• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

New Horizon: AMD New ZEN CPU Preview Event

lord

Member
Here is the thing, the 6900k is a HEDT but it isn't really targeted at people playing games. The event in question was targeted at gaming audiences, primarily, so it is almost certainly going to offer better prices than $1K.
Sure, but $250, would be amazing and it would cause a great (positive) upset in the CPU market, but I don't see that happening. They will definitely go bellow whatever Intel is charging though.
 
Here is the thing, the 6900k is a HEDT but it isn't really targeted at people playing games. The event in question was targeted at gaming audiences, primarily, so it is almost certainly going to offer better prices than $1K.

Event has no meaning on zen itself. Zen has to cover semi-custom(consoles), mobile, desktop and professional markets.

And AMD loves showing their performance against $1000 Intel cpus in games while quietly ignoring existence of highly clocked i7 quads to pryy on lack of knowledge of non enthusiast consumers.

And as far as zen is concerned it's more telling what wasn't shown (performance in cpu limited games) than meaningless benchmarks they did.
 

blu

Wants the largest console games publisher to avoid Nintendo's platforms.
Event has no meaning on zen itself. Zen has to cover semi-custom(consoles), mobile, desktop and professional markets.

And AMD loves showing their performance against $1000 Intel cpus in games while quietly ignoring existence of highly clocked i7 quads to pryy on lack of knowledge of non enthusiast consumers.

And as far as zen is concerned it's more telling what wasn't shown (performance in cpu limited games) than meaningless benchmarks they did.
Blender rendering and HandBrake encoding are two entirely CPU-bound benchmarks. Just saying.
 

Neo_Geo

Banned
if this chip is as fast as $1k intel cpu its not gping to be cheap. you guys are dreaming

If it's priced anywhere near the 6900k AMD can go ahead and chalk this up as a failure and move on to the next thing. I don't think they are dumb enough to price it over $500, maybe they are that dumb, but I really hope not. We need this to come out around a $400 price point to really cause a shake up in the market. Intel has had a monopoly for almost a decade now, and they are pricing as such.

Their $1,000 CPUs are only worth that much because there is no competition and it's the only choice at that level.
 
And as far as zen is concerned it's more telling what wasn't shown (performance in cpu limited games) than meaningless benchmarks they did.
Apart from the point already made that the benchmarks they did are CPU bound, so are the (unbenched) gaming comparisons they did. If a game on a PC with SLI Titan Xs isn't CPU bound, then no game ever is.
 
If their best chip is say $350-$400, does that not make things super interesting again in the CPU market? Also, another chip from AMD that rivals the 6700K at $225 would be AMAZING.
 

Renekton

Member
If it's priced anywhere near the 6900k AMD can go ahead and chalk this up as a failure and move on to the next thing. I don't think they are dumb enough to price it over $500, maybe they are that dumb, but I really hope not. We need this to come out around a $400 price point to really cause a shake up in the market. Intel has had a monopoly for almost a decade now, and they are pricing as such.

Their $1,000 CPUs are only worth that much because there is no competition and it's the only choice at that level.
Expecting $500 or below for a competitive 8c16t is unrealistic imo.
 

Diablos

Member
Seems like promising tech thus far but the Ryzen name is so stupid. Ugh.

Price point relative to how good it will actually be is crucial
 

tuxfool

Banned
They are also the ones which get the most out of multiple cores.

Lets then presume that because of that the 6900k wasn't boosting and was stuck at its base Frequency of 3.2 Ghz. Then we compare that to the 3.4 Ghz Ryzen, which also wasn't boosting.

The Ryzen isn't as fast, but it is pretty good, or at least close enough. And here we're not even comparing the relative TDP of each processor.

What remains to be seen is how Ryzen overclocks or what its boost frequencies look like. Also I don't know how well each of those applications take advantage of SIMD extensions, though something tells me that Intel will be superior in that area.
 

grumble

Member
To be fair, Skylake i7 (and next month's Kaby Lake) wipes the floor with 6900K on less-threaded apps as well.

