AMD's 'RYZEN' Detailed - Octacore, 3.4GHz + & Dynamic Overclocking

[Guest lilbman]

8 minutes ago, Drak3 said:

It's very likely that 3.4GHz is the base clock. Hard to market 3.4GHz+ as the only number on clockspeed if the base was 3.34GHz or less.

That + could mean anything so I assume it's the boost clock.  It would make more sense this way because a + means more or faster

[Drak3]

8 minutes ago, MageTank said:

Where are you getting that information? Last I checked, the rumors were aiming at Haswell IPC. I know there was that one cinebench run, with both clocked at 3ghz, but we didn't know anything about the settings used, or the testing methodology to actually confirm Broadwell-E IPC. Still, Haswell IPC with 8c/16t for less than $1000? That's enough to compel enthusiasts to give it a shot. 

AMD's Blender benchmark pinning a Zen 8c/16t vs an "unnamed" Broadwell-E 8c/16t, both at 3GHz. I know a single benchmark isn't a good metric, hence why I worded it in a way that is supposed to convey that it might change with future information.

And when AMD first announced Zen, Haswell IPC was the mark. But jumping from Haswell to Broadwell was not a challenge, and Zen was delayed. Broadwell is well quoted as being 3% better in terms of IPC.

According to the site's IPC test at the end of this comment, Broadwell's best bench pinned it as 3.8% better, and the worst as .3% better, with an average of 2.25%.

 

https://us.hardware.info/reviews/6215/2/intel-five-generation-ipc-test-broadwell-haswell-ivy-bridge-sandy-bridge-and-nehalem-results

[ARikozuM]

Just now, deXxterlab97 said:

Are you saying AMD name their product after... a food?

Perhaps they did. Begging for food or money?

 

P.s. a reuben is a rye sandwich with corned beef and sauerkraut.

[Drak3]

1 minute ago, lilbman said:

That + could mean anything so I assume it's the boost clock.  It would make more sense this way because a + means more or faster

That "+" means it can be higher, and in marketing, it also means that the number given is the minimum under normal circumstances.

It makes more sense that 3.4GHz is the base clock, because it can boost higher than that.

[Guest]

3 minutes ago, ARikozuM said:

Perhaps they did. Begging for food or money?

 

P.s. a reuben is a rye sandwich with corned beef and sauerkraut.

 

Given the fact that AMD been naming their product after odd stuff like a Bulldozer?

But Ryzen? WTF is Ryzen. I know it's a common Asian name. But who is Ryzen then?

[2FA]

1 minute ago, ONOTech said:

I think my biggest issue with this claim is that IPC is not static. It depends on the software you use, so there's not one, single metric that's fair to establish a set IPC for a processor. It's a dynamic trait. A blanket statement like this should definitely be taken with a grain of salt, as AMD could be measuring IPC with one piece of software that favors their architecture.

 

Of course, the likelihood of that is probably slim, but I wouldn't expect Broadwell. Expect Ivy, hope for Haswell, rejoice for anything faster.

Or don't expect anything and just wait and see...

[Guest]

8 minutes ago, Drak3 said:

That "+" means it can be higher, and in marketing, it also means that the number given is the minimum under normal circumstances.

It makes more sense that 3.4GHz is the base clock, because it can boost higher than that.

True, especially when some idiot compares CPU solely off base clock, claming FX 8350 has base of 4.0 GHZ so it out beats i5 and i7 (6700k has base of 4.0 GHZ)

[ARikozuM]

Just now, deXxterlab97 said:

 

Given the fact that AMD been naming their product after odd stuff like a Bulldozer?

But Ryzen? WTF is Ryzen. I know it's a common Asian name. But who is Ryzen then?

Does each raisin represent 1 IPC? Or does it indicate the total at 5GR? Cuz that's not a lot of raisins!

[Guest]

Just now, ARikozuM said:

Does each raisin represent 1 IPC? Or does it indicate the total at 5GR? Cuz that's not a lot of raisins!

It's AMD, what do you expect

[79wjd]

18 minutes ago, deXxterlab97 said:

Are you saying AMD name their product after... a food?

I was thinking rise-zen

[ARikozuM]

Just now, deXxterlab97 said:

It's AMD, what do you expect

Fire?

