Wednesday, December 13th 2017

AMD Confirms 2nd Generation Ryzen Processors to Debut in Q1-2018

At a press event, AMD confirmed that its 2nd generation Ryzen desktop processors will debut in Q1-2018 (before April). It also clarified that "2nd Generation" does not equal "Zen2" (a micro-architecture that succeeds "Zen"). 2nd Generation Ryzen processors are based on two silicons, the 12 nm "Pinnacle Ridge," which is a GPU-devoid silicon with up to eight CPU cores; and "Raven Ridge," which is an APU combining up to 4 CPU cores with an iGPU based on the "Vega" graphics architecture. The core CPU micro-architecture is still "Zen." The "Pinnacle Ridge" silicon takes advantage of the optical shrink to 12 nm to increase clock speeds, with minimal impact on power-draw.

AMD is also launching a new generation of chipset, under the AMD 400-series. There's not much known about these chipsets. Hopefully they feature PCIe gen 3.0 general purpose lanes. The second-generation Ryzen processors and APUs will carry the 2000-series model numbering, with clear differentiation between chips with iGPU and those without. Both product lines will work on socket AM4 motherboards, including existing ones based on AMD 300-series chipset (requiring a BIOS update). AMD is reserving "Zen2," the IPC-increasing successor of "Zen" for 2019. The "Mattise" silicon will drive the multi-core CPU product-line, while the "Picasso" silicon will drive the APU line. Both these chips will run on existing AM4 motherboards, as AMD plans to keep AM4 as its mainstream-desktop socket till 2020.
Source: WCCFTech
Add your own comment

101 Comments on AMD Confirms 2nd Generation Ryzen Processors to Debut in Q1-2018

#51
notb
"looncraz said:
Let's go from Sandy Bridge all of the way to Skylake. At 3GHz, fixed, which is where I do all of my IPC testing (as does Anandtech).
You miss the point. I don't care about IPC!
I'm only interested in how fast a CPU is (relative to power consumption, heat and price). IPC is just a weird number that hardly tells you anything about the final product.

Intel might have done just 25% in IPC, but they offered more in performance. They repeatedly refine their architecture for higher clocks and less power consumption.
AMD stressed *IPC gain* during Ryzen launch (and you clearly took the bait).
Because Ryzen's clocks are lower than those of last AM3 CPUs, performance gain is smaller than IPC increase.

I don't know if you've seen the big discussion about Ryzen launch.
Among other things (like "not impressed") I've repeatedly mentioned that almost no review compared Ryzen to Piledriver - despite the fact that Sandy Bridge was often present in these tests. I don't know why this happened. I've found it weird and smelling of conspiracy, but whatever.

But guess what. The brave Swedes did it. Scandinavians are pretty good at ignoring what other people tell them to do (not that I think anyone told TPU not to include FX in their reviews :p):
https://www.sweclockers.com/test/23613-amd-ryzen-5-1600x-och-5-1500x/4#content
So the best thing about this review is the graph below: multi-thread Cinebench R15 - AMD's favourite benchmark.
There are four 4C/8T processors: 2600K, FX-8350 (from 2011/2012) and 7700K, 1500X (from 2017).
Now look at it, think carefully and tell me once again that AMD's CPUs improved more than Intel's. I dare you. I double dare you.
Posted on Reply
#52
EarthDog
"looncraz said:
Let's be honest here: neither of us knows what Global Foundries did to achieve a 15% reduction in area. 14nm LPP is rather mature and dense as it is, so feature size changes would likely be the chosen route... by employing existing GF technology.



The IMC is a client of the IF. This means it needs to operate with discrete packet transactions. It's an absolute requirement that data is fed into a buffer before sending data (either to the memory sticks or to the requesting client) - otherwise you stand a good chance of flooding the IF with packets. It's a trade-off between bandwidth and latency. I've had to implement these things (in software) before - it's pretty simple... and VERY effective. But always with the downside of higher latency.



24%... if you include the outlier 69% boost. 19% without it.

And, yes, it includes ALL jumps.

POV-Ray is *almost* an outlier. IPC improvement is only 16.7% without it... from Sandy Bridge to Skylake. So the "fair" IPC improvement range is 16~20%.... with FOUR generations of products (well, five or six now, depending on how you count).
lol, cherry picking... 24% gaims from ivy to skylake. It would bw over 30 of sb was included.
Posted on Reply
#53
looncraz
"EarthDog said:
lol, cherry picking... 24% gaims from ivy to skylake. It would bw over 30 of sb was included.
Everything is relative to Sandy Bridge in that chart. I never cherry pick.

Look at the top of the image... where it says "Gains over Sandy Bridge"

"notb said:
You miss the point. I don't care about IPC!
I'm only interested in how fast a CPU is (relative to power consumption, heat and price). IPC is just a weird number that hardly tells you anything about the final product.
Intel might have done just 25% in IPC, but they offered more in performance. They repeatedly refine their architecture for higher clocks and less power consumption.
IPC is the core of what performance you have from a product. Sandy Bridge could hit 5GHz rather reliably. Skylake is only a slightly better overclocker. The fact that Intel gave it higher stock clocks only matters to people who never overclock.