My guess is intel smokes them moderately in gaming benchmarks due to single thread performance. AMD will counter with price and cores (both of which will squeeze margin). AMD will take a few percent of servers and make some money.
 

JaseC

gave away the keys to the kingdom.
To be fair, Skylake i7 (and next month's Kaby Lake) destroys 6900K on that as well.

To be fair also, Intel's "enthusiast" platform is a generation behind. The current slate is Broadwell-based and next year marks the debut of Skylake-X and Kaby Lake-X, although current reports indicate that the latter will be limited to four cores (presumably with HyperThreading).
 
Well, Intel's "enthusiast" platform is a generation behind. The current slate is Broadwell-based and next year marks the debut of Skylake-X.

HEDT is based on their server platform which is always a generation behind the mainstream platform. Server and workstation applications demand a higher level of validation and qualification before big companies will buy and deploy. Xeon is also on Broadwell and will not move to Skylake until 2017, followed by HEDT.

Although now with AMD preparing Ryzen, Intel is also thinking about how 8c/16t at mainstream pricing could affect them, which is why they are also hedging with Coffee Lake which may bring 6c/12t and 8c/16t to a mainstream platform. It's not clear what this would mean for HEDT since it would largely make that segment unnecessary if Coffee Lake has the same core/thread count as Skylake-X.

Intel doesn't particularly want to give up the margins they currently enjoy with HEDT versus mainstream but the real money on high core count processors has always been in servers anyways. They don't have much to lose if AMD attempts to undercut, the real question is how suicidal AMD management is because if they start a price war and lose a lot of money, they are the ones that will go bankrupt, not Intel.
 

JaseC

gave away the keys to the kingdom.
HEDT is based on their server platform which is always a generation behind the mainstream platform. Server and workstation applications demand a higher level of validation and qualification before big companies will buy and deploy.

I assume you're simply piggy-backing off my post in the interest of clarity, but just to be clear, I didn't mean to imply that its enthusiast-level product slate is a generation behind for no/a nefarious reason.
 
Lets then presume that because of that the 6900k wasn't boosting and was stuck at its base Frequency of 3.2 Ghz. Then we compare that to the 3.4 Ghz Ryzen, which also wasn't boosting.

The Ryzen isn't as fast, but it is pretty good, or at least close enough. And here we're not even comparing the relative TDP of each processor.

What remains to be seen is how Ryzen overclocks or what its boost frequencies look like. Also I don't know how well each of those applications take advantage of SIMD extensions, though something tells me that Intel will be superior in that area.
The 6900k was boosting, though..
 

FinKL

Member
They kept pointing over and over that they're beating a $1k CPU, that tells me pricing is important to them for sure.
$900 is technically "beating" them.

Holding my expectations in check, hoping for super cheap but we all know it will be on par with intels offering
 

riflen

Member
It was? On all 8 cores?

AMD used an out of the box config for the 6900k. I believe this means one core can reach 4.0 Ghz with Turbo Boost 3.0, depending on thermal conditions, but I think that's only when the driver detects a single-threaded workload. In multi-threaded workloads, it would boost all cores to 3.7 Ghz. AMD did limit the Intel system to dual-channel memory.

https://i.redd.it/r39v4xwzre3y.jpg

The Ryzen CPU was fixed at 3.4 Ghz.
 

riflen

Member
So I guess that no matter wat Price AMD goes out with Intel will always price lower and AMD will be the big loser or?

AMD are not going to take the performance crown with Ryzen. Intel will be able to charge more because performance is important to people.
There are some customers who are more cost-sensitive and will settle for 85% of the performance for 85% of the cost, for example.

It depends on which bracket you're talking about. AMD, like Intel, will want some of that sweet enthusiast money. I think AMD will try to position Ryzen in a price tier that Intel don't cover very well. If you want an 8-core Intel CPU, you have to pay about $1000. This is an obvious place for AMD to target. There is no "slightly slower 8-core CPU for less money" from Intel, but Ryzen could fill this.
 
It was? On all 8 cores?

oh I thought they said in the steam they'd locked it to base frequencies

Yeah, pretty sure the 6900K was boosted. IIRC they mentioned that they locked the frequency on their own chip since they weren't prepared to announce it's boosted frequencies yet.

