Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[looncraz]

I don't know but the OS has to play a part in this. For example, in Linux, I've found that when you have a few applications running, mostly doing light to moderate tasks - like for example watching a video while say editing a large LaTeX document, compiling the tex file from time to time shows all sorts of random behavior in the logical core utilization on an i7 3770. At any time any random core will show maximum utilization. On Windows it's mostly the first 2 cores that do the heavy lifting.

So my point is, if you are using the OC utility to boost two cores, then the OS must also know how set core affinity for the task it is currently executing, without user intervention of course.

The scheduler will purposefully try to keep threads scheduled to as few cores as possible for power savings and for allowing the CPU to turbo to its faster turbo speeds.. that's relatively simple from a kernel scheduler perspective.

What we need now is a way to set which cores are favored first. That would probably be the easiest way forward.

 

[Maverick177]

The usual suspects migrated to [H], no wonder the threads are less dull.

 

[WhoBeDaPlaya]

Interesting, in Windows 7 I would get blue screens just from switching bios setting of IDE mode into AHCI without first tweaking registry values.

Huh, I forget whether the inaccessible boot device message was on a blue screen or not.
In any event, that's the only snag I've encountered moving the same install to different rigs (test install on SSD, I'd normally prefer a clean install).

 

[nathanddrews]

Microcenter doing some strong price cuts on Intel CPU's.

 

[IEC]

Win, the consumer shall

 

[looncraz]

Not really, those prices are mostly just going back to what they were.

 

[nathanddrews]

Not really, those prices are mostly just going back to what they were.

Not when you factor in MC's bundle pricing.

 

[DrMrLordX]

How about a Nofan Icepipe CR-100A? Although that's more along the lines of an adult head, I'd say. The CR-95 might be more child-head-like in it's dimensions. Oh, and enough heatpipes? Read the name. It's made out of heatpipes. Tiny, wickless ones, but still heatpipes.

Based on its weight, I'd say no. Too much dead space in that design. I would like an enormous head-sized, high-fin-density rad with um, noisy fans? That thing is designed to be silent, and it actually weighs less than a D15. Ryzen is going to put out 200W+ for some users, so . . .

You can get an entire case that is a heatsink, like the HDPLEX H5 or Streacom DB4. I'm planning the H5 for my next build, because it can do any size board up to full ATX and also passively cool a GPU (up to 95W) with an extra heatpipe attachment. I'm hoping for ECC-supporting Ryzen chips soon (maybe these are the "PRO" models on that leaked chart?). For the GPU I'm planning a Radeon Pro WX 5100, which is the strongest one currently available that fits into the recommended TDP.

I haven't looked at those seriously, but it is an interesting concept . . .

 

[cytg111]

The CPU has been doing the work, responding to load on a given core, and recognizing when one core is loaded and the others are not, then boosting the clock on the loaded core.

What we need now is a way to choose which cores can overclock the most and create a few highly clocked cores and a few lowly clocked cores (akin to big.LITTLE architecture, except only using frequency).

I'll just use a six-core as an example:

Core #: Max turbo clock setting, for maximized performance

Core 0: 3.0Ghz
Core 1: 3.0Ghz
Core 2: 5.0Ghz <- favor heavy tasks here
Core 3: 3.0Ghz
Core 4: 5.0Ghz <- favor heavy tasks here
Core 5: 3.0Ghz

Gotcha, I didnt know that it was the CPU that responded to the load in that way, I figured it was the scheduler that assigned a high priority thread to the highest performing core.

 

[OrangeKhrush]

AMD accused of testing Intel CPU's with stock coolers, because they tested their own with stock coolers. It is AMD's fault that Intel coolers are so inefficient.

 

[french toast]

AMD accused of testing Intel CPU's with stock coolers, because they tested their own with stock coolers. It is AMD's fault that Intel coolers are so inefficient.

They cant win, even if they tested the intel rig from the north pole with chinook hovering above they would find away to complain.

 

[OrangeKhrush]

So, like apparently Cinebench is no longer a representative benchmark, this happened between all the big talking at "new horizons" and changed after the 22/2 NDA lift, it is amazing how people can change their tune.


OKAY PEOPLE LETS MOVE THESE GOAL POSTS OVER THERE.

 

[OrangeKhrush]

They cant win, even if they tested the intel rig from the north pole with chinook hovering above they would find away to complain.

That wouldn't keep an X99 CPU cool

 

[formulav8]

Hopefully this isn't true and won't happen again with Intel.

 

[DrMrLordX]

Hopefully this isn't true and won't happen again with Intel.

Yeouch. Now I'm reminded of why I buy only AMD desktop products (okay I didn't buy their RAM and SSDs, but still).