Power consumption improvements have certainly been present - but at the same time Intel started using toothpaste between the CPU die and the IHS, making the CPUs run much hotter than they should - I care far more about the heat than the power.

"notb said:

AMD stressed *IPC gain* during Ryzen launch (and you clearly took the bait).
Because Ryzen's clocks are lower than those of last AM3 CPUs, performance gain is smaller than IPC increase.
AMD, like any company, will stress the figure that makes them look best. 8 cores, 16 threads, generally superior SMT scaling, large IPC jump, 8-cores with 4.1GHz XFR, unlocked multipliers on all SKUs, etc....

AMD went from 225W CPUs to 95W CPUs while showing a massive jump in performance.

"notb said:

I don't know if you've seen the big discussion about Ryzen launch.
Among other things (like "not impressed") I've repeatedly mentioned that almost no review compared Ryzen to Piledriver - despite the fact that Sandy Bridge was often present in these tests. I don't know why this happened. I've found it weird and smelling of conspiracy, but whatever.
There are so few reviews for that because so few reviewers have Bulldozer/Piledriver CPUs for comparison. It's not a conspiracy - I've done a direct comparison with Excavator - which I compared directly to Steamroller, which I had previously compared with Piledriver (FX-8350). Even Anandtech's database lacks any direct comparisons between Bulldozer/Piledriver and Ryzen. Benchmarks change after five years.

"notb said:


But guess what. The brave Swedes did it. Scandinavians are pretty good at ignoring what other people tell them to do (not that I think anyone told TPU not to include FX in their reviews :p):
https://www.sweclockers.com/test/23613-amd-ryzen-5-1600x-och-5-1500x/4#content
So the best thing about this review is the graph below: multi-thread Cinebench R15 - AMD's favourite benchmark.
There are four 4C/8T processors: 2600K, FX-8350 (from 2011/2012) and 7700K, 1500X (from 2017).
Now look at it, think carefully and tell me once again that AMD's CPUs improved more than Intel's. I dare you. I double dare you.

FX-8350 is an EIGHT core CPU - whether you like it or not. It has a strange arrangement, but it's still eight cores. So the jump is from 633 to 1625... 156.7% faster!!

PER THREAD, which eliminates the core arrangement oddness, and the SMT advantage of Ryzen, the jump is from 98 to 162... a healthy 65.3% faster... despite 1600X/1800X operating at lower frequencies (in real world usage, the turbo+XFR of Ryzen is pretty much DOA, so we're talking about 3.7~3.8 (avg) vs 4.1~4.2GHz).

Ryzen has lower performance today than Intel's best - whether you look at IPC, frequency, or both. I doubt Pinnacle Ridge will fully close that gap, no matter how narrow - Intel is well entrenched with software optimizations in every critical program - and AMD is partly to blame for that.

AMD was able to gain so much performance over its previous generations because its previous generations were crap - they were slower than their previous architectures (Excavator even has slightly lower IPC than the Phenom II in many cases!).

Posted on Reply
#54
Melvis
"Konceptz said:
If we can get 4.0-4.2ghz stable or higher while waiting for Ryzen 2 i'll be happy. I still on my FX-8350 waiting for Ryzen +
This is pretty much me, if AMD can bring out a 8C16T CPU with base clock speed of 4GHz, turbo up to whatever then I would be very happy and would jump on it from my ageing FX-8350. I think a slight IPC increase is possible, they only need 5-7% to be on par with Intel.
Posted on Reply
#55
Captain_Tom
"Prima.Vera said:
So basically just a shrink...pfff. Was expecting better tbh...
AMD has also hinted at slight tweaks to infinity fabric and the memory controller. I know this all sounds mundane, but I personally think it will end up being a bigger deal than people realize. After all consider how competitive AMD is with Intel right now: In reality the only thing Intel has over AMD is higher clock speeds. Otherwise AMD is more efficient and cheaper to manufacture. So then add these things:

1) 10% higher clockspeeds (So 4.5GHz 8 and 16 core cpu's!)
2) Faster infinity fabric (This has often yielded substantially better gaming performance with Zen)
3) Higher memory speed support, again this has been the only thing holding back the larger Zen dies.

We are talking about the possibility of 10-20% higher performance in games, and slight bumps in all other apps. That's all it would take to knock CoffeeLake down mere months after it came out, and more importantly 7nm Zen 2 would follow-up in about a year...
Posted on Reply
#56
notb
"Captain_Tom said:

1) 10% higher clockspeeds (So 4.5GHz 8 and 16 core cpu's!)
You're comparing to XFR, not to base clocks. You shouldn't. Base clocks might go up by 10%, but boost/XFR could become less relevant.
2) Faster infinity fabric (This has often yielded substantially better gaming performance with Zen)
There's really no physical reason why Infinity Fabric could become significantly faster. The issues IF has stem not from some issues with implementation, that could be easily fixed. The problem is the design itself. IF lets AMD build cheap and efficient MCMs, but it has some problems that won't be fixed before it's redesigned.
3) Higher memory speed support, again this has been the only thing holding back the larger Zen dies.
Unlikely, since RAM speed is so important for the IF.
We are talking about the possibility of 10-20% higher performance in games, and slight bumps in all other apps. That's all it would take to knock CoffeeLake down mere months after it came out, and more importantly 7nm Zen 2 would follow-up in about a year...
And Zen 3 next year and so on. And then Zen 4! OMG!

