AMD Ryzen Clock and Turbo speeds query

[Dethroy]

BTW Does the lower TDP really affect that min frame rate? Honestly I think that the GPU is responsible for the frame rates and CPU just keeps company and has to be strong enough to keep up with the GPU. For graphics, GPU is in the front row not CPU.

Yes, the GPU is the bottleneck in most cases. But when you have a beefy GPU and you are playing less demanding games at 1,920x1,080, your CPU needs to be quick enough to feed your GPU or else you'll run into a CPU bottleneck (CPU bound scenario). This is clearly visible in the video I linked above and is also mentioned at 12m04s. Other than that, if you plan on keeping the CPU for a long period of time, by the time you move on to a newer/better GPU, all of a sudden your CPU may start to become the weak link far more often instead.

So I'd say if you want to jump on the Ryzen train right now, and you aren't constrained by cooling options due to case dimensions (I'm looking at you SFF mini-ITX cases), then I'd say, spent that extra money on the R7 1700X, because that additional 70$ investment may as well be zilch considering how many years you may keep said processor.

 

[ratirt]

Yes, the GPU is the bottleneck in most cases. But when you have a beefy GPU and you are playing less demanding games at 1,920x1,080, your CPU needs to be quick enough to feed your GPU or else you'll run into a CPU bottleneck (CPU bound scenario). This is clearly visible in the video I linked above and is also mentioned at 12m04s.

Yes I noticed that but this scenario doesn't apply to all games. Which was also said. Sometimes the 1700 was even faster than 1800X. This also depends on the game you play. Sometimes it matched the 1800X performance. Which kind of GPU they were using for the testing?

 

[Dethroy]

Yes I noticed that but this scenario doesn't apply to all games. Which was also said. Sometimes the 1700 was even faster than 1800X. This also depends on the game you play. Sometimes it matched the 1800X performance. Which kind of GPU they were using for the testing?

As I said, it depends on the situation: game/benchmark being either CPU- or GPU-bound. That's why almost all benchmarks that are run at 2,560x1,440 or higher resolutions (read: more likely to run into a GPU bottleneck), show almost exactly the same framerates across most CPUs, because current generation GPUs aren't pushing enough FPS in said scenario except for a few exceptions.

 

[Nokiron]

1. By "all" you mean all available 8 core 16 threads CPU's? Go on, check the 6900k overclocks. You'll be surprised where they start hitting the OC wall. It's around 4GHz. Going beyond that and you get stupid voltages and insane heat output.

Just a side note, I dont really agree with this. Asus stated in their OC-guide with Haswell-E that the following was the norm (5960X):

• 4.4Ghz at 1.3V was below average
• 4.5Ghz at 1.3V was average
• 4.6Ghz at 1.3V was above average

And regards to Broadwell-E (all models i presume in regards to their wording):

I mean, its pretty clear that Haswell-E has slightly better overclocking potential in relation to voltage but I would not call it stupid voltages. I mean sure, you will get some serious heat output but really, you are now pushing a 40%+ OC on 8-cores. You have to expect the cooling requirements.

 

[Aenra]

the following was the norm (5960X):

• 4.5Ghz at 1.3V was average

Show me where you found that please.

edit: disregard this.. Stupid of me, confused 5960X with 6950X

 

[Kanan]

What I meant with Rev 2 is only about its overclocking abilities. To make better use of Ryzen in games is up to Microsoft (likely fixed soon) and the game devs (takes longer but will be done too). The problem is mainly on Microsoft, game devs only have to program more core/thread usage, but you have that already in many games anyway. So I expect the biggest problems to be fixed soon.

If you overclock the Ryzen 1700 its TDP of 65W is of no consequence as OC effectively shuts it off. And with a Ryzen 1700 you want OC anyway because it's not really clocked for gaming otherwise.

