AMD launches Zen+ 12nm Ryzen and X470 motherboards

juanrga said:
With Guru3D using engineering samples for Intel chips and other tricks? Sure. Without such tricks Zen+ IPC is 8% behind skylake in Cinebench

View attachment 67930

And remember Cinebench is favouring AMD. The IPC gap is higher in other benches and when averaging we obtain about 12--25%.
Click to expand...
Can't wait to get home and spend 8 hours working on and playing Cinebench, WPrime, and Sysmark 2016.
 
juanrga said:
Optimizing for one almost always comes at the expense of the other (that is why memory controllers for GPUs and CPUs are different). That is why I expected that latency improvements to be paid in bandwidth terms.
Click to expand...

Appears this was the correct decision on AMD's part, though. Looking at detailed improvements (Anand, Toms, etc...) real world appears to gain a lot more than suffer from the change, even if synthetics suffer a little.
 
juanrga said:
With Guru3D using engineering samples for Intel chips and other tricks? Sure. Without such tricks Zen+ IPC is 8% behind skylake in Cinebench

View attachment 67930

And remember Cinebench is favouring AMD. The IPC gap is higher in other benches and when averaging we obtain about 12--25%.
Click to expand...

Don't know about these "tricks" you mention, but I do trust Toms over Guru3D. My impression is Zen+ has a broadly Broadwell (been waiting all day to type that) IPC in throughput. Somewhat less than that in latency-sensitive tasks, but improved considerably over first-gen Zen. AMD still needs those two extra cores to hang out in the 8700k bracket.
 
DuronBurgerMan said:
Don't know about these "tricks" you mention, but I do trust Toms over Guru3D. My impression is Zen+ has a broadly Broadwell (been waiting all day to type that) IPC in throughput. Somewhat less than that in latency-sensitive tasks, but improved considerably over first-gen Zen. AMD still needs those two extra cores to hang out in the 8700k bracket.
Click to expand...

Same price, hangs around the 8700k in games, murders it in productivity... seems like a winner to me.
 
DuronBurgerMan said:
Don't know about these "tricks" you mention, but I do trust Toms over Guru3D. My impression is Zen+ has a broadly Broadwell (been waiting all day to type that) IPC in throughput. Somewhat less than that in latency-sensitive tasks, but improved considerably over first-gen Zen. AMD still needs those two extra cores to hang out in the 8700k bracket.
Click to expand...

It doesn’t need extra cores. 8700k is probably boosting higher as well. In productivity it leaps ahead so you can’t reallt say it needs extra cores to keep up with 8700k. It beats it with extra cores where it matters. If you are gaming then you won’t see much difference between 6 or 8 core. That even goes for Intel vs Intel as well.
 
DuronBurgerMan said:
Don't know about these "tricks" you mention, but I do trust Toms over Guru3D. My impression is Zen+ has a broadly Broadwell (been waiting all day to type that) IPC in throughput. Somewhat less than that in latency-sensitive tasks, but improved considerably over first-gen Zen. AMD still needs those two extra cores to hang out in the 8700k bracket.
Click to expand...

I have discussed extensively the tricks used by Guru3D in my discussions with OrangeKrush. The use of engineering samples and beta BIOS for Intel chips is the most obvious move Guru3D do to reduce the performance gap between Intel and AMD. Virtually any other review site is better than Guru3D.

NKD said:
It doesn’t need extra cores. 8700k is probably boosting higher as well. In productivity it leaps ahead so you can’t reallt say it needs extra cores to keep up with 8700k. It beats it with extra cores where it matters. If you are gaming then you won’t see much difference between 6 or 8 core. That even goes for Intel vs Intel as well.
Click to expand...

2700X needs 33% moar cores to beat 8700k in multithreaded scenarios as Blender, Handbrake and so

handbrake_0.png


blender-gooseberry_0.png
 
Last edited:
NKD said:
It doesn’t need extra cores. 8700k is probably boosting higher as well. In productivity it leaps ahead so you can’t reallt say it needs extra cores to keep up with 8700k. It beats it with extra cores where it matters. If you are gaming then you won’t see much difference between 6 or 8 core. That even goes for Intel vs Intel as well.
Click to expand...

It needs the cores. But if AMD is providing the extra cores for the same money, it doesn't really matter. Price/performance is the metric that really matters most, and in that the 2700X seems competitive, if somewhat disappointing from an OC perspective.
 