By the time Zen 2 comes out, Intel will release Ice Lake - their next gen CPU years in the making. Among other things, it should give us DDR5 and PCIe 4.0. It's really hard to imagine how big the leap in performance will be. At very least I'd expect current high-end desktop performance (so desktop Ryzen and Coffee Lake) in ultrabooks paired with seamless external drives and GPUs. So maybe by 2020 there will be really no point in having a desktop at all.
And remember, that AMD will still be using AM4 SoCs.

So the big suggestion: stop wasting time dreaming what might happen in 2 or 3 years. It really takes you nowhere (maybe apart from cheering yourself up with AMD's supposed superiority).
Posted on Reply
#57
looncraz
"notb said:
You're comparing to XFR, not to base clocks. You shouldn't. Base clocks might go up by 10%, but boost/XFR could become less relevant.
3.6GHz base clocks become 4GHz base clocks. 4.1GHz Turbo+XFR becomes 4.5GHz Turbo+XFR.

Assuming we only get 10% (I expect a touch more than that from the process alone).

"notb said:

There's really no physical reason why Infinity Fabric could become significantly faster. The issues IF has stem not from some issues with implementation, that could be easily fixed. The problem is the design itself. IF lets AMD build cheap and efficient MCMs, but it has some problems that won't be fixed before it's redesigned.
Quite the contrary! IF is a multi-bus first generation protocol - there are bound to be some bottlenecks here and there that can be improved upon. On top of that, delinking the IF frequency from the memory frequency might be possible (they might always need to be related, though, so a multiplier based on the RAM/IMC frequency seems most logical).
Posted on Reply
#58
notb
"looncraz said:
I never cherry pick.
Sure. ;-)
IPC is the core of what performance you have from a product.
It's not "core performance". It's "core effectiveness" at best. And I still don't care about it. It's not the performance that I'd be able to use when I buy the CPU.
Sandy Bridge could hit 5GHz rather reliably. Skylake is only a slightly better overclocker. The fact that Intel gave it higher stock clocks only matters to people who never overclock.
So for almost everyone. Hardly anyone did that even 10 years ago, when it still had a point. Today? Pff.
The last CPU I've overclocked was an Athlon 1700+.
Power consumption improvements have certainly been present - but at the same time Intel started using toothpaste between the CPU die and the IHS, making the CPUs run much hotter than they should - I care far more about the heat than the power.
The "toothpaste" has no impact on the heat (putting aside the fact that CPU that runs 30K hotter uses wastes maybe ~5% more power).
It has a huge impact on heat transfer, but again: if you don't OC, you don't care. Intel's CPUs (even with the mediocre bundled coolers) have so much temp headroom you don't have to care about it, ever.
AMD, like any company, will stress the figure that makes them look best.
And what do you think - as a user? You'd rather buy an older tech with 5% more performance or modern tech with 10% more IPC? In other words, does the fact that AMD design is superior at this point (assuming it's true) makes you feel better?
Also, I'm really waiting for the Zen+ marketing campaign. What will be the key selling point now? % of recycled paper used for the box? :-)
8 cores, 16 threads, generally superior SMT scaling, large IPC jump, 8-cores with 4.1GHz XFR, unlocked multipliers on all SKUs, etc....
And all these resulting in performance still behind Intel's revamped Haswell.
I've bolded the funniest part. Everything else is more or less relevant, but once again you're saying that Zen is good because the predecessor was rubbish.
AMD went from 225W CPUs to 95W CPUs while showing a massive jump in performance.
Again, totally irrelevant for current AMD vs Intel choice. But you're right: they've cut power consumption significantly. This is by far the most important thing Zen gave us, but for the most part it's more about the process improvement than the design itself. Bulldozer made on Samsung's 14nm could use half the power as well.
There are so few reviews for that because so few reviewers have Bulldozer/Piledriver CPUs for comparison. It's not a conspiracy - I've done a direct comparison with Excavator - which I compared directly to Steamroller, which I had previously compared with Piledriver (FX-8350). Even Anandtech's database lacks any direct comparisons between Bulldozer/Piledriver and Ryzen. Benchmarks change after five years.
That's the weirdest argument I've seen so far. Seriously, it's so hard to find a STILL AVAILABLE part for review? Give me a break.
And some reviewers clearly managed, so maybe they are simply better at their job?
FX-8350 is an EIGHT core CPU - whether you like it or not. It has a strange arrangement, but it's still eight cores. So the jump is from 633 to 1625... 156.7% faster!! PER THREAD, which eliminates the core arrangement oddness, and the SMT advantage of Ryzen, the jump is from 98 to 162... a healthy 65.3% faster... despite 1600X/1800X operating at lower frequencies (in real world usage, the turbo+XFR of Ryzen is pretty much DOA, so we're talking about 3.7~3.8 (avg) vs 4.1~4.2GHz).
Hardwarecanucks tested 1700X without SMT - and even compared to FX-8370.
http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/74880-amd-ryzen-7-1700x-review-testing-smt-4.html
So now we have 1104/643 => 72% gain. Very good, but, and it's a big one, 1700X launched for $400 - double what you'd had to pay for FX-8370. Even today it's $300, +50%.
That's why I compared to similarly priced 1500X, like a normal consumer would. And like you would as well. You can spend the whole Friday praising AMD's improved IPC, but when you buy a CPU on Saturday, you'll be exchanging money for performance - not fairy dust for IPC.
Posted on Reply
#59
xrror
Am I the only one that read "It also clarified that "2nd Generation" does not equal "Zen2"" to mean that there won't be a technical (vs. just marketing) refresh for desktop parts yet?

