.
 
  • Like
Reactions: N4CR
like this
gintama6888 said:
AMD Leaked Spec Formula = X(likely spec) + 400MHz + 4 cores
Click to expand...

And that is what a rumor mill does.
be patient and wait for real life benchmarks.
there is a rumor of up to 15% better IPC, which would cause all of your complains moot.
 
I see where you are coming from, butt am not woried for several reasons.

1) I took less than a grain of salt with the 16C leaks. I don't see how a 16C/32T would have even been remotely possible with a 4.3 Base Clock on first round of 7nm at 135W TDP, Let alone 2nd or 3rd round. IPC improvements will always be more important than clock speed.

2) I don't see anything wrong with AMD barely beating Intel with their 7nm compared to Intels 14++. AMD can now make some money and they don't have to worry about extreme high yields. AMD is better off waiting for Intel's next move, rather than throwing as much performance out as possible and giving up all their cards.

2a) On a side note, this is probably why AMD isn't wasting their money on 16C chips if yields are lower than expected. There would be no competition at the moment so why waste on 16C when they will get better yields on 12C. AMD can release 16C when yields pick up in several months

3) AMD has at least another year or two before Intel jumps to 7nm. It's about what you need right now and soon upcoming.

4th and Final note) AMD made the 8 Core mainstream (talking about current technology) where if Ryzen was never released, 6 core CPU's would have been currently mainstream. And current CPU prices? Beautiful. You now have choices.

It could ALWAYS be worse. At least this situation is not like the Graphics GPU Industry's situation for many reasons. For starters, My 1080 Ti on 16nm fab which was released a little over 3 years ago goes just about head to head with a 7nm Vega VII.
 
While many will say bulldozer was bad, I was quite happy with my 8320@5.1ghz. I used to stream dota 2 on twitch, while gaming and it did this task amazingly well. I did cpu encoding, so quality was GREAT! Something I couldn't do on a Intel 2500k@ 4.8 ghz. Was also kinda nice hosting a game server on 4 cores, and using the other 4 for gaming. Moving on

Ryzen is a different beast. per core performance moved past all but the latest intel uarchs. Zen 2 or Ryzen 3xxx series improves in basically every area. I'm not really a fan of any company anymore I just buy what I can afford and What I think best suits my needs. I think Zen 2 will be the first time without a doubt that AMD will be on top. Some people reference the Athlon days, but that was more of trade of battles then anything else. In the tech world all companies make blunders. Its a hard game to play designing a product 4-5 years before actual release. You can be aggressive (bulldozer), or can be passive (haswell) and hope your forward thinking will pay off.

What it boils down to me is pricing. I'd love to build a 5k computer, but even with my income spending in excess of 1000$ is a major purchase (try running that by the wifey) So I usually upgrade a piece or two at a time. Biggest thing esp in the cpu side of things lately I had no desire to upgrade my cpu. Intel was rocking 3-4% ipc lifts per generation if that. MY current r7-1700@ 4ghz is all I really need, but when AMD says 15% ipc gain and few hundred more MHZ clock frequency that makes me want to upgrade. Oh it will drop in my current board? its a no brainer.

Just look at all the love the x570 boards are getting, this is unheard of for AMD. So yes I believe 100% they hit this one outta the ball park and we are in for a treat.
 
Last edited:
I just figured the plus for AMD is that the 12 core chip is using 2x 6 core chiplets which means they can use any faulty 8 core chiplets off the production line (with up to 2 bad cores = 6 good cores) and still have a saleable product. If they wanted to push a 16 core, it would take 2 perfect 8 core chiplets - which if sold as a single 3800x could net them $399 retail each vs. how much they could possibly make from the sale of a niche product like a 16 core Ryzen 9 x3999x (or whatever they would call it).
 
bigblueshock said:
On a side note, this is probably why AMD isn't wasting their money on 16C chips if yields are lower than expected.
Click to expand...

Every single tech youtuber (that I've seen) at computex has claimed that there's a 16-core being unofficially demoed.
If it's such an open secret, it's probably also just about ready for launch. Dunno exactly what they're waiting for; could be motherboards, could be yields, could be they're sandbagging HARD and want to have a "But wait, there's more!" moment and really blow the doors off.
My own guess is the 8 and 12 cores will be best for gaming, and they really want to get rid of the "...except for gaming..." line from every single review about Ryzen 1 & 2.
16 cores will obviously beat 8 for productivity, but the headlines will come from doing the improbable.
 
dvsman said:
If they wanted to push a 16 core, it would take 2 perfect 8 core chiplets - which if sold as a single 3800x could net them $399 retail each vs. how much they could possibly make from the sale of a niche product like a 16 core Ryzen 9 x3999x
Click to expand...
I think the 16 core would be cheaper to produce than two 8 cores, because the I/O die needs to exist only once.

