Ryzen "2" ?

[DragonQ]

I wonder what the new chipsets will offer, given all of the key components including the IMC are on the CPU itself.

157 to 178 is a 13% improvement in the Cinebench single-thread run. They claim the chip is running at 4.35 GHz, whereas the 1800X would be running at 4-4.1 GHz in ideal conditions. So worst case scenario, it would be around 9% better with just the clock speed bump. I suspect the additional 3-4% comes from the cache latency improvements.

The multithreaded score is only up by 8%, which is odd because that's what you'd expect if you simply got the 1800X to run at 4.0 GHz on all cores. Maybe that's what the 2800X will be: 4.0 GHz boost on all cores, up to 4.35 GHz XFR single core?

It's all very confusing when you never have any idea what clock speed chips are running at in benchmarks runs.

 

[sideways14a]

If thats accurate then its a fairly significant refresh this, the 2800x is not slow - if it can over clock as well to 4.5 or more then ... well intel will need to rush out another series of chips this year.

 

[Steveocee]

I don't think you'll need a new board for that. It's be down to the cpu. Unless I've missed some info?

I had just assumed that it would need the new chipset. Even better if not and I'll keep what I have.

 

[4K8KW10]

I had just assumed that it would need the new chipset. Even better if not and I'll keep what I have.

I also think it will depend on the CPU, not the chipset.

If thats accurate then its a fairly significant refresh this, the 2800x is not slow - if it can over clock as well to 4.5 or more then ... well intel will need to rush out another series of chips this year.

Intel already know that but they have very serious technology and process problems with the 10nm and low yields. These will eat their heads.

That FUTURE PROCESSOR in the link may be 2700, 2700X or 2800X. We don't know yet which one it is.

 

[humbug]

http://www.hwbattle.com/bbs/board.php?bo_table=hottopic&wr_id=8803

Lower latencies ram, L2, L3

Cinebench 15 Single
1800X 157
Future Processor 178

Sandra showing Turbo Speed 4,35Ghz
Not looking bad att all if real.

Yeah not bad.

For the 1800X to achieve the same score it would have to be running at:

Single: 4.525Ghz
Multi: 3.9Ghz

So if the base clock is 3.9Ghz and the boost at 4.35Ghz then IPC up lift is 0% multi and single +5%

 

[B12B6]

Yeah not bad.

For the 1800X to achieve the same score it would have to be running at:

Single: 4.525Ghz
Multi: 3.9Ghz

So if the base clock is 3.9Ghz and the boost at 4.35Ghz then IPC up lift is 0% multi and +5%

How do you go to that conclusion? CB15 Multi is 1800X@4,0 ~1700 and 4100 pretty close to 1780 +-10 and that is all cores, so propably 200mhz more on all cores than that tested no name processor + with faster ram.

 

[humbug]

How do you go to that conclusion? CB15 Multi is 1800X@4,0 ~1700 and 4100 pretty close to 1780 +-10 and that is all cores, so propably 200mhz more on all cores than that tested no name processor + with faster ram.

1645 vs 1783 is a difference of 8.1%

The 1800X all core clock is 3.6Ghz, 3.6Ghz + 8.1% = 3.9Ghz.

 

[Yaayuh!]

Worth taking into account that Cinebench benefits from L2 cache performance/latency, at least where Ryzen is concerned. This is what the CB15 performance bias does on some boards like the CH6, it tweaks the L2 cache performance.
If the latencies and/or bandwidth is better on L2 cache with the new processors, then this will show quite well in CB benchmarks but not necessarily translate elsewhere.

 

[B12B6]

1645 vs 1783 is a difference of 8.1%
The 1800X all core clock is 3.6Ghz, 3.6Ghz + 8.1% = 3.9Ghz.

But the 1800X still need 4100mhz to get 1780 score with faster memory

 

[humbug]

Yeah ^^^^ Looking at other reviews this seems about typical.