This reads to me as AMD "gently hinting" that it is not going to be releasing an new stepping/revision of Desktop Ryzen when the new mobile part hits.

I hope I'm wrong.
Posted on Reply
#60
R-T-B
"notb said:
And some reviewers clearly managed, so maybe they are simply better at their job?
Or forking out for it out of pocket, something not every reviewer does (and understandably!)
Posted on Reply
#61
Prince Valiant
"R-T-B said:
Or forking out for it out of pocket, something not every reviewer does (and understandably!)
There's also the time commitment required for another round of benching.
Posted on Reply
#62
looncraz
"notb said:

It's not "core performance". It's "core effectiveness" at best. And I still don't care about it. It's not the performance that I'd be able to use when I buy the CPU.
Single thread performance is IPC * frequency. If you don't know both values, you can't understand why the performance of a CPU is the way it is - and you don't know which way to invest going forward.

"notb said:

So for almost everyone. Hardly anyone did that even 10 years ago, when it still had a point. Today? Pff.
The last CPU I've overclocked was an Athlon 1700+.
Ryzen is an enthusiast / tweaker product. Overclocking is a major value-add feature. Intel charges more for that privilege.

"notb said:

The "toothpaste" has no impact on the heat (putting aside the fact that CPU that runs 30K hotter uses wastes maybe ~5% more power).
It has a huge impact on heat transfer, but again: if you don't OC, you don't care. Intel's CPUs (even with the mediocre bundled coolers) have so much temp headroom you don't have to care about it, ever.
I know absurdly too much about thermal behavior... but hotter is still worse - and the hotter air is more noticeable, regardless of how much actual energy is being dissipated. The whole room may not warm up as much, but the air closest to you will be warmer.

"notb said:

And what do you think - as a user? You'd rather buy an older tech with 5% more performance or modern tech with 10% more IPC? In other words, does the fact that AMD design is superior at this point (assuming it's true) makes you feel better?
I buy the product that makes the most sense for my needs. I was going to buy a $1,000 6900K if Ryzen didn't pan out. I ended up only spending $370 or so (1700X).

Now, Pinnacle Ridge should bring a ~15% boost merely a year after my upgrade. Depending on price and performance, I may be able to just plop in a new CPU for more performance... or not... and enjoy the luxury knowing that another year, yet again, I will probably have yet another CPU upgrade option available to me.

"notb said:

Also, I'm really waiting for the Zen+ marketing campaign. What will be the key selling point now? % of recycled paper used for the box? :)
Ryzen still maintains many of its advantages over Intel. Value, chief among them. Pinnacle Ridge will build upon its strengths and narrow the gap with Intel even more. The eight core CPUs should be untouchable once more until Intel released a mainstream 8-core CPU.

"notb said:

And all these resulting in performance still behind Intel's revamped Haswell.
I've bolded the funniest part. Everything else is more or less relevant, but once again you're saying that Zen is good because the predecessor was rubbish.
Zen isn't good because its predecessor is rubbish - it's good because it effectively caught up to Intel in a single bound. Ryzen came out against the likes of the 6900k. Pinnacle Ridge comes out against Coffee Lake.

"notb said:

Again, totally irrelevant for current AMD vs Intel choice. But you're right: they've cut power consumption significantly. This is by far the most important thing Zen gave us, but for the most part it's more about the process improvement than the design itself. Bulldozer made on Samsung's 14nm could use half the power as well.
The process improvement was far less important than you might think. Zen cores are INSANELY efficient. The cores, themselves, use high single digit watts, stock - HALF the power of Intel cores. Ryzen's SoC, however, has some power efficiency issues. The IMC can use 20W by itself... AT IDLE! I can cut my SoC idle power in half by reducing my memory clocks. Hopefully they address that.

My 1700X, at 4GHz 1.375V uses 12~14W per core under full load. The rest of the SoC pulls about 30~40W under full load (I'm currently rendering 4.5 hours of video, so I just looked ;-)).

"notb said:

That's the weirdest argument I've seen so far. Seriously, it's so hard to find a STILL AVAILABLE part for review? Give me a break.
And some reviewers clearly managed, so maybe they are simply better at their job?
Very few reviewers are going to both to get together all of the hardware for a dated platform few care about just to test it against a new platform. Some have - and the results are very impressive.