[sazrocks]

AMD please do this right. Intel needs more competition to keep prices down.

[Guest]

1 minute ago, djdwosk97 said:

I was thinking rise-zen

Rise zen shine Mr Freeman, rise zen shine

[Guest]

Just now, sazrocks said:

AMD please do this right. Intel needs more competition to keep prices down.

Been saving enough money, finally Intel decides to lower the price of its first gen i3 to compete with AMD

[Drak3]

4 minutes ago, deXxterlab97 said:

Given the fact that AMD been naming their product after odd stuff like a Bulldozer?

But Ryzen? WTF is Ryzen. I know it's a common Asian name. But who is Ryzen then?

Zen is potentially AMD's ticket of ryzen back to the top.

 

2 minutes ago, deXxterlab97 said:

True, especially when some idiot compares CPU solely off base clock, claming FX 8350 has base of 4.0 GHZ so it out beats i5 and i7 (6700k has base of 4.0 GHZ)

Well, if AMD is right about Zen's IPC, comparing the base clocks might be a half way decent metric when comparing against Broadwell-E.

[ARikozuM]

1 minute ago, sazrocks said:

AMD please do this right. Intel needs more competition to keep prices down.

Forget pricing, performance is what we need.

[Guest]

Just now, Drak3 said:

Zen is potentially AMD's ticket of ryzen back to the top.

-snip

 

Well, if AMD is right about Zen's IPC, comparing the base clocks might be a half way decent metric when comparing against Broadwell-E.

Then it just gives us higher clock speed, more cores and less IPC

 

100 cores, 200 threads at 10GHZ. With idle temp at 80

 

 

[Guest]

1 minute ago, ARikozuM said:

Forget pricing, performance is what we need.

*value. I would not spend 1000$ for a xeon chip just to get 2 more fps

[Drak3]

6 minutes ago, ONOTech said:

I think my biggest issue with this claim is that IPC is not static. It depends on the software you use, so there's not one, single metric that's fair to establish a set IPC for a processor. It's a dynamic trait. A blanket statement like this should definitely be taken with a grain of salt, as AMD could be measuring IPC with one piece of software that favors their architecture.

 

Of course, the likelihood of that is probably slim, but I wouldn't expect Broadwell. Expect Ivy, hope for Haswell, rejoice for anything faster.

The big issue is that consumers and reviewers have no way of consistently measuring IPC, because it is a static trait of CPUs. The problem is that every single piece of software we have to test it with is optimized for select generations from each brand (or just one brand), or can even be optimized for select CPUs. We also don't have the time or equipment to eliminate every variable.

[Drak3]

3 minutes ago, deXxterlab97 said:

Then it just gives us higher clock speed, more cores and less IPC

 

100 cores, 200 threads at 10GHZ. With idle temp at 80

 

 

WHOA PARTNER! WHO TOLD YOU OF AMD'S SUCCESSOR OF ZEN: YIN-YANG.

[RagnarokDel]

52 minutes ago, HKZeroFive said:

snip

It's just risen

[patrickjp93]

It's 16MB of L3 cache and 512KB of L2 per core vs. Intel's 20MB of L3 and 256KB of L2 per core on the 5960X/6900K

[79wjd]

1 minute ago, patrickjp93 said:

It's 16MB of L3 cache and 512KB of L2 per core vs. Intel's 20MB of L3 and 256KB of L2 per core on the 5960X/6900K

Source? 

 

Ha, who am I kidding

[patrickjp93]

8 minutes ago, Drak3 said:

The big issue is that consumers and reviewers have no way of consistently measuring IPC, because it is a static trait of CPUs. The problem is that every single piece of software we have to test it with is optimized for select generations from each brand (or just one brand), or can even be optimized for select CPUs. We also don't have the time or equipment to eliminate every variable.

Yes they do. It's called use Agner Fog's scripts he built to do exactly that.

[patrickjp93]

1 minute ago, djdwosk97 said:

Source? 

 

Ha, who am I kidding

http://www.anandtech.com/show/10591/amd-zen-microarchiture-part-2-extracting-instructionlevel-parallelism

 

You lose again.