TR Forums

derFunkenstein wrote:

No doubt you'll be able to create "125W" CPUs by overclocking. And who outside of AMD yet knows what XFR does to the TDP if there's sufficient cooling, so maybe I'm jumping the gun.

Problem is most leaks thus far show Zen to (at best) overclock as well as Broadwell, which itself is a terrible overclocker compared to Haswell and Skylake. Unless they do some sort of respin, like Skylake to Kaby Lake, maintained 4.5GHz clocks are not going to realistic for most users.

ultima_trev wrote:

derFunkenstein wrote:

No doubt you'll be able to create "125W" CPUs by overclocking. And who outside of AMD yet knows what XFR does to the TDP if there's sufficient cooling, so maybe I'm jumping the gun.

Problem is most leaks thus far show Zen to (at best) overclock as well as Broadwell, which itself is a terrible overclocker compared to Haswell and Skylake. Unless they do some sort of respin, like Skylake to Kaby Lake, maintained 4.5GHz clocks are not going to realistic for most users.

I got 4.0 GHz on all 4 cores stable at stock voltage on my i7-5775c. I could push it further if I wanted to, but I don't really see a reason to do so. It smashes all the benchmarks. What more do you need?

Uh... that chart is comparing a Intel CPU with an iGPU to a AMD CPU without one, right? In that case, Zen does not have the size advantage.

DancinJack wrote:

Mr Bill wrote:

Unboxing later today.
http://techreport.com/news/31507/join-u ... w-kit-live

Not necessarily related to this particular thread, but I thought it warranted a comment. It's clear that the powers that be are listening to the feedback in some of the recent threads about concern for TR, and I thought they should be commended for that.

That unboxing was depressing as all hell. Including a water cooler? Really, AMD? Come on.

Well, I guess the good news is that we know RyZen doesn't OC for crap!

At least TR got a review sample, before release, AND we got extra content because of it. Hard to be too upset. Even if RyZeN does OC for crap.

DancinJack wrote:

At least TR got a review sample, before release, AND we got extra content because of it. Hard to be too upset. Even if RyZeN does OC for crap.

Well, we can see that AMD must be massively confident that it OC's well on air! /s

Vhalidictes wrote:

Uh... that chart is comparing a Intel CPU with an iGPU to a AMD CPU without one, right? In that case, Zen does not have the size advantage.

It's just the cores that are smaller, nothing else on die is smaller. Density is not higher either, they just took out some resources that are barely used. Cache sizes are bigger where Intel doesn't need them to be so large. There really is no advantage here for us.

It might be an advantage for APUs in things like Project Scorpio where it doesn't need to be optimized for so many different uses and they can dedicate more die real estate to whatever they want. AMD has been working to make CPUs more modular. Intel has been racing to the bottom to turn everything into an SoC.

Vhalidictes wrote:

Well, I guess the good news is that we know RyZen doesn't OC for crap!

That might be true, but Jeff said most sites got Noctua air coolers. That tells me there's not much correlation between TR getting that sweet, sweet EK Predator and Ryzen's OC headroom.

derFunkenstein wrote:

Vhalidictes wrote:

Well, I guess the good news is that we know RyZen doesn't OC for crap!

That might be true, but Jeff said most sites got Noctua air coolers. That tells me there's not much correlation between TR getting that sweet, sweet EK Predator and Ryzen's OC headroom.

Yep. Clearly obvious by the Noctua brackets still on the motherboard itself, no less.

DeadOfKnight wrote:

It's just the cores that are smaller, nothing else on die is smaller. Density is not higher either, they just took out some resources that are barely used. Cache sizes are bigger where Intel doesn't need them to be so large. There really is no advantage here for us.

That is it - half of Intel's quad-core parts is a GPU that is effectively deadweight for us gaming enthusiasts. AMD just used it for CPU cores. Intel can of course do the same in their next generation. Will they?

_ppi wrote:

DeadOfKnight wrote:

It's just the cores that are smaller, nothing else on die is smaller. Density is not higher either, they just took out some resources that are barely used. Cache sizes are bigger where Intel doesn't need them to be so large. There really is no advantage here for us.

That is it - half of Intel's quad-core parts is a GPU that is effectively deadweight for us gaming enthusiasts. AMD just used it for CPU cores. Intel can of course do the same in their next generation. Will they?

They have a line of CPUs without a GPU every generation... Maybe you missed them?

-yes, but it's the enthusiast consumer level that's being compared. It's also *more* than half of the core reserved for GPU on those CPU's (..kaby lake and skylake).