DuronBurgerMan said:
It needs the cores. But if AMD is providing the extra cores for the same money, it doesn't really matter. Price/performance is the metric that really matters most, and in that the 2700X seems competitive, if somewhat disappointing from an OC perspective.
Click to expand...

Does is not outperform 8700k in productivity stuff? If it was performing the same then yes you would have a point. It’s out performing the 8700k with those extra cores. In gaming it’s pointless. You could say the same thing for Intel 8 core chips vs 8700k.

What you should say is it lacks little clock speed. Other than that it’s not using more cores to catch up to 8700k. It actually outperforms it where the cores matter.
 
juanrga said:
I have discussed extensively the tricks used by Guru3D in my discussions with OrangeKrush. The use of engineering samples and beta BIOS for Intel chips is the most obvious move Guru3D do to reduce the performance gap between Intel and AMD. Virtually any other review site is better than Guru3D.



2700X needs 33% moar cores to beat 8700k in multithreaded scenarios as Blender, Handbrake and so

View attachment 67954

View attachment 67955
Click to expand...

Thanks for proving my point it’s faster. Oh it needs more core to be faster. No dah! if it was equal or slower with more cores in majority of benchmarks you would have a point lol.
 
NKD said:
Does is not outperform 8700k in productivity stuff? If it was performing the same then yes you would have a point. It’s out performing the 8700k with those extra cores. In gaming it’s pointless. You could say the same thing for Intel 8 core chips vs 8700k.

What you should say is it lacks little clock speed. Other than that it’s not using more cores to catch up to 8700k. It actually outperforms it where the cores matter.
Click to expand...

You're preaching to the choir, mang. All I'm saying is AMD needs a few extra cores to play in the same league. But there isn't anything wrong with that. It's like saying Ford needs 8 cylinders to keep up with a turbocharged 6 cylinder from another manufacturer. Doesn't matter so much. End result, price/performance matters.
 
juanrga said:
I have discussed extensively the tricks used by Guru3D in my discussions with OrangeKrush. The use of engineering samples and beta BIOS for Intel chips is the most obvious move Guru3D do to reduce the performance gap between Intel and AMD. Virtually any other review site is better than Guru3D.
Click to expand...

Whatever. It's not one of my usual goto sites anyway, regardless of who does or does not like it. I trust the old players, mang. Toms, Anand, [H]ard.
 
DuronBurgerMan said:
You're preaching to the choir, mang. All I'm saying is AMD needs a few extra cores to play in the same league. But there isn't anything wrong with that. It's like saying Ford needs 8 cylinders to keep up with a turbocharged 6 cylinder from another manufacturer. Doesn't matter so much. End result, price/performance matters.
Click to expand...

LOL I am not. I don't even have AMD haven't had them in the last decade. What AMD doesn't have is as good of clock speed as Intel. That is the problem. Intel keeps hush hush about turbo on their CPUs now. I can pretty much bet 8700k is clocking over 4ghz under load on all cores as long as it has decent temps which review sites definitely do. As I said if 2700x was struggling to keep up and was not ahead at all then yes I would agree with you that it needs more core to keep up. More cores to beat it is different scenario than more cores to keep up with it. If more cores beat it for the same price then I will take it lol. If its barely getting close in every scenario or little slower then I wouldn't even consider buying it. Not that I am upgrading now anyways. Zen 2 on 7nm is when I consider switching from my 7600k. So far I don't see a point for me to ugprade. 7600k will go in secondary computer and the old parts will get retired. lol.
 
NKD said:
LOL I am not. I don't even have AMD haven't had them in the last decade. What AMD doesn't have is as good of clock speed as Intel. That is the problem. Intel keeps hush hush about turbo on their CPUs now. I can pretty much bet 8700k is clocking over 4ghz under load on all cores as long as it has decent temps which review sites definitely do. As I said if 2700x was struggling to keep up and was not ahead at all then yes I would agree with you that it needs more core to keep up. More cores to beat it is different scenario than more cores to keep up with it. If more cores beat it for the same price then I will take it lol. If its barely getting close in every scenario or little slower then I wouldn't even consider buying it. Not that I am upgrading now anyways. Zen 2 on 7nm is when I consider switching from my 7600k. So far I don't see a point for me to ugprade. 7600k will go in secondary computer and the old parts will get retired. lol.
Click to expand...