If there was a shortage of perfect 8-core chips, AMD could have put two 4-cores in the 3700X/3800X instead, yet they chose not to.
otg said:
Dunno exactly what they're waiting for; could be motherboards, could be yields, could be they're sandbagging
Click to expand...
My bet is on motherboards. Give board partners more time to figure out which existing mobos meet the power requirements and perform validation.
 
Intel has been on this revision of 14nm+++++++++++++++++++++. Okay may be not that many plusses but you know what I mean. And Intel's 14nm is like 10nm equivalent of samsung and tsmc.

TSMC 7nm is still fresh and AMD has the first product on it. Obviously, you will get better and better as time goes on. This is AMD's first attempt at 7nm, and they have been stuck with GF process that was fine but never really was designed for higher clocks, it was more for efficiency.

AMD has the right mindset. Get as much per core IPC as they can out of each iteration and frequency will come over time with process improvements.
 
gintama6888 said:
Personally I'm at disappointment at Ryzen 3000 series specs so far, because I was expecting leaked specs(yes I'm a fool for believing leaks from known AMD fanboys) or at least similar..

According to leaked specs, we were suppose to get four CPU's(8 cores, 12 cores & 2x 16 cores) running @ 4.7-5.1GHz. and 3 of em(12 cores & 2x 16 cores) were clearly better than i9 9900K at lower price.
View attachment 164516
But we didn't get ANY 16 core CPUs, and on average Ryzen 3000 series are 400MHz slower than leaked specs.
We only get one 12 core CPU @ 4.6 GHz. Rest are 6-8 core CPUs @ 4.2-4.5GHz.. sigh

AMD Leaked Spec Formula = X(likely spec) + 400MHz + 4 cores
Click to expand...

Really? Did you really believe the leaks you were going to get a frickin 16 core chip running at 5.1ghz boost? Common man if you can't keep your expectations in check that is your fault. Leaks are not facts. That's why they say take them with a grain of salt, looks like you didn't lol.
 
otg said:
Dunno exactly what they're waiting for; could be motherboards, could be yields, could be they're sandbagging HARD and want to have a "But wait, there's more!" moment and really blow the doors off.
Click to expand...

Many people that will buy the 12 core, are exactly the kind of people that will upgrade to the 16 core later. Why sell once, when they can sell twice?
 
Snowdog said:
Many people that will buy the 12 core, are exactly the kind of people that will upgrade to the 16 core later. Why sell once, when they can sell twice?
Click to expand...

This. I think they will probably release 16 core down the road when they can bring the efficiency even higher and squeeze the best clocks from it, to bring it out under best case scenario. They basically have no need for it right now, beating intel on core and efficiency as we speak.
 
drescherjm said:
They paid way way too much for the ATI acquisition.
Click to expand...
At the time? Ya probably. In the long run? Nope.
That IP is already worth so much more. They own the consoles. They have the best APU's. And they still compete in discrete. Going forward CPU and GPU tech is needed.
There are only so many patents and Nvidia and AMD own most. Intel is having to spend big $$ to re-enter a space they long abandoned.
Imagine ATi's worth in 25 years if they were still in the game?
Nvidia is just hoping that ARM will allow them to hang on as a major player.
 
NKD said:
This. I think they will probably release 16 core down the road when they can bring the efficiency even higher and squeeze the best clocks from it, to bring it out under best case scenario. They basically have no need for it right now, beating intel on core and efficiency as we speak.
Click to expand...

It may also be a strategy to wait until Threadripper 2 is out so AMD doesnt crush thir current hottest selling HEDT processor 2950x.

Possibly they may wait to release the desktop 16 as a path for AM4 while the 32core 3950x or whatever name it will have is sold as a path for hedt. I'd do that if I were them. They dont want the reputation of shitting on thier hedt customers like Intel does.
 
Keep in mind that boost speeds are not XFR speeds. We may very well still see a 5ghz XFR speed on a single core.
 
I think AMD simply does not need a 16C to be ahead of Intel in the mainstream platform. Intel will not be able to release a faster CPU for a long time .
 
drescherjm said:
I think AMD simply does not need a 16C to be ahead of Intel in the mainstream platform. Intel will not be able to release a faster CPU for a long time .
Click to expand...

Intel: We are behind on cores, add two more...!

AMD: Release the Kraken, er, 16 core parts...!
 
Intel: We are behind on cores, add two more...!
Click to expand...

This strategy worked both times they tried it. Intel released a CPU with better performance than AMD had. However I don't see it working with their 10nm process.
 
drescherjm said:
ATI was on the verge of bankruptcy when AMD purchased them.
Click to expand...

yeah i think a lot of people don't realize just how doomed ATi really was when AMD bought the company.. not to mention how much more AMD had to spend just to build it back up afterward, damn near killed AMD and would of left us with just intel and nvidia. worse case if they hadn't of bought ATi we would been left with just nvidia for discrete gpu's.
 
