Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[GroundZero7]

AMD rolled the memory controller, but the PHY is supposedly from RAMBUS (physical interface circuitry).

It appears that the PHY, the IMC, and the infinity fabric are all up to snuff - it would take all three to work nearly perfectly for the results we are seeing.

If that 99% memory efficiency figure is right there is some serious secret sauce
AMD put in there.

 

[looncraz]

Zen overclocks to 11. niiiiiiiice

It's not a very good thing that 4.4Ghz is considered "Extreme" overclocking.

But, it makes sense, that's > 150W @ 60C in a 192mm^2 die... not even water cooling will keep it at 60C, so it will likely be 70C @ 150W, which means the frequency then falls to 4.25~4.3GHz... on water... or power increases to 180W+ ... which you simply aren't going to manage to pull out of a small die like Ryzen very easily.

Going to six core changes things only slightly - it removes about 20W. Going to quad core (one CCX), though, will save a bit extra power but increase power density, which is bad for clocks. 2+2 will use more power but spread out the heat, making it easier to cool, and likely leading to the highest overclocks (and 4MB of L3 per core...).

 

[PhonakV30]

ok Here Link for Ryzen 1800X
Memory bandwidth : http://ranker.sisoftware.net/show_r...e4dceadceddae3c5b78aba9cf99ca191b7c4f9c9&l=en

It's possible ? that's insane efficiency.
Integer Memory Bandwidth : 33.94GB/s
Float Memory Bandwidth : 34.05GB/s

LOL! It's almost 100%!! If true , then You don't need DDR4 3000

 

[Glo.]

ok Here Link for Ryzen 1800X
Memory bandwidth : http://ranker.sisoftware.net/show_r...e4dceadceddae3c5b78aba9cf99ca191b7c4f9c9&l=en

It's possible ? that's insane efficiency.
Integer Memory Bandwidth : 33.94GB/s
Float Memory Bandwidth : 34.05GB/s

LOL! It's almost 100%!! If true , then You don't need DDR4 3000

You do, but the effects will be insane for gaming .

 

[.vodka]

Damn, Sandra tells the same as the rumor! I can now see why AMD decided to go dual channel with this,

DDR4 3000 dual channel = 48GB/s
DDR4 3200 as ASUS supports out of the box DC = 51.2GB/s
DDR4 3600 as rumored DC = 57.6GB/s

even higher when BIOSes are ready and done a few months down the line? That's Intel quad channel HEDT bandwidth before memory overclocking, holy crap


So AMD fixed their caches, fixed their memory subsystem... I can see Naples doing just fine in this department with each die having its two memory channels... hope the GMI links can keep up

 

[GroundZero7]

ok Here Link for Ryzen 1800X
Memory bandwidth : http://ranker.sisoftware.net/show_r...e4dceadceddae3c5b78aba9cf99ca191b7c4f9c9&l=en

It's possible ? that's insane efficiency.
Integer Memory Bandwidth : 33.94GB/s
Float Memory Bandwidth : 34.05GB/s

LOL! It's almost 100%!! If true , then You don't need DDR4 3000

Good point..I wonder if those CL10 2400 sticks will bottleneck.

 

[inf64]

Skylake-X should help them in three ways:

1. Higher IPC than Broadwell-E (probably bigger gap between BDW-E and SKX-X than there is between BDW-E and SKL-S/KBL-S here, due to changes in cache sizes/bandwidth).

2. Skylake has a better SMT implementation than Broadwell/Haswell do, so SMT yield should improve, boosting MT competitiveness for a given core/frequency count.

3. Skylake-X should clock a lot better than Broadwell-E due to both process improvements as well as a much better physical implementation (Broadwell-E's was quite poor).

That is the ~15-20% I mentioned in my post. That should be aggregate perf. improvement over 6900K. Still that doesn't make SKL-X much better value proposition than 6900K because perf./$ ratio will still be vastly better on AMD side. AMD made a "quantum" IPC leap with Zen and it seems they are going to do yearly IPC bumps on this core which spells bad news for intel. Their next core that will give us some measurable IPC jump is ~2 years away, just like 2 next iterations of Zen are.

 

[PhonakV30]

Good point..I wonder if those CL10 2400 sticks will bottleneck.