That said, you can very well buy the Ryzen now and a 1700 too, but be aware that the CPU is still very new and it needs some time to be fully utilised, though I expect the biggest problems (that is windows scheduler) will be fixed very soon. This is based on what the software dev said in the video I already linked before.

 

[uuuaaaaaa]

https://siliconlottery.com/ has binned Ryzen cpu's!

it seems that there is a pretty good chance that any ryzen 7 cpu will do 3.9GHz.

 

[EarthDog]

Below its own boost...

 

[Kanan]

Below its own boost...

Boost isn't for all cores and not all the time, a 3.9GHz overclock is a overclock. Intended troll or simply no knowledge of how turbo works in CPU architectures of the last decade or so?

 

[EarthDog]

Boost isn't for all cores and not all the time, a 3.9GHz overclock is a overclock. Nice troll post btw.

I'm well aware how their boost and xfr works. You are not able to overclock all cores past their boost/xfr... that's sad.

 

[Aenra]

Thanks for the link, had forgotten about SL. I'd buy that.. https://siliconlottery.com/collections/frontpage/products/1800x40g

30 bucks up, but a guaranteed 4.0Gigs

edit: mobo depending, lol.. (yes, i'm still butthurt)

 

[Vayra86]

Zen+ should be 7nm also.

Dream on. 7nm is so far away, anyone who puts it on the roadmap before the year 2020 is a blatant liar, or an extreme optimist.

IF we see 7nm before 2020 it will be in the mobile space. Not x86.

@Kanan why always so serious, buddy? Relax

 

[ratirt]

Dream on. 7nm is so far away, anyone who puts it on the roadmap before the year 2020 is a blatant liar, or an extreme optimist.

IF we see 7nm before 2020 it will be in the mobile space. Not x86.

@Kanan why always so serious, buddy? Relax

I can get a 14nm if they would improve things or 2 . and get better performance than now.

 

[Dethroy]

Ryzen's launch is very polarizing indeed, but that shouldn't lead to forum members attacking each other personally. We can do better...

 

[RejZoR]

Dream on. 7nm is so far away, anyone who puts it on the roadmap before the year 2020 is a blatant liar, or an extreme optimist.

IF we see 7nm before 2020 it will be in the mobile space. Not x86.

@Kanan why always so serious, buddy? Relax

It took ages for Intel and AMD to come to 14nm, depending on process, they might shrink it down to 10nm, but I think they'll just stick with 14nm for quite a while.

 

[Tatty_Two]

Some just don't or won't learn, if this mess does not calm down to something more civil from here on in things will get sporty real quick.

 

[Kanan]

It took ages for Intel and AMD to come to 14nm, depending on process, they might shrink it down to 10nm, but I think they'll just stick with 14nm for quite a while.

Gf will be on 7nm quite soon, but the 7nm Gf is talking about isn't the same as Intels 7nm it's rather comparable to the 10nm of Intel. 14nm of GF isn't quite as good as Intels 14nm too, but it's good enough I'd say.

 

[RejZoR]

Yeah, I know they measure things differently. Samsung also has their own measurement of nodes. Gets kinda confusing for average users who know a thing or two, but not enough to really understand it.

 

[Kanan]

Yeah, I know they measure things differently. Samsung also has their own measurement of nodes. Gets kinda confusing for average users who know a thing or two, but not enough to really understand it.

Yeah I think it all started with 16/14nm ff stuff of Tsmc/GF to get weird.

 

[Vario]

Dream on. 7nm is so far away, anyone who puts it on the roadmap before the year 2020 is a blatant liar, or an extreme optimist.

IF we see 7nm before 2020 it will be in the mobile space. Not x86.

@Kanan why always so serious, buddy? Relax

AMD claims the next revision of Zen is 7nm from Global Foundries. Not a dream on my part. Just repeating what the chip manufacturers are saying. If it doesn't come to be true, frankly I don't give a shit because I plan to be on Ivy bridge another 5 years anyway at this rate.