  • Like
Reactions: blkt
like this
sirmonkey1985 said:
yeah i think a lot of people don't realize just how doomed ATi really was when AMD bought the company.. not to mention how much more AMD had to spend just to build it back up afterward, damn near killed AMD and would of left us with just intel and nvidia. worse case if they hadn't of bought ATi we would been left with just nvidia for discrete gpu's.
Click to expand...

Have some links to confirm that? My recollection is that they had a bad year, but still brought net cash to the deal, which you wouldn't see from a company on the verge of bankruptcy.
 
drescherjm said:
I think AMD simply does not need a 16C to be ahead of Intel in the mainstream platform. Intel will not be able to release a faster CPU for a long time .
Click to expand...

My thought on this is it's likely AMD is trying to get as much out of two launches as they can.

Probably were going to originally have a lineup closer to adoredtv list, but when they found several of those low yield products would only be besting the other higher yield products they had lower in the stack and Intel already in the dust they decided to hold off on those models and make as much off the 8 and 12 core models as they could before later dropping all of those products prices and releasing the real lineup with the 16 core r9 3950 and r93950X products, not to mention keeping more 8-core chiplets for the impending threadripper.

It's essentially the Intel playbook when they could charge a massive premium for anything with more than 4 cores.
 
Zarathustra[H] said:
Why is it AMD needs a 7nm process in order to be competitive with Intel's 14nm chips? By all accounts all else being equal, a 7nm chip should be crushing a 14nm chip. Smaller process size means less heat and power, which means you can crank things up more. The fact that AMD is only catching up to Intel's 14nm chips on per core performance on a 7nm process strongly suggests that all things are not equal.
Click to expand...

If this was 10 years ago this would maybe be true, but it's not. Just look at how well a 4770k (22nm) continues to hold up against Intel's latest 14nm++, and how Broadwell was not an upgrade from Haswell. Surely if process node mattered as much as you are claiming this would be utterly impossible?

So maybe Intel's 22nm->14nm was a fluke. Let's go back further. How about the i7-2700K (32nm) vs. i7-4770k (22nm). Again, huge process change. And yet, both are 4c/8t, 3.5ghz base, and 3.9ghz boost despite both being the top of Intel's consumer line. And the TDP difference is only 10w, so it's not like one was a huge power reduction, either. Overall performance difference around 10%.

The reality is process node barely matters anymore. It's not enabling significant jumps in clock speeds. It took 7 years, 2 major process node hops, and multiple refinements to go from 3.9ghz boost to 4.9ghz boost. It's not enabling much in the way of IPC gains, as that's not a "how many transistors can you fit" problem anymore. What it *is* doing is allowing cache sizes & core counts to increase, but that's about it. Single-thread performance largely hit a ceiling. This is it, this is all you're going to get without something major changing.

AMD's current architecture may be better than the highly flawed Bulldozer architecture, but the above means that the architecture itself is still WELL behind Intel's, and that they are competitive solely because they currently have the process advantage.
Click to expand...

IPC is the better indicator of architecture and Zen's IPC was only around 10% behind Skylake's. *IF* AMD delivers on that 15% IPC increase that would put AMD's architecture *ahead* of Intel's by ~5%. But either way it's clearly nonsense to claim AMD's is "well behind" Intel's and that they only are competitive because of a process advantage. They are already pretty close on IPC at a process *disadvantage*

But as elaborated on above process was really more about clock speed than architecture, and these days it's really just core counts with incremental clock speed increases.
 
gintama6888 said:
Personally I'm at disappointment at Ryzen 3000 series specs so far, because I was expecting leaked specs(yes I'm a fool for believing leaks from known AMD fanboys) or at least similar..

According to leaked specs, we were suppose to get four CPU's(8 cores, 12 cores & 2x 16 cores) running @ 4.7-5.1GHz. and 3 of em(12 cores & 2x 16 cores) were clearly better than i9 9900K at lower price.
View attachment 164516
But we didn't get ANY 16 core CPUs, and on average Ryzen 3000 series are 400MHz slower than leaked specs.
We only get one 12 core CPU @ 4.6 GHz. Rest are 6-8 core CPUs @ 4.2-4.5GHz.. sigh

AMD Leaked Spec Formula = X(likely spec) + 400MHz + 4 cores
Click to expand...

A 16 core CPU will be released, and has appeared in some early benchmarks. It just doesn't look like it's part of the July 7th release. So don't be disappointed there. IIRC, even the leakers suggested that the 16 core part might not be part of the initial release.

Your disappointment is a little misguided, in the sense that though clockspeeds were lower than the leaks, most of us expected that the IPC increase would NOT be 15%, despite some early "leaks" to that effect. There was a lot of skepticism on that. So we got more IPC than was anticipated, but lower clockspeed than was leaked. That's a wash.

I never really expected 5.0GHz anyway. My expectations were max boost ~4.7GHz as documented here, and we got 4.6. Pretty close!
 
You must log in or register to reply here.
Back
Top