Lol I see the source of the confusion. I'm not saying it needs *more than it has*. I'm saying Zen architecture needs 8 cores to keep up with Intel's 6, and so on and so forth down the line. (2600x looks good next to the 8600k because of extra threads, etc...).

I'm not saying the 2700x doesn't have enough as-is. It does quite well, trading blows with the 8700k. After reading all the reviews today, I regard the results as pretty much a tie - to be defined more by use case than which is 'better' than the other.
 
DuronBurgerMan said:
I'm not saying the 2700x doesn't have enough as-is. It does quite well, trading blows with the 8700k. After reading all the reviews today, I regard the results as pretty much a tie - to be defined more by use case than which is 'better' than the other.
Click to expand...

They were already almost there- I'd happily have run an R7 v1 if the single-thread performance were there; the multi-thread performance just doesn't mean as much. But they are getting closer!
 
IdiotInCharge said:
They were already almost there- I'd happily have run an R7 v1 if the single-thread performance were there; the multi-thread performance just doesn't mean as much. But they are getting closer!
Click to expand...

Well in typical AMD fashion they botched Ryzen V1 launch. Even trying to find the reviews after they fixed their bios (and got a huge perf increase) is hard.

Looks like they finally did a launch right. Anandtech, hardocp and pcper all show Ryzen v2 kicking ass. I went with the 2700x over the 6800k or 7820k...

The discrepencies between the reviews seem to be if they have the Intel security holes patched or not from what I can tell. Intel gets crushed. Maybe they shouldn't have sat on their asses and jerked off to their fortune for so long.

This cluster of events couldn't have happened to a nicer company IMO. ;)
 
Last edited:
Alas, gaming wise still a little wary.

The charts and benchmarks being posted for gaming do look promising at face value, and it genuinely does look like AMD has addressed things that were covered during and may have hindered the original Ryzen launch. Credit to AMD there (and thanks to those that were the canaries in the mine for the rest of us). But then reading or listening to the reviewers commenting that "remember that this game/engine has a frame cap" and seeing instances of non-K i3's and i5's numbers converging with those of the i7 8700K and R7 2700X in some of the charts posted thanks to some other bottleneck, lessens their significance, and I cannot escape two thoughts...

1. What if Intel does manage to get a 8/16 i7 based on Coffee Lake or better out shortly?
2. Will these charts look different with graphics cards beyond the GTX 1080 and 1080 TI?

I know, I know... there will also be something better "just around the corner", and waiting or trying to future proof is a fool's errand, but as someone that upgrades my graphics card far more often than I have my CPU/motherboard combo, I give it more time than I should. That's probably the problem though... lack of competition having kept my current i5 relevant for as long as it has. That things might progress faster now that there is competition again, I might not be able to keep getting away with that.

Definitely a step up from what I have. Maybe I'll go back and look at the <1% lows and see if I can convince myself that way.
 
NKD said:
Thanks for proving my point it’s faster. Oh it needs more core to be faster. No dah! if it was equal or slower with more cores in majority of benchmarks you would have a point lol.
Click to expand...

I think you didn't understand what I said neither what he said.

DuronBurgerMan said:
Lol I see the source of the confusion. I'm not saying it needs *more than it has*. I'm saying Zen architecture needs 8 cores to keep up with Intel's 6, and so on and so forth down the line. (2600x looks good next to the 8600k because of extra threads, etc...).
Click to expand...

Your point was clear.
 
So the HFR review is out. My prediction


There is no 2600 at launch, so only 2700X and 2600X were reviewed

getgraphimg.png


So Pinnacle Ridge is about 2--3% faster than I expected.
 
IdiotInCharge said:
They were already almost there- I'd happily have run an R7 v1 if the single-thread performance were there; the multi-thread performance just doesn't mean as much. But they are getting closer!
Click to expand...
We all know that when you buy 8c/16t the most important thing is single thread and multi-thread does not matter.
 
New post from The Stilt addresses a ton of questions that were posed in this thread around XFR, differences between 370 and 470 boards, latencies, practical difference between 12 and 14nm, etc. Really good technical read as usual: https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-72#post-39391302

I was especially interested in the info on individual core binning and how slight variability between chips is inherent to getting out every last drop of performance for these CPUs. Very awesome work.
 
juanrga said:
With Guru3D using engineering samples for Intel chips and other tricks? Sure. Without such tricks Zen+ IPC is 8% behind skylake in Cinebench

And remember Cinebench is favouring AMD. The IPC gap is higher in other benches and when averaging we obtain about 12--25%.
Click to expand...

