Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[scannall]

Impressive but too little too late. Hopefully the release drives down the I7 prices so I can upgrade!

It's very unlikely that Intel will lower prices, particularly on their more expensive SKU's.

 

[AtenRa]

These are all CPU only. For CPU+iGPU (APU) you'll have to wait till mid year until Raven Ridge arrives. Raven Ridge's iGPU is Vega based, so it'll have the latest and greatest video capabilites on AMD products.

Both products use the same AM4 motherboards, you'd just need one with video outputs.

Yea be careful people because some High-End AM4 motherboards only support CPU only RYZEN.

 

[CatMerc]

If you can clock things just right, absolutely.

Phenom II had the same type of behavior. My X3 @ 3.2GHz was faster than it should have been when coming from 2.8Ghz - not massively, mind you. I think Ryzen has a LOT of independently clocked areas. The command fabric, data fabric, I/O plane, memory controllers, PCI-e, GMI, L3, CCX common pathways, each core, etc... ALL seem to be on different frequencies. Most likely they are mostly just referenced to the bus speed (hopefully except PCI-e and I/O, which should always be at a fixed speed).

Craziest thing is that it looks (from the die shot) that the L2 caches may be on a different frequency than the cores (may be on the same as the L3 while still being able to be disabled...). That can have all kinds of effects on performance as frequency increases.

How can you tell?

 

[Tup3x]

Yea be careful people because some High-End AM4 motherboards only support CPU only RYZEN.

Those do (usually) have USB-C though so shouldn't it be possible to get displayport out from it?

 

[zinfamous]

Hmmm... where do you live? Mind you: I've never traveled more than about 5,000 miles for a customer.

I have a creepy house in a creepy neighborhood north of DC. It actually has a super creepy basement, too (no joke).

....So, fair warning.

 

[AtenRa]

Those do (usually) have USB-C though so shouldn't it be possible to get displayport out from it?

They dont support other CPUs only R3-5-7 Ryzen, they dont support any APUs. Even if they have USB-C, APUs will not work on those Motherboards.

 

[.vodka]

They dont support other CPUs only R3-5-7 Ryzen, they dont support any APUs. Even if they have USB-C, APUs will not work on those Motherboards.

What makes you think so?

AM4 is a unified socket, you should be able to install an APU in an AM4 motherboard without video outputs yet be able to use it as a pure CPU just fine.

FM2 had specific CPUs with the iGPU fused off, but I don't think there were motherboards without video outputs to test this.

 

[AtenRa]

What makes you think so?

AM4 is a unified socket, you should be able to install an APU in an AM4 motherboard without video outputs yet be able to use it as a pure CPU just fine.

FM2 had specific CPUs with the iGPU fused off, but I don't think there were motherboards without video outputs to test this.

It was said by Motherboard makers (unless I got something wrong).

 

[scannall]

It was said by Motherboard makers (unless I got something wrong).

Probably a bios thing though. Something that will be addressed down the road.

 

[Insert_Nickname]

Yea be careful people because some High-End AM4 motherboards only support CPU only RYZEN.

They support APUs just fine. You're just not able to use the APUs video outputs.

There are a few PCIe and SATA quirks to watch out for however.

 

[looncraz]

How can you tell?

From the power gating arrangements and the demarcation between the L2 and the core and the L3 and the L2 (which is harder to see, but it's there). Looks like an interface to adapt to different frequencies.

And from the numbered slides here: http://files.looncraz.net/zen/

Especially this one:

L2 & L3 have their own voltage domain separate from the cores:

 

[AtenRa]

Ok example,

Asrock Fatal1ty X370 Professional Gaming and X370 TaiChi only support CPU only Ryzen

Asrock Fatal1ty X370 Gaming K4 and X370 Killer SLI supports both Ryzen and APUs (BristolRidge)

 

[burninatortech4]

Still waiting on high end mATX options. The few I've seen either don't have VRM heatsinks or have a weak VRM in the first place.

 

[formulav8]

AMD Ryzen Initial Shipments to Eclipse One Million

 

[looncraz]

I have a creepy house in a creepy neighborhood north of DC. It actually has a super creepy basement, too (no joke).

....So, fair warning.

Ewww... D.C..

I don't plan on going that way for as long as possible.

 

[bjt2]

From the power gating arrangements and the demarcation between the L2 and the core and the L3 and the L2 (which is harder to see, but it's there). Looks like an interface to adapt to different frequencies.

And from the numbered slides here: http://files.looncraz.net/zen/

Especially this one:

L2 & L3 have their own voltage domain separate from the cores:

According to the japanese guy, the separate voltage domains are to support some tricks that allow to use lower Vdd to power the CPU. Usually there is a minimum voltage to use with L2 to maintain content. But if you want to allow a very low power state of the CPU still retaining the cache (the famous Pmin state), you can't go under that value, so the leakage is higher than necessary. With separate Vdd you can power the various parts with the necessary Vdd. And with the DLDO you don't even need separate power rails, because the LDO will provide the necessary voltages...

