AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[tential]

65W for 8 cores running @3.7 GHz seems a little too optimistic, no?

Already a winner for me. That's the chip I'll put in my server.
The gaming done off that server doesn't have to be 4k/60fps as long as it's 1080p/60fps on console like settings it's ok.

If the perf/OC is right.... I'd be scared to see what Zen 2 looks like. Will be interested to see how this progresses over the next 2 years as I look for an upgrade for my gaming rig.

 

[lopri]

I really hate this turbo speed advertising.

 

[jpiniero]

I really hate this turbo speed advertising.

It makes sense since the chips can (in theory) run at that speed with all cores.

 

[Doom2pro]

My budget for my Premium VR build was $500 for CPU/Heatsink Combo... I guess I'll have to add room now for an AIO cooler or something, ohh well

 

[itsmydamnation]

I really hate this turbo speed advertising.

Maybe it isn't but amd have gone to down clocking based off load. So int / light amount of AVX/sse keeps the clocks at 4 and then it drops based on sensemi logic.

That would align with some of the things amd have said about zen

 

[lolfail9001]

Maybe it isn't but amd have gone to down clocking based off load.

Would explain suspect TDP for 1700 too.

 

[zinfamous]

If those are base clocks this is going to be a bloodbath. Pricing alone so far is setting up the market for a bloodbath. Those ridiculous prices Intel charges for their HEDT CPUs and mainstream i7s are going to come crashing down fast. AMD is giving more cores to a LONG stagnated 4c8t market, and more than decent performance to go with it.

At the new horizon demo and at least at those two real world workloads (Blender and Handbrake), 3.4GHz fixed 8c16t Ryzen was equal or a tiny bit faster than a stock 6900k (3.2/3.7GHz). We know there's a 3.6GHz base, 3.9GHz turbo sample out there. Remember canardpc's early results using a 3.15/3.4GHz A0 early sample? Moving base clocks to at least 3.6GHz already closes most of the gaps here relative to the 6900k. 4GHz base in the halo part is certainly faster across the board than a 3.2/3.7GHz 6900k if not close to the 10 core 6950k. 6900k performance for half the price or less. Now I understand why key executives are leaving Intel, along with that hilarious note avoiding the elephant in the room. They've seemingly been caught with their pants down.

7700k is another thing, although it provides half the cores it's clocked absurdly high and can do 5GHz when delidded. 4c or 6c Ryzen probably won't touch this.. Yet AMD is selling you a 8c16t 3.7??GHz CPU for say, $400 while $350 buys you a 4.5-5GHz 4c8t CPU. Some may have no use for the extra cores, some may... yet games are increasingly multithreaded and there are cases where 6c or 8c HSW-E or BDW-E are ahead of 6700k/7700k. This is not 2008-2010... games are catching up with the amount of cores available today and getting good use out of them. Not to mention the ocassional video encode or whatever... more strong cores at this stage is always welcome.

Come on, someone leak some key benchmarks so this is all finally over... BTW, finally there's some hint of competition! It's been a long 10 years... Phenom II was good... this is shaping up to be better.

All of this is exactly what I am hoping for and assuming going forward.

8c/16t shouldn't be a direct and complete competitor to 7700k, but it should off a choice for certain customers that want no-fail gaming (with potential incredible upside on the long-term), with absolute unmatched HEDT performance for half the price.

That, alone, is such an incredible deal, because, think about it: ff this thing can perform at the exact same specs of a 6700k ($300) or maybe 5% less at worse on comparable applications (games), and punch above a 6900k ($1200) about 5-10% of performance at best on comparable workloads (Premiere Pro), and all for ~$370-400?

Holy shit man--that is a game changer. spending $400 bucks on a single CPU that covers the entire range of the current $350-1200 range of the high end. I mean, that is incredible value if it pans out on benches like that. Feels like hype train a bit, but, man: I do hope you are right.

 

[.vodka]

Agreed. Should it end up this way, it's a much needed back to competition scenario where the consumer wins.

On top of all this, remember the entire Ryzen lineup is unlocked. No K tax. The 1700 is a perfect CPU to be paired with a beefy motherboard to clock as far as it'll go for cheap, the same could be said for the cheapest 4c8t Ryzen.

Motherboards shouldn't be too expensive, as we have dual channel memory (vs HEDT's 4 channels) and most of the I/O and southbridge functionality is right on the CPU. A well overengineered VRM will handle that 1700 just fine with decent cooling...


I don't think leaks can be stopped at this point... come on...



Intel sure has the $65 G4560+iGPU which is an awesome CPU I intend to build many cheap yet powerful rigs for many people in the coming months with to which AMD has no response to right now apart from AM4 Bristol Ridge... Raven Ridge is far away, but I think a good old CPU only product can and will make quite the splash in the market as it is right now and has been for quite a while.

 

[lolfail9001]

On top of all this, remember the entire Ryzen lineup is unlocked. No K tax.

Unless of course AMD decided to make TDP limit on non-X models hardcoded, so any overclocking is useless. The 65W is highly suggestive of aggressive throttling, after all.