"notb said:

Hardwarecanucks tested 1700X without SMT - and even compared to FX-8370.
http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/74880-amd-ryzen-7-1700x-review-testing-smt-4.html
So now we have 1104/643 => 72% gain. Very good, but, and it's a big one, 1700X launched for $400 - double what you'd had to pay for FX-8370. Even today it's $300, +50%.
That's why I compared to similarly priced 1500X, like a normal consumer would. And like you would as well. You can spend the whole Friday praising AMD's improved IPC, but when you buy a CPU on Saturday, you'll be exchanging money for performance - not fairy dust for IPC.
The 1700X was competing with Intel's 6900K - not Piledriver. Intel was charging $1,000 for that level of performance. $400 was a BARGAIN!
Posted on Reply
#63
notb
"R-T-B said:
Or forking out for it out of pocket, something not every reviewer does (and understandably!)
I'm not sure what are the terms with reviewing PC stuff. Do well-known review sites get to keep CPUs? It's not a huge cost for AMD nor Intel.
If not, there are other means of obtaining one - asking around, for example. Quite a lot of people on this forum have (and even still use) Piledrivers.
And they could always ask AMD ("Could you send us an FX? We want to show how great this new stuff really is!")

In the end we have the option of simply buying the missing stuff. 8350 + mobo is what... $150? I don't think this is a fortune in US or Western Europe.
Having a business results in costs. Doing things properly increases the costs even further. That's how things works, don't they?
"Prince Valiant said:
There's also the time commitment required for another round of benching.
OK. So let's say some reviewers are more committed to their job. This satisfies me as well. ;-)
"looncraz said:
Single thread performance is IPC * frequency. If you don't know both values, you can't understand why the performance of a CPU is the way it is - and you don't know which way to invest going forward.
No. At best single thread performance is proportional to IPC*frequency.
Ryzen is an enthusiast / tweaker product. Overclocking is a major value-add feature. Intel charges more for that privilege.
Now this is an interesting statement - one I didn't expect to see, but gladly welcome. :)
I know absurdly too much about thermal behavior... but hotter is still worse - and the hotter air is more noticeable, regardless of how much actual energy is being dissipated. The whole room may not warm up as much, but the air closest to you will be warmer.
Irrelevant. CPU temperature has no impact on your room's temperature, assuming the heat created is the same. Simply don't touch the CPU.
And honestly, if you're feeling intense heat from your PC while using it, you should really think about the desk setup. It's not healthy.
Now, Pinnacle Ridge should bring a ~15% boost merely a year after my upgrade. Depending on price and performance, I may be able to just plop in a new CPU for more performance... or not... and enjoy the luxury knowing that another year, yet again, I will probably have yet another CPU upgrade option available to me.
So you're in a tiny group of people who actually consider replacing a CPU every year. I mean: it's marginal even in this community, let alone globally.
You are a tweaker, you see Ryzen as a tweaker product. It's a perfect match. I'm not against that.
What I am against is criticizing Intel for providing a product cycle matching needs of a vast majority of population.
The 1700X was competing with Intel's 6900K - not Piledriver. Intel was charging $1,000 for that level of performance. $400 was a BARGAIN!
So why did we see all those 3-5 year old Intel CPUs in the reviews?
Is Ryzen competing with 2500K and 4770K?

Posted on Reply
#64
looncraz
"notb said:
I'm not sure what are the terms with reviewing PC stuff. Do well-known review sites get to keep CPUs? It's not a huge cost for AMD nor Intel.
If not, there are other means of obtaining one - asking around, for example. Quite a lot of people on this forum have (and even still use) Piledrivers.
And they could always ask AMD ("Could you send us an FX? We want to show how great this new stuff really is!")
Why would a reviewer go through all of that extra work? Besides, there ARE reviews that compare them. And, as you've so delightfully brought up repeatedly, Ryzen must be compared to the competition. Piledriver/Bulldozer aren't even remotely close to being competitive with Ryzen.

I would happly take a Ryzen 5 1400 over an FX-8370. That's a ~15% or so better comparison than the 2600k and the FX-8370 was back in the day.

"notb said:

No. At best single thread performance is proportional to IPC*frequency.
Fair enough, IPC is a complex number which varies by program. But the overall average IPC of a processor with a wide array of applications is nearly all you need to know to get a general idea of processor single-threaded performance at a given frequency. SMT, core count, bandwidth, and so on all play their parts in overall system performance.

"notb said:

Irrelevant. CPU temperature has no impact on your room's temperature, assuming the heat created is the same. Simply don't touch the CPU.
And honestly, if you're feeling intense heat from your PC while using it, you should really think about the desk setup. It's not healthy.
A 65W 95C heat source will eject warmer air than a 125W 50C heat source. Over time, the 125W heat source will do a better job of warming a LARGE amount of air, but that 95C heat source will transfer energy more efficiently - causing more localized heating. Meaning the inside of the case is hotter, the air around the computer is hotter, and so on. Transient versus equilibrium, if you will.

"notb said:

So you're in a tiny group of people who actually consider replacing a CPU every year. I mean: it's marginal even in this community, let alone globally.
You are a tweaker, you see Ryzen as a tweaker product. It's a perfect match. I'm not against that.
I am, yes, and so are many others. If not every year, then every few years - as progress permits. When I upgrade, my parts are sold down the chain - I usually upgrade at little to no cost to myself.

