Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[GroundZero7]

And in 2 months what degradation-free OC headroom does SR have.

As long as temps stay cool degradation shouldn't be a problem. I'v had my Phenom II X6 OC to 4ghz for 6 years, and ran 2 1950xtx's at 100C for 4.

If AMD is still using the high temperature bump material it always has then we should be in good shape.

 

[Valantar]

On the 14nm up to +80% clock is expected around 0.9V at same TDP.

That assumes linear scaling and identical architecture, the first of which is impossible, and only slightly negated by the second.

Also, I've never seen anyone actually test power draw on an FX 9800P, but I don't believe for a second that it runs 4 threads flat out at 2.7GHz and 15W. Carrizo isn't bad, but that would put it on par with Kaby Lake for efficiency. Not happening. Of course, AMD and Intel TDPs are not the same thing. I'd expect that chip to spike up to at least 25W under any sort of load. And even with the (seemingly extremely impressive) efficiency of Zen, a 4c8t chip at 15W seems too good to be true, at least at usable clock speeds.

 

[majord]

FX 9800P on 28nm BULK, with 15W TDP/12W CDP, has: 4 XV cores with 2.7 base 3.6 turbo (probabily you are right on CPU throttling under base on high GPU load) plus 512 SP at 758MHz+
On the 14nm up to +80% clock is expected around 0.9V at same TDP. So 512SP at same power should gain at least 300MHz. Let's say that we use 900MHz 512SP. We can expect that some W can be allocated to the CPU, so to have 2.7-3GHz base 4c8t Zen...

bit optimistic on the base clock.. but I think given what we've seen so far, you wouldn't be too far off.. We know 3Ghz 4/8t base is possible @ ~35w (one core complex of an R7 1700) . by the time APU's launch.. on a more matured process. I can at least see 15w @ 2.5Ghz being a reality.

 

[lolfail9001]

dude you are telling as if electromigration is a problem which only AMD CPUs face.

Nope, i am telling you that Broadwell-E had degradation issues from pushing as little as 1378mV. So yes, i would like to see safe voltages first on Ryzen. Thought Polaris would teach you to be objective but w/e.

What does that mean? Loss of possible OC room due to pushing a processor at near max?

Sort of.

As long as temps stay cool degradation shouldn't be a problem. I'v had my Phenom II X6 OC to 4ghz for 6 years, and ran 2 1950xtx's at 100C for 4.

If AMD is still using the high temperature bump material it always has then we should be in good shape.

Voltage-related degradation happens on any temperature.

 

[A///]

Even stock, right? I've heard the average stock CPU will keep for 10-15 years before the internal architecture degrades and it dies.

 

[tamz_msc]

And in 2 months what degradation-free OC headroom does SR have.

That MSI manual should already give an indication of what you can achieve. It won't take long to arrive at a general consensus on what voltages are going to be safe to run for 24x7 OC.

 

[bjt2]

That assumes linear scaling and identical architecture, the first of which is impossible, and only slightly negated by the second.

Also, I've never seen anyone actually test power draw on an FX 9800P, but I don't believe for a second that it runs 4 threads flat out at 2.7GHz and 15W. Carrizo isn't bad, but that would put it on par with Kaby Lake for efficiency. Not happening. Of course, AMD and Intel TDPs are not the same thing. I'd expect that chip to spike up to at least 25W under any sort of load. And even with the (seemingly extremely impressive) efficiency of Zen, a 4c8t chip at 15W seems too good to be true, at least at usable clock speeds.

+80% was taken from official GF statement (http://www.bitsandchips.it/english/...obal-foundries-pdf-on-the-14nm-finfet-process), moreover it seems that vega can clock higher at same process. So 512SP (Vega) at 1GHz on 14nm FF probabily draw much less than 512SP (old arch) at 758MHz. Similarly, Ryzen is more efficient than XV at same process, so I think that 2.7-3.0GHz 4c8t Ryzen plus 512SP at 900MHz vega arch could be feasible...

 

[GroundZero7]

Voltage-related degradation happens on any temperature.

I never said it didn't. Chips are designed to operate at specific temperatures. Usually around 55-65C. Increased voltage increases the internal temperature of the chip which is dissipated via heatsink. If the chip remains below that during overclocking very little/no extra damage is incurred.