 

[.vodka]

It's a possibility, yet why go state that all your CPUs are unlocked only to shoot yourself in the foot that way? Not that AMD has never done such a thing before... Overclocker's dream, among some

Either way, that will get clarified soon. If it ends up that way, well, the 1700x isn't that much more expensive, and still highly disruptive.

 

[inf64]

The 65W is highly suggestive of aggressive throttling, after all.

65W is indication of low base clock . Think 3.3Ghz or so. Turbo of 3.7Ghz is ST turbo I believe and all core turbo is somewhere halfway between those two figures (~3.5Ghz or so). When manually OCIng the TDP spec means nothing, AMD won't hard limit you to 65W. You are voiding warranty by touching the manual OC in BIOS so they don't care, just like intel. They(intel) will take from you some smallish amount of $$ if you'd like to get additional insurance in case you burn something, which is a fair deal imo.

 

[beginner99]

How can you establish which system actually has the upper hand when you have no clear idea for a Ryzen 6c system cost?

True. My point was merely that for gaming and future proofing it is not that clear which CPU will be the better choice if priced similar. We can profit from more cores or from the iGPU. Both is possible.

Agree.

DX12 adoption should help accelerate better utilisation of more cores.

Or dx12 multi-adapter can make use of iGPU, which Ryzen lacks compared to 7700k. I'm not taking sides just a consideration I'm making because I will buy Ryzen or 7700K (or maybe Skylake-X). These prices don't mean much yet because stuff where I live usually is costlier, especially AMD stuff (difference between US prices and where I live is always much higher for AMD parts).

 

[coercitiv]

Unless of course AMD decided to make TDP limit on non-X models hardcoded, so any overclocking is useless. The 65W is highly suggestive of aggressive throttling, after all.

Even with a TDP limit set in place, overclocking would be far from useless in lightly threaded scenarios when not all cores are working (and considering the number of cores, many tasks will enter the light category). In heavy MT loads the CPU would be performing nicely anyways.

The 65W TDP rating sounds more like making sure there's an SKU for cheaper non oc boards as well. It may also allow AMD to include a cheaper cooler, if they decide to go against Intel's decision to remove coolers from oc SKUs. Many reasons can be given for the TDP rating, and frankly the "hardcoded" anti-oc explanation is the least likely.

[EDIT] From the lambda-tek.com links the 65W TDP comes with a Wraith cooler while the 95W TDP SKU comes without cooler.

 

[krumme]

True. My point was merely that for gaming and future proofing it is not that clear which CPU will be the better choice if priced similar. We can profit from more cores or from the iGPU. Both is possible.

Or dx12 multi-adapter can make use of iGPU, which Ryzen lacks compared to 7700k. I'm not taking sides just a consideration I'm making because I will buy Ryzen or 7700K (or maybe Skylake-X). These prices don't mean much yet because stuff where I live usually is costlier, especially AMD stuff (difference between US prices and where I live is always much higher for AMD parts).

I dont understand - so you expect an igpu to help 2 years down the line with gaming in systems with 1080ti like gpu perf running a 4c 8t vs 8c 16t system?

I take the 8c solution every single day. Igpu ? It seems to me more and more unnessesary perspectives comes up. Pci lanes, memory channels, sata port and now igpu? All for nothing. Its like people looking for an wrong excuse to buy a certain brand. Its just like when k6 k7 and k8 is released. Its just flat out better value. Thats how it is when a non brand product have to sell and enthusiast can profit from it because we have the knowledge unlike other consumers - so dont waste your knowledge getting lost in the marketing wood.

Its the rule of the game that less brand have to sell on price and why everyone want to build a brand.

 

[krumme]

Even with a TDP limit set in place, overclocking would be far from useless in lightly threaded scenarios when not all cores are working (and considering the number of cores, many tasks will enter the light category). In heavy MT loads the CPU would be performing nicely anyways.

The 65W TDP rating sounds more like making sure there's an SKU for cheaper non oc boards as well. It may also allow AMD to include a cheaper cooler, if they decide to go against Intel's decision to remove coolers from oc SKUs. Many reasons can be given for the TDP rating, and frankly the "hardcoded" anti-oc explanation is the least likely.

[EDIT] From the lambda-tek.com links the 65W TDP comes with a Wraith cooler while the 95W TDP SKU comes without cooler.

Do we have any itx boards comming up early on release?

 

[cytg111]

Don't you guys find it strange that the clock speeds are listed? Turbo State clocks, anyone would advertise? Its very easy to get in trouble with some "unsatisfied" clients this way... don't you think?

What if, the listed clocks, are actually... base clocks?

And turbo goes to 7?

 

[bjt2]

Don't you guys find it strange that the clock speeds are listed? Turbo State clocks, anyone would advertise? Its very easy to get in trouble with some "unsatisfied" clients this way... don't you think?

What if, the listed clocks, are actually... base clocks?

It's impossible, because 14nm is a low power process and moreover i would be right... So it's impossible.