"notb said:

What I am against is criticizing Intel for providing a product cycle matching needs of a vast majority of population.
Intel has provided practically no benefit from one generation to the next since Sandy Bridge. Prior to that, 10~15% IPC and a frequency bump was pretty standard.

Sandy Bridge? 10~15% IPC bump, 400~800MHz better overclocking (5GHz pretty common) - the last big jump.
Ivy Bridge? 3~5% performance improvement - worse overclocking. Higher stock clocks just to make up for their failure.
Haswell? A decent IPC jump for some programs, decent overclocking. Higher stock clocks just to make up for their failure.
Skylake? 5~7% IPC boost, sometimes. Slightly better overclocking. Higher stock clocks just to make up for their failure. (5GHz overclock pretty common... still)
Coffee Lake? Two more cores, no other real improvements... maybe slightly better overclocking. (5GHz still the benchmark OC frequency).

Four cores. Eight threads. Every single generation, until Coffee Lake, for the mainstream.

Intel should have had mainstream six core CPUs with Haswell, in the very least. That's my biggest personal gripe - aside from Intel milking overclockers with 'k' edition chips which DISABLED features (TSX, virtualization, etc...).

"notb said:

So why did we see all those 3-5 year old Intel CPUs in the reviews?
Is Ryzen competing with 2500K and 4770K?
Because Intel has remained effectively stagnant for people who overclock. If you had a 4.5GHz+ Sandy Bridge CPU Intel offered NOTHING worth buying without spending an absolute fortune. A five year dry spell is NOT good.

If Intel had added 15% IPC each generation and pushed frequency and efficiency just enough to stay at Sandy Bridge levels, then comparisons would only need to be with Intel's then-current Skylake and whatever generation of CPU with which Ryzen compared favorably.

Ryzen competes only in the new build market. But you have to convince people that they need/want a new build. People with older AMD systems **KNOW** that their systems are/were FAR behind - they don't/didn't need a reminder. So comparing Ryzen with chips available for purchase on the new and used market is what makes/made sense.

"notb said:


Now who's cherry picking?

Nobody ever said Ryzen was better than Intel across the board. Different tools for different jobs.

7700k vs 1800X - 1800X can be as much as 75% faster... or about 35% slower, but usually being very close - or WAY ahead.... and generally more efficient.
6900k vs 1800X - 1800X can be as much as 13% faster... or about 25% slower, but usually being very close... and HALF the price... and more efficient.

(Both of the above % ignore outliers, such as Ryzen's other-worldy 7-zip performance versus 7700k or Rocket League results for the 6900k).

An eight core CPU should be compared with another eight core CPU, IMHO. The 1800X makes a mockery of the 6900K - not in raw performance, but in pure, simle, value for the same, or similar, performance (outside of games - which has been getting addressed more and more since these release benchmarks).

For someone considering upgrading or buying a new system, AMD finally makes sense to consider. Not to mention that we're only looking at the high end mainstream here - most will want to look at the Ryzen 5 CPUs - where fast and efficient 6 core CPUs with SMT can be had for cheap without any of Intel's artificial restrictions or bad business ethics.
Posted on Reply
#65
notb
"looncraz said:
Why would a reviewer go through all of that extra work?
Because it's their job!
Besides, there ARE reviews that compare them.
Which only shows that it is possible and all these arguments about FX rarity are fabricated.
And, as you've so delightfully brought up repeatedly, Ryzen must be compared to the competition. Piledriver/Bulldozer aren't even remotely close to being competitive with Ryzen.
But aren't current FX owners curious how much they would gain by moving to Zen? Maybe not so much? Maybe a Ryzen 3, that costs as much as they paid for their FX few years ago, isn't much faster?
A 65W 95C heat source will eject warmer air than a 125W 50C heat source. Over time, the 125W heat source will do a better job of warming a LARGE amount of air, but that 95C heat source will transfer energy more efficiently - causing more localized heating. Meaning the inside of the case is hotter, the air around the computer is hotter, and so on. Transient versus equilibrium, if you will.
Still no. And now you've moved from equilibrium thermodynamics to dynamical systems. But earlier you were talking about total heat and now you're slowly losing that idea.
Once again: we have a CPU that produces THE SAME AMOUNT OF HEAT at 50*C and 95*C. And you've just silently ignored the constant heat property. You say: it's hotter, so it'll make air around hotter. And this is true. But if it was constantly making air around it hotter than the 50*C one, it would produce more heat. So there's only one explanation: it's making the air hotter, but the amount of heated air is smaller. Only this way you can guarantee conservation of energy.