Most sites shown the Ryzen to be neck and neck with Skylake. Tom's may have had loose memory timings?? However, the 2600x actually managed to tie the 8700k in multithread (around 1400) despite being clocked slightly lower. This is due to AMD smt being more efficient.


_mockingbird said:
The Ryzen 7 2700X has managed to beat the Core i7-7820X across the board in games apparently.
Click to expand...

Overall, they trade blows. I realize the 2700x is competing against the 8700k, but it is a bit unfair to expect it to match the CFL in games, even a similiar clocks. This will never happen with the Mesh interconnect against a Ring Bus. Now when you compare the 2600x vs the 7800x, gaming performance is nearly identical.

If AMD release a 2900x, it will be the long lost twin to the 7820x across the board.
 
So... should I sell the 6700K/Z170 and go with a 2700x?

Gaming is much less important to me than it was in 2015 when I built the 6700k system.
 
Pieter3dnow said:
We all know that when you buy 8c/16t the most important thing is single thread and multi-thread does not matter.
Click to expand...

That's exactly what I said!

Or not even close, lol.

For me, both are important, but over four cores I'll take extra single-core performance before more cores. That's why I'm running an 8700k and not a 1700x, and why I'm personally not interested in trading for a 2700x. That's because, for me, the most strenuous thing I do that is time sensitive is gaming. If I did enough compiling or content creation that processing time was money, I'd own a TR rig.
 
Silent.Sin said:
New post from The Stilt addresses a ton of questions that were posed in this thread around XFR, differences between 370 and 470 boards, latencies, practical difference between 12 and 14nm, etc. Really good technical read as usual: https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-72#post-39391302

I was especially interested in the info on individual core binning and how slight variability between chips is inherent to getting out every last drop of performance for these CPUs. Very awesome work.
Click to expand...

Indeed, I did learn more about he confirms that Pinnacle Ridge uses same Zen cores than Raven Ridge, that Glofo claims about 12LP were pure unadultered BS, that IPC gain is 1.5%, that real TDP of 2700X is 140W, than safe OC is about 4.2GHz,...
 
Nightfire said:
This is due to AMD smt being more efficient.
Click to expand...

AMD SMT implementation isn't more efficient.

_mockingbird said:
You said that Ryzen 7 2700X would be slower than Core-7820X.

You failed.

View attachment 68172
Click to expand...

So I make a statement about applications and you ignore what I said and that the review proves it is slower (318.6 points for R7 2700X vs 328.8 points for i7-7820X), then link to games result and claim I "failed". ROFL

Better start apologizing for all what you said about CPCHardware, because reviews confirmed that they were correct.

pendragon1 said:
so youre just pointing out you were wrong?
Click to expand...

Sure, considering the uncertainty that I estimated


there is a 0--1% variation between my prediction and measurement. It is obvious that I failed miserably. Only the people that predicted 4.8GHz stock clocks out of a magic 12LP node, and 10% IPC improvements got it ALL right!

lzupk.jpg
 
juanrga said:
AMD SMT implementation isn't more efficient.



So I make a statement about applications and you ignore what I said and that the review proves it is slower (318.6 points for R7 2700X vs 328.8 points for i7-7820X), then link to games result and claim I "failed". ROFL

Better start apologizing for all what you said about CPCHardware, because reviews confirmed that they were correct.



Sure, considering the uncertainty that I estimated


there is a 0--1% variation between my prediction and measurement. It is obvious that I failed miserably. Only the people that predicted 4.8GHz stock clocks out of a magic 12LP node, and 10% IPC improvements got it ALL right!

View attachment 68236
Click to expand...

Why would I apologize for being right?

Before launch, you claimed that Ryzen 7 2700X would be barely faster than the Core i7-6900K.

You are always “right”, because you make so many predictions and when you get some of them wrong, you pretended like that never happened.
 
Just ordered 2700X with MB from Newegg to augment my 5820k. Last AMD CPU I owned was x2 3800. Excited to be back on team green.
 
dzyp said:
Still play classic OSP sometimes but not many on and few good servers. Did everyone move over?
Click to expand...
CROM CTF is still going. There's a discord for QL pickups that is pretty dead, but we have gotten a few games recently.
Not really anyone moved over. QL CTF is a completely different community. You'd know bogdog, though... played with him a few nites ago
 
You must log in or register to reply here.
Back
Top