Dr. Lisa Su, AMD president and CEO

“The five-year amendment further strengthens our strategic manufacturing relationship with GLOBALFOUNDRIES while providing AMD with increased flexibility to build our high-performance product roadmap with additional foundries in the 14nm and 7nm technology nodes. Our goal is for AMD to have continued access to leading-edge foundry process technologies enabling us to build multiple generations of great products for years to come.”

 

[TheLostSwede]

Dream on. 7nm is so far away, anyone who puts it on the roadmap before the year 2020 is a blatant liar, or an extreme optimist.

IF we see 7nm before 2020 it will be in the mobile space. Not x86.

@Kanan why always so serious, buddy? Relax

TSMC will start 7nm production this year http://www.electronicsweekly.com/news/business/7nm-tape-outs-tsmc-q2-2017-01/

Sure, not x86, nor is it "true" 7nm apparently, but hey...

GloFo's current 14nm is in the same boat as TSMC, it's more like Intel's 20nm, so their shrink from "14" to "7" is more like 20 to 14. In other words, they'll most likely deliver as promised, but it's a matter of what different companies call things...

 

[ratirt]

TSMC will start 7nm production this year http://www.electronicsweekly.com/news/business/7nm-tape-outs-tsmc-q2-2017-01/

Sure, not x86, nor is it "true" 7nm apparently, but hey...

GloFo's current 14nm is in the same boat as TSMC, it's more like Intel's 20nm, so their shrink from "14" to "7" is more like 20 to 14. In other words, they'll most likely deliver as promised, but it's a matter of what different companies call things...

I wouldn't be so sure about what you just said. Not all components in intel is 14/16nm but it's not that huge difference as you described. It's different shrinking process and some elements stay larger then others but still it's not that much different.
BTW. I kinda don't care if it's 14 or 7nm. for me it can be 22 but if it delivers twice as much performance as current CPU's with reasonable price tag and TDP I don't really care Although I know those things come with shrinking process a bit. Well architecture also has a saying in it though.

 

[TheLostSwede]

I wouldn't be so sure about what you just said. Not all components in intel is 14/16nm but it's not that huge difference as you described. It's different shrinking process and some elements stay larger then others but still it's not that much different.
BTW. I kinda don't care if it's 14 or 7nm. for me it can be 22 but if it delivers twice as much performance as current CPU's with reasonable price tag and TDP I don't really care Although I know those things come with shrinking process a bit. Well architecture also has a saying in it though.

So if you're not sure, why did you point it out?

Does this help convince you? Admittedly that's GloFo LPE, not LPP, but hey...

 

[Kanan]

So if you're not sure, why did you point it out?

Does this help convince you? Admittedly that's GloFo LPE, not LPP, but hey...

Thanks, that's exactly what I meant. Seeing this again, 7mm Samsung should be equal to 10nm Intel or maybe a tad better, TSMCs node is still inferior to both and if they don't up their game the difference will increase too.

 

[eidairaman1]

Yes, the GPU is the bottleneck in most cases. But when you have a beefy GPU and you are playing less demanding games at 1,920x1,080, your CPU needs to be quick enough to feed your GPU or else you'll run into a CPU bottleneck (CPU bound scenario). This is clearly visible in the video I linked above and is also mentioned at 12m04s. Other than that, if you plan on keeping the CPU for a long period of time, by the time you move on to a newer/better GPU, all of a sudden your CPU may start to become the weak link far more often instead.

So I'd say if you want to jump on the Ryzen train right now, and you aren't constrained by cooling options due to case dimensions (I'm looking at you SFF mini-ITX cases), then I'd say, spent that extra money on the R7 1700X, because that additional 70$ investment may as well be zilch considering how many years you may keep said processor.

I'D rather have a GPU bottleneck than cpu so that the cpu can keep that gpu fed.

Honestly though bottleneck is too easy a term to throw around and really shouldn't be worried about because all circuits have to deal with physical limitations anyway.