If you put your PC in a virtual sphere, the same amount of heat will be transferred though its surface - no matter what happens inside.
In other words: equilibrium thermodynamics works and you're just thinking way to much.
Intel has provided practically no benefit from one generation to the next since Sandy Bridge. Prior to that, 10~15% IPC and a frequency bump was pretty standard.
Intel is not concentrating on IPC, nor are its clients.
People care about other things: like will that ultrabook fit in my purse? Yes, many Intel-powered ultrabooks have been compatibile with women's purses for years. IPC what? Is that a new fragrance?
Intel should have had mainstream six core CPUs with Haswell, in the very least. That's my biggest personal gripe - aside from Intel milking overclockers with 'k' edition chips which DISABLED features (TSX, virtualization, etc...).
Because Intel has remained effectively stagnant for people who overclock.
Intel has more important things to care about than overclockers. Seriously. If you like manual OC (so not the automatic kind that manufacturers include in the design), you should be happy it's still possible. Just live with it.
Overclocking importance is constantly shrinking - at some point it will be as marginal as tuning cars.
Ryzen competes only in the new build market. But you have to convince people that they need/want a new build. People with older AMD systems **KNOW** that their systems are/were FAR behind - they don't/didn't need a reminder. So comparing Ryzen with chips available for purchase on the new and used market is what makes/made sense.
Piledriver is still available. 2500K is not. Why do I see 2500K in a review, not an FX-8350? This has been my exact question for months. Can you answer it or not?[/quote]
Posted on Reply
#66
TheLaughingMan
Money says the APUs come out first. Then Gen2 or the desktop parts late 3rd quarter.
Posted on Reply
#67
Prince Valiant
"notb said:
Piledriver is still available. 2500K is not. Why do I see 2500K in a review, not an FX-8350? This has been my exact question for months. Can you answer it or not?
Ask the reviewers you're so dissatisfied with.
Posted on Reply
#68
looncraz
"notb said:
Because it's their job!
Which only shows that it is possible and all these arguments about FX rarity are fabricated.
But aren't current FX owners curious how much they would gain by moving to Zen? Maybe not so much? Maybe a Ryzen 3, that costs as much as they paid for their FX few years ago, isn't much faster?
I can only tell you why I, and the many reviewers who have already said why, didn't do it.

FX owners already know/knew where they stand/stood.

"notb said:

Still no. And now you've moved from equilibrium thermodynamics to dynamical systems. But earlier you were talking about total heat and now you're slowly losing that idea.
Once again: we have a CPU that produces THE SAME AMOUNT OF HEAT at 50*C and 95*C. And you've just silently ignored the constant heat property. You say: it's hotter, so it'll make air around hotter. And this is true. But if it was constantly making air around it hotter than the 50*C one, it would produce more heat. So there's only one explanation: it's making the air hotter, but the amount of heated air is smaller. Only this way you can guarantee conservation of energy.

If you put your PC in a virtual sphere, the same amount of heat will be transferred though its surface - no matter what happens inside.
In other words: equilibrium thermodynamics works and you're just thinking way to much.
I have exclusively talked about transient thermals. Equilibrium thermals will always favor the higher source. Transients favor the higher temperature (to the fourth power!). Think of a laptop - do you want your laptop base to run at 80C or would you prefer it to be 50C? Do you care, at that point, how many joules of heat are being dissipated per second?

Computers operate in 'peaks' not generally sustained loads. A CPU that rapidly warms to 75~90C will dump its energy more quickly than one that warms only to 50C. It will warm the surrounding air / materials faster, and you will notice it.

But, yes, if everything is kept equal, except heat output, the higher heat output will always produce greater energy/heat transfer... EVENTUALLY. The air in the core of that sphere of yours would be much hotter with a 65W heat source at 95C than a 65W heat source at 50C. The outside of the sphere would be the same.. if averaged over time. But you will feel the heat faster with the 95C heat source (Stefan-Boltzmann Law).

"notb said:

Intel is not concentrating on IPC, nor are its clients.
People care about other things: like will that ultrabook fit in my purse? Yes, many Intel-powered ultrabooks have been compatibile with women's purses for years. IPC what? Is that a new fragrance?
Yes, but I - and nearly everyone here - don't care about those people. We care about enthusiasts, modders, DiYers, ourselves, and the state of the technical markets.

Those other people, though, DO care about performance metrics of their own. Mostly fiscal performance. They want an affordable product that is faster than what they bought last time. Laptops are commodity mobile devices - and we're talking about Ryzen desktop devices.

"notb said:

Intel has more important things to care about than overclockers. Seriously. If you like manual OC (so not the automatic kind that manufacturers include in the design), you should be happy it's still possible. Just live with it.
Overclocking importance is constantly shrinking - at some point it will be as marginal as tuning cars.
Then why does Intel go out of their way to court overclockers? And then go out of their way to HARM them by artificially removing features from 'K' series products?

Intel could just as easily have let multipliers unlocked, which is EASIER, and left overclockers alone. Instead, they decided to milk them.

"notb said:

Piledriver is still available. 2500K is not. Why do I see 2500K in a review, not an FX-8350? This has been my exact question for months. Can you answer it or not?
It's quite simple, really. I still have access to Sandy Bridge and later CPUs. I have access to Excavator. I do NOT have access to Bulldozer or Piledriver. This is pretty normal - Intel systems have had ample performance and overclocking since Sandy Bridge, so those systems are abundant and still in play. Excavator is the product AMD used as a point of comparison for Ryzen - so making sure you have access is important.