I'll watch it again to make sure I didn't miss anything, but yeah...

Edit: yeah, the 6900K was "out of the box" the only chip they disabled boost on was their own.

So yeah, the Zen beat a comparable Intel that was boosted to a higher frequency. Since the thread count and everything else was identical, that would mean the the Zen would need higher IPC to match the higher clocked Intel... Unless they pulled some shenanigans like putting a shitty cooler on it to keep it throttled or something, which is getting into conspiracy theory territory.
 

Oreoleo

Member
I'd like that too.

There's basically nothing to summarize. AMD will have CPU's that can compete with modern Intel CPU's for the first time in like a decade including one with 8 cores/16 threads. Whether they are better/worse/cheaper/more expensive is still anyone's guess. That's just about all anyone knows.
 

dr_rus

Member
Lets then presume that because of that the 6900k wasn't boosting and was stuck at its base Frequency of 3.2 Ghz. Then we compare that to the 3.4 Ghz Ryzen, which also wasn't boosting.

The Ryzen isn't as fast, but it is pretty good, or at least close enough. And here we're not even comparing the relative TDP of each processor.

What remains to be seen is how Ryzen overclocks or what its boost frequencies look like. Also I don't know how well each of those applications take advantage of SIMD extensions, though something tells me that Intel will be superior in that area.

The 6900K is also an 8c/16t processor, but I understand the assumption to the contrary as the 5930K is 6c/12t.

What I've meant is that this type of benchmark is hardly relevant to judge CPUs performance in actual games. Just wait for proper benchmarks, most of stuff which is out right now are either controlled leaks or engineering samples results.
 

JaseC

gave away the keys to the kingdom.
What I've meant is that this type of benchmark is hardly relevant to judge CPUs performance in actual games.

Ah, right.

Just wait for proper benchmarks, most of stuff which is out right now are either controlled leaks or engineering samples results.

Oh, I'm acutely aware that it's not wise to draw a conclusion from a single point of data.
 

Durante

Member
What remains to be seen is how Ryzen overclocks or what its boost frequencies look like. Also I don't know how well each of those applications take advantage of SIMD extensions, though something tells me that Intel will be superior in that area.
I think what primarily remains to be seen is sequential (and "mediocre scaling parallel") performance. While more applications and games become more parallelized, in a huge amount of real-world use cases you are still partially or even entirely limited by sequential performance.
 

Mr.Mike

Member
How quickly do compiler vendors incorporate new CPU instructions anyway? And even if they did, I don't recall ever seeing different builds of software available for download for specific versions of a certain architecture, but I do know that new CPU instructions are added with new CPU's all the time. But if somebody is distributing pre-compiled software they'd have to build for the lowest common denominator anyway, no?

I suppose if you're compiling stuff on your own you could set the compilation flags to target the very specific processor your have, but very few games have their sources available for download. I guess if you're using a package manager / digital games downloading platform they could have a bunch of different builds for specific iterations of an architecture and send you the most appropriate one. I don't recall ever hearing about this being done though.

Anther question for those who would know. Hyper-threading basically means that the CPU is able to switch execution from one thread on a core to the other thread on a core very quickly, but you'd never have more than your number of physical cores actually executing code at the same time, correct? So would hyperthreading be very useful for workloads with lots of short-lived threads, but largely useless for workloads with long lived threads?
 
How quickly do compiler vendors incorporate new CPU instructions anyway? And even if they did, I don't recall ever seeing different builds of software available for download for specific versions of a certain architecture, but I do know that new CPU instructions are added with new CPU's all the time. But if somebody is distributing pre-compiled software they'd have to build for the lowest common denominator anyway, no?

I suppose if you're compiling stuff on your own you could set the compilation flags to target the very specific processor your have, but very few games have their sources available for download. I guess if you're using a package manager / digital games downloading platform they could have a bunch of different builds for specific iterations of an architecture and send you the most appropriate one. I don't recall ever hearing about this being done though.