 

[looncraz]

According to the japanese guy, the separate voltage domains are to support some tricks that allow to use lower Vdd to power the CPU. Usually there is a minimum voltage to use with L2 to maintain content. But if you want to allow a very low power state of the CPU still retaining the cache (the famous Pmin state), you can't go under that value, so the leakage is higher than necessary. With separate Vdd you can power the various parts with the necessary Vdd. And with the DLDO you don't even need separate power rails, because the LDO will provide the necessary voltages...

Yes, this is no doubt the case. But then how do you determine what frequency you will use for the L3? L2 is easier - likely running at core clock, though not necessarily. The L3 will need to run at some other frequency. It may just match the highest clocked core, but I think it has a fixed multiplier based on the bus speed.

That should mean that bus-based overclocking may result in better scaling that frequency-based overclocking.

 

[biostud]

http://wccftech.com/amd-ryzen-1700x-benchmarks-leaked-beats-kaby-lake-ipc/

wccftech though....

 

[unseenmorbidity]

http://wccftech.com/amd-ryzen-1700x-benchmarks-leaked-beats-kaby-lake-ipc/

wccftech though....

userbench + wccftech = negative credibility.

 

[looncraz]

http://wccftech.com/amd-ryzen-1700x-benchmarks-leaked-beats-kaby-lake-ipc/

wccftech though....

Yeah, the data actually shows a 6% IPC deficit against Kaby Lake... which, frankly, is a miraculous result from AMD.

 

[DrMrLordX]

My prediction for q1 and q2 is the bean counters and amd - margin - cfo will discover the desktop market is not a fixed and ever declining market.

They already know about that. Much of the market decline after '06 came straight out of their hide. But not all of it.

Water cooling is for plebians. My heatsink block is the size of a child's head.

If they would sell something like that to me, I would probably buy it. Assuming it had enough heatpipes to work. Hell I would have to buy an entire case around a heatsink like that, but it would be awesome.

What makes you think so?

AM4 is a unified socket, you should be able to install an APU in an AM4 motherboard without video outputs yet be able to use it as a pure CPU just fine.

FM2 had specific CPUs with the iGPU fused off, but I don't think there were motherboards without video outputs to test this.

FM2/FM2+ had outputs on everything, correct. Regardless an APU should work in any motherboard with a UEFI that supports it. You should be able to use the iGPU for compute functions if nothing else.

Display output via USB-C might even work.

AMD Ryzen Initial Shipments to Eclipse One Million

Wow, I was not expecting them to move that many units. Their pricing strategy is going to get them serious revenue. Remember that's 1 million units of 1700/1700X/1800X and the Pro CPUs!

 

[bjt2]

Yes, this is no doubt the case. But then how do you determine what frequency you will use for the L3? L2 is easier - likely running at core clock, though not necessarily. The L3 will need to run at some other frequency. It may just match the highest clocked core, but I think it has a fixed multiplier based on the bus speed.

That should mean that bus-based overclocking may result in better scaling that frequency-based overclocking.

Yes. The L3 probabily clocks as the NB, like INTEL and BD. AMD promised 5x BW for the L3. The bus is 4X and the clock should be +25%. But what was the clock on BD? The NB clock, that was 2.4GHz. 2.4+25%=3GHz. On BD the NB should go at least at half frequency of the fastest core so if for Zen is the same, up to 6GHz the NB can also stay at default. But obviously the scaling will not be linear if you raise only CPU clocks...

 

[unseenmorbidity]

Looks like Gibbo did something wrong.

Gigabyte motherboard going above 3000MHz

http://i.imgur.com/nR4nIIl.png

 

[Ryzenenthusiast]

Note it somewhere;

Editors of the one of the most influential hardware website of Turkey who benchmarks 1800x states that "Overclock is no problem at all. Though we can not say a certain frequency due to NDA, I can say that it is impressive.". Also in somewhere else of their livestream one of the guys stated that "In some benchmarks AMD(!). It seems ironical yes but AMD is presenting a CPU performance that Intel can not keep up with even with their 10 core 6950x!"

 

[.vodka]

Note it somewhere;

Editors of the one of the most influential hardware website of Turkey who benchmarks 1800x states that "Overclock is no problem at all. Though we can not say a certain frequency due to NDA, I can say that it is impressive.". Also in somewhere else of their livestream one of the guys stated that "In some benchmarks AMD(!). It seems ironical yes but AMD is presenting a CPU performance that Intel can not keep up with even with their 10 core 6950x!"

Nice! Could you post a link to where you got that?

It supports the idea that the R7 1700 is made up of lower quality dies that obviously couldn't fit the targeted clockspeeds/voltage and 95w bin of the R7 1700x and 1800x, and as such runs perfectly well underclocked at 3.0/3.7GHz to fit the 65w bin.

This also supports the rumor/report that being lower quality dies, obviously requiring higher voltages to get to 4 GHz, 4GHZ R7 1700 were cooking the VRMs of lower end boards but not the monster VRM of the Crosshair VI Hero.

If true, the $70 extra for the R7 1700x is warranted if it can easily clock past 4GHz.

Looks like Gibbo did something wrong.

Gigabyte motherboard going above 3000MHz

http://i.imgur.com/nR4nIIl.png

Excellent.

Man I can't wait for actual reviews.