Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[lolfail9001]

Is it confirmed that asus crosshair IV hero is the best oc board?

VI, actually.

And yes, it is.

 

[CentroX]

VI, actually.

And yes, it is.

Whohoo. Ill try and get my 1800X to 4.3 on water this friday.

 

[lolfail9001]

Whohoo. Ill try and get my 1800X to 4.3 on water this friday.

Rest in Piece CentroX's house.

 

[Karnak]

P. S. XFR was totally working, unless you want to claim CPU-Z 1.78.3 lies.

And you wan't to claim the reviewers are lying?

In this study, the prototype (known as engineering samples) of a motherboard we have still not quite ready. It was not yet ready motherboard BIOS and several limitations, including in support of XMP and overclocking configurations there. However, these problems will be resolved once the final version of the product.

 

[Atari2600]

Talking about goalpost moving, i should probably compile list of reactions to Ryzen memory latency.

It'll be interesting to see what the effect of the latency hit is.

It will also be interesting to see how much more bandwidth efficient (if any) Ryzen is vs. the Intel IMC. With the leaks so far pointing toward massive or minor differences (in bandwidth), we've no real idea where it lies never mind the effects of the tradeoff.

 

[lolfail9001]

And you wan't to claim the reviewers are lying?

Where is "XFR" mentioned in this sentence? Also, XFR is disconnected from mobo BIOS. And it is not really "overclocking", if we are real.

 

[alexruiz]

I think that the more leaked "reviews" we see, the better Ryzen looks!
The detractors all just cling to an specific absolute number and claim "it couldn't beat the blue chip company, it sucks!" without even analyzing the numbers and how the numbers were generated.
The Iranian review, in my eyes, even for final silicon would paint Ryzen in a good light. It is right there with the competition.
If we consider that is was a ES chip on old BIOS running DDR4-2133 makes it more impressive.

For the cooler, that Notus 200 looks a lot like a knock off of the DeepCool Gammaxx 200. The Gamaxx 200 is rated at 95W only. Overclocking the chip would get get consumption higher, it will thermally throttle.

 

[lolfail9001]

It will also be interesting to see how much more bandwidth efficient (if any) Ryzen is vs. the Intel IMC

Few pages ago, around 2 gigs/sec more than Skylake at similar memory timings and same frequency.

I mean, make no mistake, i want to believe that if XFR was enabled it was not working at all (together with boost), just to make Ryzen look better and CPU-Z was acting up. But we will learn it all in ~24 hours, won't we?

 

[french toast]

I know lets magnify in on latency sub tests and fritzchess, obviously if it fails there thats a deal breaker for sure.

 

[CentroX]

Cpu shortcomings/bugs can be patched via microcode right?

 

[lolfail9001]

I know lets magnify in on latency sub tests and fritzchess, obviously if it fails there thats a deal breaker for sure.

It is a dealsalter for me, because my workload is memory latency sensitive and may in fact be similar to fritz (conceptually it is not that different)

Also, there's the whole deal with power consumption in P95 shooting up from ~120W power delta at stock (accounting for PSU and VRM, it can be stretched to fit 95W TDP) to over 200W at just 4Ghz. That thing actually hits heat density issues at just 4Ghz, that's kind of insane. And spells big trouble for my dreams of ITX Ryzen build running at 3.7Ghz all core.

 

[scannall]

I know lets magnify in on latency sub tests and fritzchess, obviously if it fails there thats a deal breaker for sure.

Some of the complaints are just crazy. Gee, it misses on an obscure and mostly useless Fritch thing. Guess I better run out and spend 600 or 700 more on an Intel setup. Nah, don't think so. I'd rather add another GTX 1080 to the system instead.

 

[tamz_msc]

Cpu shortcomings/bugs can be patched via microcode right?

Bugs yes, shortcomings no. For example the AIDA L3 writes on Ryzen is low compared to Skylake, which might be due to it being victim instead of write-back. Microcode patches won't change that.

 

[richierich1212]

We should not post for 24 more hours since official reviews are not out yet. Let's also not post about coffeelake and Skylake-X since those reviews aren't out yet as well. Let's just not post at all. Keep your damn enthusiasm for PCs to yourself since it's a dead market anyways.