It isn't the voltage, or current that causes degradation, it's the temperature. Even under LN2 pushing that amount of power through a processor creates hot spots that kill the chip. That's why they fail.

I probably have 3 years clocked on my 4ghz Thuban @ 200W(I'v had it for 7 years). It has been at the limit of it's OC the entire time, at under 50C.

You can argue all you want but you are wrong again.

 

[tamz_msc]

Nope, i am telling you that Broadwell-E had degradation issues from pushing as little as 1378mV. So yes, i would like to see safe voltages first on Ryzen. Thought Polaris would teach you to be objective but w/e.

Electromigration-induced failure do not occur due to voltages, but due to a combination of temperature and current. "X volts is too little or too high" doesn't tell us anything unless you list the current spec to go along with it. Perhaps BD-E was designed to have low tolerances for the CPU Vcore, which is not unreasonable since they are all based off Xeon dies.

 

[lolfail9001]

Electromigration-induced failure do not occur due to voltages, but due to a combination of temperature and current.

And CPUs have nice property of pulling more current with more voltage.

It isn't the voltage, or current that causes degradation, it's the temperature.

Actually, it is both current and temperature. Talk about pot and kettle. Voltage is directly complicit here.

I probably have 3 years clocked on my 4ghz Thuban @ 200W(I'v had it for 7 years). It has been at the limit of it's OC the entire time, at under 50C.

Sort of outdated knowledge when talking about 14nm lithography with inherent hot spots due to SMT, don't you think?

Perhaps BD-E was designed to have low tolerances for the CPU Vcore, which is not unreasonable since they are all based off Xeon dies.

But can you claim that the same is not true for Ryzen without any stats to back it up?

 

[GroundZero7]

Electromigration-induced failure do not occur due to voltages, but due to a combination of temperature and current. "X volts is too little or too high" doesn't tell us anything unless you list the current spec to go along with it. Perhaps BD-E was designed to have low tolerances for the CPU Vcore, which is not unreasonable since they are all based off Xeon dies.

Here read..Dutch Geko, and Byriel explain it far better than I could. Thermal cycling will kill a CPU long before electromigration in almost all scenarios. At cooler temperatures it is practically negligible, unless you are pushing absurd voltages.

https://www.reddit.com/r/askscience/comments/1321ei/is_there_an_expected_lifespan_for_computer_cpus/

 

[lolfail9001]

Here read..Dutch Geko, and Byriel explain it far better than I could. Thermal cycling will kill a CPU long before electromigration in almost all scenarios. At cooler temperatures it is practically negligible, unless you are pushing absurd voltages.

And NBTI will make it unusable long before thermal cycling. You are right, they did give a good explanation of what is up and why OCing Broadwell/Ryzen beyond turbo clocks is even possible.

 

[tamz_msc]

And CPUs have nice property of pulling more current with more voltage.

So does the light-bulb.

But can you claim that the same is not true for Ryzen without any stats to back it up?

I made a guess as to why BD-E is like that. What does the voltage tolerance of Ryzen got to do with it?

 

[lolfail9001]

I made a guess as to why BD-E is like that. What does the voltage tolerance of Ryzen got to do with it?

Because we do not know whether Ryzen is tolerant of high voltages. What i am talking all along is that i would rather wait and see how Ryzen and voltage interact before i'd bother pushing it to the limit. Not that i will, what i want can probably be done with slight undervolt, if anything.

 

[Glo.]

Sure. And the GPU itself is far from the whole thermal load of a dGPU. Let's assume that the GPU itself represents 50W (66%) of the 75W TDP. That means a 512SP version at full tilt, no RAM or anything included, is 25W. Alone.

Base clock numbers go out the window once the iGPU gets any sort of load, remember that. With Intel as an example, their 15W >2.5GHz base clock chips often dip well below 1.5GHz when the iGPU kicks in and that's with a far smaller GPU. My desktop A8-7600 (65W) power throttles when the GPU (384SP) is under load, although not by that much (from 3.1GHz down to ~2.4-2.6). APUs are optimized to utilize their thermal window to the fullest extent possible. As such, it would be silly for the CPU not to have a higher base clock when the GPU is idle, as they have cooling to spare. I'd love to see actual clock speeds for a Carrizo chip with the GPU under load. They'd definitely not be 2.7GHz.