Again though, I'm not sure it makes much difference beyond power efficiency at **lighter** loads (..which is a big deal for huge server installations, but really not one at all for enthusiast use). My guess is that both the Intel consumer enthusiast processors (K) and AMD's will top-out very similarly in freq.. BUT it's likely that the AMD architecture requires more advanced cooling solutions to get there.

_ppi wrote:

DeadOfKnight wrote:

It's just the cores that are smaller, nothing else on die is smaller. Density is not higher either, they just took out some resources that are barely used. Cache sizes are bigger where Intel doesn't need them to be so large. There really is no advantage here for us.

That is it - half of Intel's quad-core parts is a GPU that is effectively deadweight for us gaming enthusiasts. AMD just used it for CPU cores. Intel can of course do the same in their next generation. Will they?

I'll say that the GPU is not 'deadweight', primarily because it can be used to power secondary displays that are not used for gaming. However, they don't need to be so 'powerful' for the desktop SKUs- enough to run outputs with the proper decoding blocks for media would be great, and replace those extra almost completely useless GPU units with x86 cores!

..for most enthusiasts the integrated GPU is probably worse than "dead weight". Not only will it *not* be used (by most enthusiasts), but added voltage (beyond that of cores/memory/etc.) is being required even if there is no load on that portion of the CPU.

CScottG wrote:

..for most enthusiasts the integrated GPU is probably worse than "dead weight". Not only will it *not* be used (by most enthusiasts), but added voltage (beyond that of cores/memory/etc.) is being required even if there is no load on that portion of the CPU.

Well, it didn't bother my 2500k to power two monitors at 4.5GHz for years on end with +10% core voltage. I'm just now working on this 6700k after narrowing down a bit of instability to this Corsair RAM not wanting to run at its XMP settings, so we'll see if it makes a difference after that.

CScottG wrote:

-yes, but it's the enthusiast consumer level that's being compared. It's also *more* than half of the core reserved for GPU on those CPU's (..kaby lake and skylake).

Again though, I'm not sure it makes much difference beyond power efficiency at **lighter** loads (..which is a big deal for huge server installations, but really not one at all for enthusiast use). My guess is that both the Intel consumer enthusiast processors (K) and AMD's will top-out very similarly in freq.. BUT it's likely that the AMD architecture requires more advanced cooling solutions to get there.

There's so much AMD hype that people are forgetting Intel has been fighting with one hand tied to their backs by "wasting" transistors/die space on the relatively huge iGPU on SKL/KBL, and BW-E is still ~1 billion transistors less than Zen despite having 2 more cores, 2x the memory channels, 8MB more L3 cache and more PCIe lanes.

strangerguy wrote:

There's so much AMD hype that people are forgetting Intel has been fighting with one hand tied to their backs by "wasting" transistors/die space on the relatively huge iGPU on SKL/KBL, and BW-E is still ~1 billion transistors less than Zen despite having 2 more cores, 2x the memory channels, 8MB more L3 cache and more PCIe lanes.

There's more than one way to count transistors. Remember when AMD changed how they counted Bulldozer, and the number dropped by 40%? http://www.anandtech.com/show/5176/amd- ... 12b-not-2b

Regarding the IGP- to be honest, that's just because Intel is selling laptop chips into the desktop market, while AMD are selling server chips into the desktop market.

CScottG wrote:

..for most enthusiasts the integrated GPU is probably worse than "dead weight". Not only will it *not* be used (by most enthusiasts), but added voltage (beyond that of cores/memory/etc.) is being required even if there is no load on that portion of the CPU.

I should point out though that not all us enthusiasts play games. I don't, so I don't care about GPU performance. I like having the Intel iGPU because it's perfectly adequate for the job, and has good Linux support. I want high CPU performance though, which is why I have the 4790K, and I'm using the iGPU. I could also use the PEG ports for something else.

To answer the OP's question after looking at some reviews...

In gaming? It doesn't look like Intel ever lost the lead.

In highly threaded productivity workloads? It may take Intel some time to beat them in Price\Performance here. Ryzen is very solid in this regard.

So AMD made a CPU that performs great in synthetics.

Gaming performance is pretty inconsistent though. Guru3D had the biggest game suite thus far. Games like The Division or Rise of the Tomb Raider make this CPU look pretty bad although the other games it performed pretty close to the 5960X.

So AMD re-released the 5960X, three years later. Still better than Bulldozer, which was essentially Netburst re-released a decade later.

Gaming looks functional enough- if you're looking for grunt in other areas without giving up too much in gaming (depending on what you're coming from), RyZen looks like a solid release. And once overclocked, if you're into that, there's very little difference between the three parts released- the X models have a small boost function, usually <100MHz, that adds very little, so putting the screws to an R7 1700 would do just fine for most.