Weird... HOLD ON! I see this :
AMD FCH; 16GB Corsair CMK16GX4M2B3000C15 DIMM PC24000

It's DDR4 3000 : This link but if you look at this link , It's running at 2133Mhz.

 

[tamz_msc]

Nearly every Sandra screenshot I've seen has 3000MHz rated memory run at 2133MHz CL16 or CL17. Early BIOSes maybe?

 

[GroundZero7]

Weird... HOLD ON! I see this :
AMD FCH; 16GB Corsair CMK16GX4M2B3000C15 DIMM PC24000

It's DDR4 3000 : This link but if you look at this link , It's running at 2133Mhz.

No I'm talking about the supposed +25% thouroughput efficiency of the Ryzen DDR4 controller.

2400 should perform close to 3000. With CL10 it could be a very good combination. Lower latency than 3200 CL14. If it doesn't bottleneck it would be great.

 

[Dresdenboy]

If those are all of the options, then it seems that the rumors of needing balanced CCXes is true.

Still, I would have liked to be able to select WHICH of the cores in a CCX are disabled - so I could find my worst two and turn them off for testing

Missing options actually don't imply a need. They might just be worse than those presented, as an unbalanced configuration might simply mean, that threads, which happen to run in one CCX, might have to share L3 and all the CCX interfaces' bandwidths, while that single or two threads in the other CCX are happy with 8MB L3 on their own. This might just cause unbalanced performance.

 

[otinane]

http://ranker.sisoftware.net/show_r...e4dceadceddae3c5b78aba9cf99ca191b7c4f9c9&l=en

Does that mean:

A)Ryzen IMC = topkek

B)Quad channel is only for weak Intel cpu

C)Bandwidth/Latency > Speed for Ryzen

 

[DrMrLordX]

Not seeing any indication that Ryzen has onboard graphics so this launch is kind of meh for me. Downsides of having a slim case.

Wait for Raven Ridge. I was almost in the same boat as you . . . had AMD launched Raven Ridge at the same time as Ryzen, I'd be buying Raven Ridge.

That is some pretty freaking insane efficiency if true!

Who is providing the memory controller for AMD here? Rambus? Infinity fabric at work....?

Um, RAMBUS? Which is sad, since I still sort of hate those guys. Sort of. Surprise surprise they actually made something worth a darn for once.

AMD rolled the memory controller, but the PHY is supposedly from RAMBUS (physical interface circuitry).

It appears that the PHY, the IMC, and the infinity fabric are all up to snuff - it would take all three to work nearly perfectly for the results we are seeing.

Oh okay, so the IMC is in-house. Sort of makes sense.

Smart too, AMD really had to nail it on the IMC if they were going to support an 8c/16t CPU with dual-channel configs. That sort of protects them from silly OEMs that will bundle Ryzen with DDR4-2133 and the like.

http://ranker.sisoftware.net/show_r...e4dceadceddae3c5b78aba9cf99ca191b7c4f9c9&l=en

Does that mean:

A)Ryzen IMC = topkek

Possibly . . .

B)Quad channel is only for weak Intel cpu

Raven Ridge would still benefit from that. As would any other APU.

C)Bandwidth/Latency > Speed for Ryzen

Ehh that's vague. Remember DDR4 ratings can improve both bandwidth and latency at higher values, assuming timings remain unchanged.

 

[DrMrLordX]

Yeah, I know, I was just showing you the biggest heatsinks I've ever seen Wonder what the result would be if they stuck some fins on it, and added a fan mount or three. Those mini heatpipes are interesting, though. wonder how effective they are, given the size and not having wicks.

At that point I think they'd be better off just doing a redesign. Something with lanyards or what have you to take weight off the mobo.

Oops technically not a double-post but I coulda fit this response into the above reply . . . oh well.

 

[itsmydamnation]

Um, RAMBUS? Which is sad, since I still sort of hate those guys. Sort of. Surprise surprise they actually made something worth a darn for once.

They always did thats what made them such good patient trolls. RDRAM came out in 96. DDR SDRAM was like 2000.

 

[DrMrLordX]

Eh don't get me started on RDRAM.