I'm not saying Ryzen/Raven Ridge won't give us some awesome APUs. I'm simply saying this: Don't expect miracles. 15W will not be the sweet spot for full-APU load, not by a long shot. It will no doubt be more than enough for a decent CPU. But with a GPU, even 25W will be a massive improvement - either throttling the CPU less, or allowing it to run full-tilt with 10W to spare for the GPU alone. 35W or more would allow this to really shine. GPUs are power hungry. That's just the way it is.

No higher than ~1200MHz, at least. And sure, cutting back a few hundred MHz will save you power. But there's also data suggesting that Polaris power draw flattens out below ~900MHz. So there's not much to save. I hope Raven Ridge has Vega-based iGPUs, but I'm not betting on it. And even so, you'd need 15-20W for a 512SP GPU alone for it to get up to speed. Fitting it inside a 15W TDP APU and not expecting throttling is a pipe dream.

35W designs are only for mobile. Desktop are bigger, and have almost 3 times higher thermal envelope.

That doesn't answer the question of CPU clocks under GPU load. Again: any GPU load means CPU base clocks go out the window. Any at all. If it's a 35W 4c8t chip with a 3GHz base and 11CUs (704SP), I'd expect it to allocate anywhere between 50 and 75% of its power to the GPU under heavy loads - which would mean significant CPU throttling. I'd be very impressed if it maintained 2GHz on all cores.

You forgot one thing. It is Engineering sample. Early one, to add to that. How does it relate to Eng Samples of Ryzen CPUs? We expected normally 3.5 GHz at best for the CPUs, and we ended up with 3.4 GHz base 8C/16T CPU@95W TDP.

I suggest waiting, for end results. Yes, you are correct overall that APUs will throttle. There is a very good reason why AMD states that the Bristol Ridge APUs have 900 MHz base clock, and boost clock of 1108 MHz. In typical circumstances they do not achieve them.
But with Ryzen APUs, I think situation will be "slightly" different. There is ginormous amount of thermal and power sensors on the chips, to keep them within thermal envelope, and adjust the clocks accordingly to workload, and get the best of everything.

This is my personal opinion: 35W 4C/8T+12 CU design will have core clocks of 3.0/3.3 GHz and 0.9/1.2 GHz on the GPU.
95W 4C/8T+16CU+4GB HBM2 design will have 3.6/4.0 on CPU, and 1.1/1.4 GHz on the GPU.

 

[Valantar]

+80% was taken from official GF statement (http://www.bitsandchips.it/english/...obal-foundries-pdf-on-the-14nm-finfet-process), moreover it seems that vega can clock higher at same process. So 512SP (Vega) at 1GHz on 14nm FF probabily draw much less than 512SP (old arch) at 758MHz. Similarly, Ryzen is more efficient than XV at same process, so I think that 2.7-3.0GHz 4c8t Ryzen plus 512SP at 900MHz vega arch could be feasible...

To a certain degree, sure. But under full CPU and GPU load? Again, no. Simply because your comparison couldn't perform at spec under that kind of load either. I'd be surprised if the 2m4t XV ran at much beyond 1GHz under full GPU load at 15W. Also, I'm very wary of directly equating 2 XV modules to 2 Zen cores. Sure, there's a 65W 3GHz 8c16t SKU for Ryzen. Even if that's not binned for low power, expecting more than a 10% difference from that is very optimistic. As such, expecting Raven ridge to require less than 30W at 3GHz flat out for 4c8t seems illogical. And that's without a GPU, of course.

But yes, sure, it could happen. I'm just not letting myself get overcome with excitement quite yet. Ryzen is already looking nigh on miraculous. Expecting even more just seems like a recipe for disappointment.

 

[tamz_msc]

Because we do not know whether Ryzen is tolerant of high voltages. What i am talking all along is that i would rather wait and see how Ryzen and voltage interact before i'd bother pushing it to the limit. Not that i will, what i want can probably be done with slight undervolt, if anything.

Broadwell was a new node for Intel, Ryzen is a completely new node for AMD, so it wouldn't surprise me if voltages are a bit conservative, at least initially.

 

[Herr Kutz]

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

 

[GroundZero7]

And NBTI will make it unusable long before thermal cycling. You are right, they did give a good explanation of what is up and why OCing Broadwell/Ryzen beyond turbo clocks is even possible.