And it looks like there may be scheduling issues with AMD's interpretation of hyperthreading (or Microsoft's interpretation of AMD's interpretation) as there are reports that turning off AMD HT boosts game benchmarks measurably, so we may see a boost in the not too distant future.

But as it stands today, for gaming, the 7700k is where it's at.

ultima_trev wrote:

So AMD re-released the 5960X, three years later.

...at 1/3 the current street price of a 5960X. That counts for something.

just brew it! wrote:

ultima_trev wrote:

So AMD re-released the 5960X, three years later.

...at 1/3 the current street price of a 5960X. That counts for something.

For now, it sure does. I have to wonder, though- could AMD have a quad-channel 'workstation' variant coming, maybe with more PCIe lanes? Could they have a version coming that has ECC support? Do they have a stepping in the works that clocks higher?

Also, the break with ECC support from the past is a bit perplexing here; AMD had to know that people were going to want to use the CPUs for 'workstation' builds, especially home workstations, and could have easily thundered into the market just by including ECC (or say charging another US$50-100 for an equivalent SKU that also supports ECC).

Airmantharp wrote:

just brew it! wrote:

ultima_trev wrote:

So AMD re-released the 5960X, three years later.

...at 1/3 the current street price of a 5960X. That counts for something.

For now, it sure does. I have to wonder, though- could AMD have a quad-channel 'workstation' variant coming, maybe with more PCIe lanes?

I suspect that those features will be reserved for the Naples variant, on a different socket. Quad-channel in particular requires a LOT of additional pins.

Airmantharp wrote:

Could they have a version coming that has ECC support?

I hope so.

Airmantharp wrote:

Do they have a stepping in the works that clocks higher?

Your guess is as good as mine.

Edit: I'd be surprised if they don't though. Since these CPUs seem to have pretty reasonable TDPs out of the gate, they ought to be able to eke out at least a few hundred MHz without requiring exotic cooling solutions.

Airmantharp wrote:

Also, the break with ECC support from the past is a bit perplexing here; AMD had to know that people were going to want to use the CPUs for 'workstation' builds, especially home workstations, and could have easily thundered into the market just by including ECC (or say charging another US$50-100 for an equivalent SKU that also supports ECC).

They may yet release something like this, assuming the AM4 socket has the extra pins for the parity bits. I've been looking for information on the AM4 pinout but haven't been able to find anything; presumably it is only available under NDA. I miss the days when AMD used to publicly release the engineering data sheets for their CPUs.

CScottG wrote:

..for most enthusiasts the integrated GPU is probably worse than "dead weight". Not only will it *not* be used (by most enthusiasts), but added voltage (beyond that of cores/memory/etc.) is being required even if there is no load on that portion of the CPU.

The good news is that modern CPUs will shutdown sections of the package that aren't-in-use, so hopefully the power draw shouldn't be an issue (unless the OS turns the iGPU on for some reason).

Airmantharp wrote:

I'll say that the GPU is not 'deadweight', primarily because it can be used to power secondary displays that are not used for gaming. However, they don't need to be so 'powerful' for the desktop SKUs- enough to run outputs with the proper decoding blocks for media would be great, and replace those extra almost completely useless GPU units with x86 cores!

Thanks, Airmantharp! I'll set this up tonight - various weird issues with brand-new AMD drivers and multiple monitors will all go away if I just connect my other monitors to the iGPU. I'm amazed that I didn't think of this already.

Firestarter wrote:

just brew it! wrote:

If I was a bettin' man I'd probably short AMD (or buy put options)

it's always "buy the rumor, sell the news" isn't it?

...and AMD is down over 4% based on the initial RyZen reviews. Called that one right.

Wall Street bought into the hype, and was convinced AMD had an "Intel killer" on its hands.

just brew it! wrote:

Firestarter wrote:

just brew it! wrote:

If I was a bettin' man I'd probably short AMD (or buy put options)

it's always "buy the rumor, sell the news" isn't it?

...and AMD is down over 4% based on the initial RyZen reviews. Called that one right.

Wall Street bought into the hype, and was convinced AMD had an "Intel killer" on its hands.

Buy on the rumor.
Sell on the news.

chuckula wrote:

just brew it! wrote:

Firestarter wrote:

it's always "buy the rumor, sell the news" isn't it?

...and AMD is down over 4% based on the initial RyZen reviews. Called that one right.

Wall Street bought into the hype, and was convinced AMD had an "Intel killer" on its hands.

Buy on the rumor.
Sell on the news.

Firestarter beat you to it by several days.

Sold my AMD stock before closing yesterday!