 

[bjt2]

Unless of course AMD decided to make TDP limit on non-X models hardcoded, so any overclocking is useless. The 65W is highly suggestive of aggressive throttling, after all.

I think that this is not the case, and i try to explain why.
Until now the process of CPU binning was stochastic. You suppose that the dices near the center are faster, then try it to see if they work and then package.

But now AMD has implemented a very powerful technology: XFR. I think that it can be used for precise binning. How? Cut the chip, package it, put an air cooler, make it run a standard power virus load (EG multithreading prime 95 like), in a controlled environment (EG: a 35C room) on 8c, 6c 4c CCX0 and 4c CCX1. Then record the mean clock achieved in each situation. Now subtract a fixed percentage, e.g. 10% and round to 100MHz. These are the base clock that you can advertise for this particular sample and for which you can bin. You must choose the core number only based on projected price.

Now let's return on our CPUs. If the binning is done as I suggested, you may buy a 3.7GHz rated chip, but in the same conditions (MB, cooler, room temperature), no way that it can clock like a 4GHz rated chip, because if it was rated inferior, it is a truly inferior chip.

 

[bjt2]

TDP limits can surely be disabled

And turbo goes to 7GHz?

I know that it's a rethoric question, but i will answer: with XFR IMHO it's not possible anymore to specify a turbo max frequency, because there is not max. If the cooler is very bad or very chocked with dirt, probabilty the CPU will not go much up than base clock, that they can guarantee. They can't guarantee any upper clock.
Conversely, most of the ES have a rated turbo max, because in that samples the XFR was disabled and only the turbo old style was available...

 

[cytg111]

I think that this is not the case, and i try to explain why.
Until now the process of CPU binning was stochastic. You suppose that the dices near the center are faster, then try it to see if they work and then package.

But now AMD has implemented a very powerful technology: XFR. I think that it can be used for precise binning. How? Cut the chip, package it, put an air cooler, make it run a standard power virus load (EG multithreading prime 95 like), in a controlled environment (EG: a 35C room) on 8c, 6c 4c CCX0 and 4c CCX1. Then record the mean clock achieved in each situation. Now subtract a fixed percentage, e.g. 10% and round to 100MHz. These are the base clock that you can advertise for this particular sample and for which you can bin. You must choose the core number only based on projected price.

Now let's return on our CPUs. If the binning is done as I suggested, you may buy a 3.7GHz rated chip, but in the same conditions (MB, cooler, room temperature), no way that it can clock like a 4GHz rated chip, because if it was rated inferior, it is a truly inferior chip.

I think that is a somewhat.. overthought line of reasoning. I agree that a 65watt TDP rated chip is weird given an envoriment where you can auto overclock in very small increments and ignore rated TDP at will...Why even bother with the 65watt tag? - 65watt is too hot for a laptop right? Cant be that?

 

[lolfail9001]

If it ends up that way, well, the 1700x isn't that much more expensive, and still highly disruptive.

Fair enough.

 

[Atari2600]

I think that is a somewhat.. overthought line of reasoning. I agree that a 65watt TDP rated chip is weird given an envoriment where you can auto overclock in very small increments and ignore rated TDP at will...Why even bother with the 65watt tag? - 65watt is too hot for a laptop right? Cant be that?

'cos you can charge slightly more for it than a 95W rated chip at the same clock speed.

The user is aware of the efficiency improvement and will be more accepting of a price tag $20 higher.

 

[bjt2]

I think that is a somewhat.. overthought line of reasoning. I agree that a 65watt TDP rated chip is weird given an envoriment where you can auto overclock in very small increments and ignore rated TDP at will...Why even bother with the 65watt tag? - 65watt is too hot for a laptop right? Cant be that?

The 65W TDP can hide a culprit. The silicon quality is extremely variable. There are high leakage parts and low leakage parts. Some articles points to the fact that the new 14nm is not much worse than older processes in this aspect. But it is somewhat noticeable. And have conseguences: high leakage parts are the one in which the mean Vt of the transistors is lower. This means that at same Vcore will go higher in frequency. And usually they will overclock higher, also. This means also that at high TDP they are better than lower leakage chips. Where they don't shine is lower clock zone, because of high leakage.
Low leakage chips instead are best for low clocked chips, but do not overclock well.

In short: low TDP chips are probabily low leakage chips, that, having higher mean Vt, will clock less, but will be cooler at low frequency.

 

[coercitiv]

Why even bother with the 65watt tag?

It includes a cooler for that rating and also ensures compatibility with certain budget friendly motherboards. It also conveniently matches mainstream Intel parts, making it a perfect candidate to show off the efficiency of Zen power management.

65watt TDP rated chip is weird given an environment where you can auto overclock in very small increments and ignore rated TDP at will...

Why bother with the 91W TDP rating on top i7 when many will take it past 120W at will? Every CPU needs a TDP rating for their stock config, I thought we we past that on the forums.

 

[LTC8K6]

It makes sense since the chips can (in theory) run at that speed with all cores.

Could be single core turbo though?