1626 vs 1782 is a difference of 9.4%

3.6Ghz + 9.4% = 3.94Ghz

So ok slightly lower scaling than the apparent leak chart so perhaps the "Future Processor" is running at about 4.1Ghz to achieve that score of 1782 or its 3.9Ghz and the IPC is up 5%.

The leak doesn't jump out at me as fake, unlike the other one, because to get those single threaded scores you would have to be running Ryzen 1 at over 4.5Ghz and that is probably unlikely.

If the Sandra boost speed is to be believed the boost is 4.35Ghz which would mean there is a 5% IPC uplift as well as those higher clocks.

We can speculate this:

3.9Ghz all cores (+8% Clocks) + 5% IPC up lift: total +13% Multi.
4.35Ghz single core XFR boost (+6% XFR) + 6% IPC up lift +12% Single

That would fit nicely.

Ninja edit

 

[Curlyriff]

Ninja edit

It is all speculation and numbers never end up like any of us show haha. Just sit and wait instead of speculate is my view these days

 

[humbug]

What i find hilarious about all this.

FX-8370, AMD's previous 8 core, now granted it doesn't have SMT and 2 core pairs share an L2 each but it is an 8 core.
Look at the performance difference at the same sort of speed, its insane. i make Ryzen 1 having about 185% higher IPC, lol!

It is all speculation and numbers never end up like any of us show haha. Just sit and wait instead of speculate is my view these days

Crap! its fun to speculate

 

[4K8KW10]

What i find hilarious about all this.

FX-8370, AMD's previous 8 core, now granted it doesn't have SMT and 2 core pairs share an L2 each but it is an 8 core.
Look at the performance difference at the same sort of speed, its insane. i make Ryzen 1 having 186% higher IPC, lol!

Argh!
That FX-8370 behaves like a quad core with poor clustered multi-threading.

 

[humbug]

Argh!
That FX-8370 behaves like a quad core with poor clustered multi-threading.

The FX series CPU's are changeable depending on what you want from them, how you program your execution for them.

By design they have 8 Integer units (Cores) and 8 128Bit FP Units or they can be combined to form 4 256Bit FP units.

So depending on whats asked of it its either an 8 core 4 thread CPU or an 8 core 8 thread CPU, Cinebench treats it as an 8 core 8 thread CPU.

 

[DarkHorizon472]

My guess is with good cooling the maximum for Ryzen 2 on a good chip will be 4.4ghz overclocked. Which is a good enough result, any more than that and Intel really has a problem.

I get about 1650 in cinebench.

 

[Curlyriff]

Crap! its fun to speculate

I used too and then I realised we are on OCUK

 

[humbug]

I used too and then I realised we are on OCUK

lol...

 

[4K8KW10]

The FX series CPU's are changeable depending on what you want from them, how you program your execution for them.

By design they have 8 Integer units (Cores) and 8 128Bit FP Units or they can be combined to form 4 256Bit FP units.

So depending on whats asked of it its either an 8 core 4 thread CPU or an 8 core 8 thread CPU, Cinebench treats it as an 8 core 8 thread CPU.

4 module 4 thread or 4 module 8 thread. But in each case the definition for a thread differs.
A module is a dual-core but with some parts of the cores cut and shared among them.

 

[humbug]

4 module 4 thread or 4 module 8 thread. But in each case the definition for a thread differs.
A module is a dual-core but with some parts of the cores cut and shared among them.

Yes, L2, i did say that in the original post you quoted.

Lets not get hung up on this.

 

[Killer7xx]

Just shows how much Intel are relying on their process technology. 12nm LP from GloFo is just an improved version of 14nm LPP, probably barely better than the 14nm process that Intel used for Skylake. If AMD had access to the 14nm++ process Intel are currently using, Zen would be reaching 5 GHz.

It's gonna be fun to see AMD compete with Intel next year using equal process technology. I'm expecting that with Zen 2, they'll bring at least 10-15% IPC over Zen and we should see 5 GHz overclocks being quite common.