Except that Ryzen has a lot finer voltage regulation than any previous processor out there. That combats NBTI.

Global foundries has improved it's NBTI resistance, and I think 14nm is using sigma stack to eliminate PBTI.

 

[Valantar]

35W designs are only for mobile. Desktop are bigger, and have almost 3 times higher thermal envelope.

Exactly. They're the exact same silicon (well, except that my A8-7600 is Kaveri, not Carrizo). Yet even the 65W version throttles. What do you then expect of the 35W version?

You forgot one thing. It is Engineering sample. Early one, to add to that. How does it relate to Eng Samples of Ryzen CPUs? We expected normally 3.5 GHz at best for the CPUs, and we ended up with 3.4 GHz base 8C/16T CPU@95W TDP.

I suggest waiting, for end results. Yes, you are correct overall that APUs will throttle. There is a very good reason why AMD states that the Bristol Ridge APUs have 900 MHz base clock, and boost clock of 1108 MHz. In typical circumstances they do not achieve them.
But with Ryzen APUs, I think situation will be "slightly" different. There is ginormous amount of thermal and power sensors on the chips, to keep them within thermal envelope, and adjust the clocks accordingly to workload, and get the best of everything.

This is my personal opinion: 35W 4C/8T+12 CU design will have core clocks of 3.0/3.3 GHz and 0.9/1.2 GHz on the GPU.
95W 4C/8T+16CU+4GB HBM2 design will have 3.6/4.0 on CPU, and 1.1/1.4 GHz on the GPU.

It's an engineering sample, yes. It's also a rumor. No corroboration = no reason to believe it when it looks too good to be true.

And in the end, although I'll expect the more fine grained control of Ryzen/RR to have an impact, you'll still be power limited in the end, no matter how fine grained clocks are controlled. With both CPU and GPU fighting for a measly 15W, the question is how efficiently and correctly it will be distributed.

And while both the CPU and iGPU parts have base clocks, it's erroneous to read those two as simultaneous under 100% load. Typically the CPU base clock is for any CPU-heavy workload (w/little to no GPU load), while the GPU is given priority so that it can sustain its base clock at the cost of CPU base clock. There doesn't exist a single mobile APU/SoC in the world capable of sustaining both base CPU and GPU clocks under sustained load. Not one. That's simply not how they're specified (and specifying them in that way would waste thermal headroom in other scenarios).

Tjey can, as you say, "adjust the clocks accordingly to workload, and get the best of everything". That's what throttling is. Can Ryzen/RR do it better than most? Hopefully/possibly. But they can't run two 15W components at full tilt without generating more than 15W of heat. That's physics for you.

 

[lolfail9001]

Except that Ryzen has a lot finer voltage regulation than any previous processor out there. That combats NBTI.

Global foundries has improved it's NBTI resistance, and I think 14nm is using sigma stack to eliminate PBTI.

Combats, perhaps it does, but is it successful? We'll see in few months. I expect quite a few pushing Ryzen to the "VRM is cooking" level.

 

[GroundZero7]

Combats, perhaps it does, but is it successful? We'll see in few months. I expect quite a few pushing Ryzen to the "VRM is cooking" level.

Especially people using WC with no fan on the VRM's. I think that might bite quite a few people that press the XFR button with second tier motherboards.

 

[lolfail9001]

Especially people using WC with no fan on the VRM's. I think that might bite quite a few people that press the XFR button with second tier motherboards.

XFR does not have a button, that's the entire point. But yeah, second tier mobos may just be an cat in the bag, especially since they will not be sent for early reviews.

 

[GroundZero7]

XFR does not have a button, that's the entire point. But yeah, second tier mobos may just be an cat in the bag, especially since they will not be sent for early reviews.

XFR has to be enabled in windows. That came out in the Donanimhaber leak.

 

[scannall]

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

No IGP is a good thing for a lot of us. There are some great slim graphics cards though.

https://www.newegg.com/Product/Prod...rebbr=1&cm_re=gtx_1050-_-14-137-081-_-Product
and
https://www.newegg.com/Product/Prod...rebbr=1&cm_re=gtx_1050-_-14-137-079-_-Product

For a couple of examples.