The few thousand people with Piledriver 8-core CPUs ALREADY know where they stand. They don't need to be told. They know that a STOCK 2700K outperforms their CPUs across the board. A product that comes out and whips the 2700k is OBVIOUSLY that much faster than what they have.
Posted on Reply
#70
Gmr_Chick
"EarthDog said:
Get a room...
Are you kidding, Dog? This is so fun to watch :D

My money is on looncraz :peace:
Posted on Reply
#71
Vayra86
"looncraz said:

Intel has provided practically no benefit from one generation to the next since Sandy Bridge. Prior to that, 10~15% IPC and a frequency bump was pretty standard.

Sandy Bridge? 10~15% IPC bump, 400~800MHz better overclocking (5GHz pretty common) - the last big jump.
Ivy Bridge? 3~5% performance improvement - worse overclocking. Higher stock clocks just to make up for their failure.
Haswell? A decent IPC jump for some programs, decent overclocking. Higher stock clocks just to make up for their failure.
Skylake? 5~7% IPC boost, sometimes. Slightly better overclocking. Higher stock clocks just to make up for their failure. (5GHz overclock pretty common... still)
Coffee Lake? Two more cores, no other real improvements... maybe slightly better overclocking. (5GHz still the benchmark OC frequency).

Four cores. Eight threads. Every single generation, until Coffee Lake, for the mainstream.

Intel should have had mainstream six core CPUs with Haswell, in the very least. That's my biggest personal gripe - aside from Intel milking overclockers with 'k' edition chips which DISABLED features (TSX, virtualization, etc...).
See this is where it all gets derailed in your train of thought. I think you have good insight on the matter, but you forget to compare products within their own timeframe and above all, their own reality. PC performance relies on so much more than even just hard numbers like IPC, clocks, cores etc.

Above everything else it relies on the applications you have for it. In the end it comes down to "is there a market for this". Simple as that. Back in 2010-11 there was no market for a 5 Ghz beast called Sandy - proven by the millions who still run this CPU and enjoy smooth computing for everything they throw at it, and the extremely LOW % of people who actually run this CPU at 5.0.

The whole period in between, from IB > Kaby Lake? That was 100% AMD failing to deliver - and all this time, they too had their 5 Ghz CPU on the shelf, except it didn't sell and it wasn't good. Gaming performance is a great example, again, of why Coffee Lake TODAY is such a relevant CPU: AMD delivered Ryzen with the higher and much sought after core counts, but they forgot that performance is not just IPC and cores, but also clocks. While ever since SB, Intel has that triplet under very tight, ever tighter control. Its that degree of control precisely where AMD drops the ball. Another reason CFL is so relevant today and would not have been six years ago, is because today we see actual multithreading and applications that are actually eating all those threads at the highest clocks, such as high refresh gaming.

You talk down alot on overclocking on Intel (have to pay for it, still stuck on 5 Ghz, Satan's toothpaste under IHS), but the reality is that its competitor clocks 20% lower even with the best intentions and greatest hardware. That's 20% lower than what was 'the norm' 7 years ago (!!!) but at least you've got soldered heatspreaders. GREAT.

Stop overselling AMD, it doesn't belong, it just looks stupid once you get down to the hard numbers; whether those numbers are performance, features, upgradability and whatever else you feel like dragging into the mix. The bottom line doesn't change and neither do the numbers of whichever performance benchmark you run. Ryzen still trails in single thread performance and AMD still suffers from it. The gap is smaller, but Ryzen is not the better CPU today, and I doubt 2nd gen is going to make it so.
Posted on Reply
#72
Frick
Fishfaced Nincompoop
"notb said:

Piledriver is still available. 2500K is not. Why do I see 2500K in a review, not an FX-8350? This has been my exact question for months. Can you answer it or not?
For what it's worth, Sweclockers uses the FX 8350, that FX 9XXX, A10-7870K, Ivy Bridge, Sandy Bridge and even the Phenom II X6, and they do some tests with the CPUs at the same clocks even.
Posted on Reply
#73
notb
"Frick said:
For what it's worth, Sweclockers uses the FX 8350, that FX 9XXX, A10-7870K, Ivy Bridge, Sandy Bridge and even the Phenom II X6,
Yes, I've already mentioned this review. It is excellent, as are many other articles they've written - at least technically, because I have to rely on Google Translate...
and they do some tests with the CPUs at the same clocks even.
Not a huge fan, to be honest. I don't understand what useful information it provides. But hey, I don't care about IPC, so I'm not the target. :-)
Posted on Reply
#74
Vayra86
"notb said:
Yes, I've already mentioned this review. It is excellent, as are many other articles they've written - at least technically, because I have to rely on Google Translate...

Not a huge fan, to be honest. I don't understand what useful information it provides. But hey, I don't care about IPC, so I'm not the target. :)
IPC only matters for scientific purposes really. It has no bearing on end performance and never has. Its similar to this ridiculous mention of Tflops for GPUs, implying a relation where there is none.
Posted on Reply
#75
EarthDog
Many tests would beg to differ... IPC and clockspeed are important. If IPC wasn't important, we'd all still be running SB at 5 GHz, no?

There are real gains to be had from IPC in both real world applications, testing, and benchmarking.

Am I missing what you are saying here? I jumped in the middle of this silly p1ssing contest between notb and loon. :p
Posted on Reply
Add your own comment