 

[OrangeKhrush]

Hopefully this isn't true and won't happen again with Intel.

Just about every big wig District Attorney would love to have their day in Court with Intel, Intel can only go so far, any breach of anti trust again and the law suits will mount.

 

[railven]

Not really, those prices are mostly just going back to what they were.

Yeah someone post that on another thread and got excited for a price war. Seeing they used Microcenter as their source, I instantly got deflated.

Hell, Microcenter had better deals in the past ($50 bundle, the i7-4c/4t was normally $250-300, that's where the i5-4c/4t now sits).

Here is hoping a price war does start, but that article is click-bait as hell.

 

[looncraz]

Gotcha, I didnt know that it was the CPU that responded to the load in that way, I figured it was the scheduler that assigned a high priority thread to the highest performing core.

To be fair, I oversimplified it

The OS sets P-states, parks cores, schedules the threads on the cores in accordance to policy, and so-on. But the actual response to those events is determined by the CPU.

 

[OrangeKhrush]

Old habits die hardest, Intel back to doing what they do best.

 

[jpiniero]

Yeah someone post that on another thread and got excited for a price war. Seeing they used Microcenter as their source, I instantly got deflated.

Hell, Microcenter had better deals in the past ($50 bundle, the i7-4c/4t was normally $250-300, that's where the i5-4c/4t now sits).

Here is hoping a price war does start, but that article is click-bait as hell.

$259 for the 6700K with the $30 mobo discount is unusual.

 

[mikk]

i think instead of cinebench favoring amd it "lucked" out and ryzen favored cinebench...

Based on the leaked Ryzen benchmarks this is not the case. Most likely Cinebench is very close to median, as expected. That was the case for Bulldozer as well.

 

[railven]

$259 for the 6700K with the $30 mobo discount is unusual.

Not from Microcenter. Best bang for buck for Intel was always Microcenter.

Hell, I remember I got my 4790K for $250 with $40 off bundle...wonder if it's in my order history. The motherboard bundle use to be $50 and it slowly got smaller as the actual price of CPUs get pricier too.

Wonder if it's my order history...

EDIT: There it is:

Placed: 12/26/2014 4:18:11 PM
Recipient:
Order Type: In Store
Intel Core i7-4790K Devil's Canyon 4.0 GHz LGA 1150 Boxed Processor

• SKU: 302273
• Condition: New
• Status: Completed
• Transaction #: <redacted>

• Quantity: 1
• Price Each: $249.99
• Total Price: $249.99

[ MSI Z97-Gaming 5 LGA 1150 ATX Intel Motherboard

• SKU: 248427
• Condition: New
• Status: Completed
• Transaction #: <redacted>

• Quantity: 1
• Price Each: $119.99
• Total Price: $119.99

Note the price on the motherboard, $120 brand new, that was after the $40 bundle savings.

EDIT #2:
Oh man I'd love to back to those days of <$300 i7's and $50-off bundles! Buying an Intel CPU+Board anywhere but Microcenter was just out of the question. I lucked out it's 15mins from my job, so I've been buying from them for over a decade.

 

[CatMerc]

nVidia achieved higher frequencies, that's all. AMD, too, can play that game (stay tuned) and has MUCH more experience with high clock speeds... on 14nm LPP

Intel and AMD count FinFet transistors differently. Seemingly, AMD includes every fin while Intel only counts functional groups (which may consist of two or three fins).

AMD is well ahead in density and performance per area on the CPU front.

Edit: nvm broken logic

 

[ElFenix]

An NDA is a very distinct legal contract, a review embargo is more of a gentleman's agreement (which CAN have legal sway in civil court, but you have to prove damages to be made whole).

If you violate an NDA there are prescribed remedies for your actions that can involve monetary or disciplinary retribution to which you have agreed in advance and are possibly WELL in excess of any damages. There are situations where violating this can even be considered a criminal matter (corporate espionage, for example).

Violating an embargo is much less of an issue and there are no punitive action agreement in advance and it is NEVER a criminal matter.

Who said anything about a criminal matter?

And gentlemen's agreement? What the heck kind of thing do you think this is? Some kind of handshake and a nod?


(Knowing full well that a handshake deal resulted in a multibillion dollar dollar verdict against Texaco)

 

[CatMerc]

Who said anything about a criminal matter?

And gentlemen's agreement? What the heck kind of thing do you think this is? Some kind of handshake and a nod?

Almost?
In the case of review samples and such in this industry, it more or less amounts to that. There is rarely anything you need to sign. It's just an email saying "Hey is it cool if we give you information you can't talk about until X day at Y time?", and if you reply yes you get the information.

If you break a company's trust, you no longer get samples from anyone, so it goes against your interests to break it.