 

[tamz_msc]

It is a dealsalter for me, because my workload is memory latency sensitive and may in fact be similar to fritz (conceptually it is not that different)

Also, there's the whole deal with power consumption in P95 shooting up from ~120W power delta at stock (accounting for PSU and VRM, it can be stretched to fit 95W TDP) to over 200W at just 4Ghz. That thing actually hits heat density issues at just 4Ghz, that's kind of insane. And spells big trouble for my dreams of ITX Ryzen build running at 3.7Ghz all core.

If it's not for you get Intel then.

You do realize P95 uses AVX right? Who runs it on an ITX system for hours just to see how high the cores boost?

 

[Atari2600]

It is a dealsalter for me, because my workload is memory latency sensitive and may in fact be similar to fritz (conceptually it is not that different)

What is your workload?

 

[lolfail9001]

You do realize P95 uses AVX right?

You do realize that in Ryzen case, Prime95 only consumes a smidge more than Cinebench, because of design trade-off? Basically the P95 consumtion is a good "worst-case" ballpark for any 100% CPU usage scenario on Ryzen.

And i compile software on my PC, too!

What is your workload?

Symbolic execution. Different thing on paper, but concept of choosing state to run is not dissimilar to what Fritz does, not to mention that state sets naturally can get fairly titanic in sizes (hell, a sample test can run in billions already) and you naturally have to jump from one end of the set to another. On each iteration.

 

[lobz]

What is your workload?

Whatever that may be, since last week it consists of fritzchess and the passmark's prime numbers and physics. I really think he should get a Nehalem with first gen ddr, it had super low memory latency. No need fooling around with anything else

 

[unseenmorbidity]

VI, actually.

And yes, it is.

Where was it confirmed?

 

[AtenRa]

AMD's pricing is really the star of this show -- they sensed that enthusiasts wanted "more cores" and wanted those cores to be decent and not priced to the moon, and they appear to have delivered.

What's surprising to me is how Intel has not yet cut prices on Broadwell-E. They have to know that nobody is going to want to buy those now, except maybe the top 10 core.

1, HEDT is only 1% or even lower
2, With KabyLake release in January nobody would go for X99
3, Everyone knew Ryzen was launching in Q1
4, Everyone knows SkyLake-X is coming in Q3

1+2+3+4 = Even without Ryzen, you would be stupid to even consider X99 today.
And Intel doesnt even have to sweat, they had Broadwell-E in the market for almost a year now, they made up more than what they spend for R&D. No need to lower prices now, they would sell less BD-E anyway, better sell less at those huge margins than cannibalize KabyLake sales with price cuts

 

[tamz_msc]

You do realize that in Ryzen case, Prime95 only consumes a smidge more than Cinebench, because of design trade-off? Basically the P95 consumtion is a good "worst-case" ballpark for any 100% CPU usage scenario on Ryzen.

And i compile software on my PC, too!


Symbolic execution. Different thing on paper, but concept of choosing state to run is not dissimilar to what Fritz does, not to mention that state sets naturally can get fairly titanic in sizes (hell, a sample test can run in billions already) and you naturally have to jump from one end of the set to another. On each iteration.

Fritz scales almost linearly with frequency, so you're better off with higher clock speeds.

 

[lolfail9001]

Fritz scales almost linearly with frequency, so you're better off with higher clock speeds.

Of course it scales linearly with frequency when you are comparing identical uarches, man.

 

[PhonakV30]

Performance is ok , still above 40% IPC also Power consumption is pretty damn good.lower than 7700K.

 

[french toast]

Fritz scales almost linearly with frequency, so you're better off with higher clock speeds.

same uarch though

Damn beat me to it

 

[lolfail9001]

Where was it confirmed?

1. The only board Gibbo pushed 1700 to 4Ghz. Granted, these folks managed to do it on B350, but with 1700X.
2. That buildzoid's posts, only approving of Crosshair preorders, hints that if it is not The Best, it is at least One Of The Best.