 

[Teizo]

I am a bit suspect as to why AMD is sending Noctua coolers with an Industrial NF-F12 2000rpm fan with their press kit. Are these things going to run hot? I figured they would send their Wraith Cooler.

https://www.youtube.com/watch?v=JaTvSI0K_vA

https://www.youtube.com/watch?v=rF2phtH8v2g

On a side note...All black Noctua fans ftw.

 

[Puffnstuff]

I find it refreshing that AMD has finally produced a CPU worthy of notice. We haven't seen one since the original 1ghz Athlon hit the scene back in the late 90's. I did run an Opteron 170 until C2D hit the market and then I switched back to Intel.

 

[french toast]

I am a bit suspect as to why AMD is sending Noctua coolers with an Industrial NF-F12 2000rpm fan with their press kit. Are these things going to run hot? I figured they would send their Wraith Cooler.

https://www.youtube.com/watch?v=JaTvSI0K_vA

https://www.youtube.com/watch?v=rF2phtH8v2g

On a side note...All black Noctua fans ftw.

You do make a fair and valid point, i feel they also should be using stock coolers, on the other hand it could be argued they are getting their sh*t together and giving the best impression of themselfs, unlike in the past.

 

[KTE]

INTEL has had a settlement with AMD and the national authority (I quite remember a settlement also in Europe) and promised to do not repeat those actions and in return it obtained a lower punishment. But if INTEL repeats the crime, it will be punished twice.

You can't deny it, because it was punished by at lest USA and Europe. I don't remember the fees, but they were reduced because in exchange INTEL settled with AMD with more than 1 billion dollar and the relaxation of the terms on the x86 license, more notabily the obligation to own a FAB.

What's that got to do with what I said?

You're just repeating what we all know.

Sent from HTC 10
(Opinions are own)

 

[Teizo]

You do make a fair and valid point, i feel they also should be using stock coolers, on the other hand it could be argued they are getting their sh*t together and giving the best impression of themselfs, unlike in the past.

I suspect it probably has something to with the fact that they are not telling the truth about TDP. It is not 65 watt and 95 watt. If you notice, the new Wraith coolers are designed for 95w TDP and 140w TDP.

So, AMD sent a high quality cooler and a high rpm industrial fan to make sure the chips are cooled adequately to see XFR kick in. I'm not sure how much headroom there is really going to be beyond that frequency wise. Otherwise they would have sent their new Wraith coolers to show off as well.

We will see Tuesday, but that did raise a red flag and grab my attention.

 

[agouraki]

I suspect it probably has something to with the fact that they are not telling the truth about TDP. It is not 65 watt and 95 watt. If you notice, the new Wraith coolers are designed for 95w TDP and 140w TDP.

We will see Tuesday, but that did raise a red flag and grab my attention.

they most likely have to count in XFR and have some headroom....giving 95tdp cooler on same watt cpu might throttle it.

 

[french toast]

I suspect it probably has something to with the fact that they are not telling the truth about TDP. It is not 65 watt and 95 watt. If you notice, the new Wraith coolers are designed for 95w TDP and 140w TDP.

So, AMD sent a high quality cooler and a high rpm industrial fan to make sure the chips are cooled adequately to see XFR kick in. I'm not sure how much headroom there is really going to be beyond that frequency wise. Otherwise they would have sent their new Wraith coolers to show off as well.

We will see Tuesday, but that did raise a red flag and grab my attention.

Certainly a fair question, to be fair the wraith coolers are probably designed for xfr and some overclocking headroom, the power measurements AMD showed did show R7 1700 65w chip consuming considerably less power than the i7 6900k, which was worked out to be right inline with 65w, that was with a wraith cooler as with all benchmark results amd demonstrated.

Your right though we need to see reviews.

 

[cytg111]

though overvolted to a degree at the horizon demo it did draw the same as the 6900k under load - that is a 140watt part IIRC.

 

[bjt2]

What's that got to do with what I said?

You're just repeating what we all know.

Sent from HTC 10
(Opinions are own)

I should have quoted only the first sentences...
I was saying that INTEL was already convicted in the past for bribing or better, blackmailing IT companies to not offer or delay AMD products.
See, if i know that a person was convicted for pedophilia, i would rather not leave him alone with my childs...
In other words, i don't trust INTEL. This is matter of shareholder, stock values etc... They can't permit AMD to gain market. Especially because they have not a new product to counteract...