Anther question for those who would know. Hyper-threading basically means that the CPU is able to switch execution from one thread on a core to the other thread on a core very quickly, but you'd never have more than your number of physical cores actually executing code at the same time, correct? So would hyperthreading be very useful for workloads with lots of short-lived threads, but largely useless for workloads with long lived threads?
Even before compilers, the first question we should be asking ourselves is when will the Windows OS support it? Both early reviews for Bulldozer and Skylake didn't necessarily show all the performance improvements from the processors due to the OS not being updated at the time of release.
 
What was this slide all about at the conference?

C0FMDAlUQAQ-am3.jpg:large
 
Unless they are hitting the Ryzen cpu be used on the Xbox One Project Scorpio?
Could you not be bothered to actually watch the event? Here's what was said about that slide:

Lisa Su said:
And we are also working with Microsoft on Project Scorpio, which will launch next year. All of these systems are powered by AMD.

That "all of these systems" collectively refers to PS4 slim, PS4 Pro, Xbox One S, and Scorpio. So no, this is not evidence that Zen will be used in Microsoft's 2017 console, any more than the presence of those older devices in the presentation was evidence that they do.
 
Edit: FAKE


Rumour is fake apparently, thread were the rumour started got deleted from Baidu forum and a mod confirmed it was fake. The benchmarks were run on a E5 2660.


Not sure if it's been discussed yet, but there are new leaked benchmarks... wccftech, grain of salt, all that...

http://wccftech.com/amd-ryzen-zen-cpu-benchmarks-leak/

Code:
AMD RYZEN vs Core i7-7700K in Cinebench R15:

i7-7700K stock 966 cb

i7-7700K (5GHz) 1083 cb

Ryzen (8/16) 3.4 GHz 1188cb
There were also 6900 and 6950 listed, but both of them had ~ scores... meaning they were round-abouts or estimates...

Code:
AMD RYZEN vs Core i7-7700K in Fritz Chess:

FRITZ Chess Relative Score	
Core i7-7700K (Stock) 35.52
	
Core i7-7700K (5GHz 41.44	

Core i7-6900K (Stock) 47.80	

Core i7-6950X (Stock) 51.50	

AMD RYZEN (8/16) 36.86
 
Not sure if it's been discussed yet, but there are new leaked benchmarks... wccftech, grain of salt, all that...

http://wccftech.com/amd-ryzen-zen-cpu-benchmarks-leak/

Code:
AMD RYZEN vs Core i7-7700K in Cinebench R15:

i7-7700K stock 966 cb

i7-7700K (5GHz) 1083 cb

Ryzen (8/16) 3.4 GHz 1188cb
There were also 6900 and 6950 listed, but both of them had ~ scores... meaning they were round-abouts or estimates...

Code:
AMD RYZEN vs Core i7-7700K in Fritz Chess:

FRITZ Chess Relative Score	
Core i7-7700K (Stock) 35.52
	
Core i7-7700K (5GHz 41.44	

Core i7-6900K (Stock) 47.80	

Core i7-6950X (Stock) 51.50	

AMD RYZEN (8/16) 36.86

Looks pretty good, but the Fritz Chest benches are strange, the scaling looks pretty off to me..
 
Not sure if it's been discussed yet, but there are new leaked benchmarks... wccftech, grain of salt, all that...

http://wccftech.com/amd-ryzen-zen-cpu-benchmarks-leak/

Code:
AMD RYZEN vs Core i7-7700K in Cinebench R15:

i7-7700K stock 966 cb

i7-7700K (5GHz) 1083 cb

Ryzen (8/16) 3.4 GHz 1188cb
There were also 6900 and 6950 listed, but both of them had ~ scores... meaning they were round-abouts or estimates...

Code:
AMD RYZEN vs Core i7-7700K in Fritz Chess:

FRITZ Chess Relative Score	
Core i7-7700K (Stock) 35.52
	
Core i7-7700K (5GHz 41.44	

Core i7-6900K (Stock) 47.80	

Core i7-6950X (Stock) 51.50	

AMD RYZEN (8/16) 36.86

Remember when WCCFTech was reporting that the Fury X was going to bury the Titan X? Good times.
 
Top Bottom