[News] AMD Ryzen 5 1600X Cinebench R15 Score Close To Intel Core i7 6800K

[vario]

The main thing that is often overlooked is ipc per core per clock.

So we will need to pay close attention to that.

If ipc of amds 4.0 beats haswell 4.2 then reality is you only need 4.3 to be equal to 4.5.

Saying you expect Xx speeds is pointless without knowing this factor.

The analogy of dropping cores or ht is plausible to an extent.

Example 1.3v gets me 4.5 ht on or 5.0 ht off but we are talking a 4 core chip. That result with 8 cores ht on would likely be halved...making gains from threading on and off less drastic and not as beneficial.

I think RPGwizard compares 1700x and clocks of it to 7700K so kabylake, i doubt zen will have ipc of kabylake :-/ even excluding AVX2.
One more thing is, recent leaks suggests that smt on zen works better than intel ...
Anyhow, i do a lot of encoding and also can have like 2 VM`s open while doing a bunch of other stuff so i want threads, but also gaming performance to be in range with intels , but a 4.2 ghz haswell like performance is pretty much all i need.I dont use SLI and 144hz monitor.
But clocks wise, im pretty sure there will be a wall around 4.0 - 4.3. Expecting 4.8ghz clocks is kinda unrealistic.Im speaking 24/7 prime stable ofc.
Im pretty sure if AMD could do it, they would do 8/16 4.0/4.4 HALO ZEN FX part for 999$ .

[AliG]

Who's to say that won't be eventually coming though? It's not like the Black Edition chips launched immediately with the main product family.

As far as we know a year down the road an 140W X1900 Black Edition at 3.7/4.2 (or whatever) could appear. Right now they are (sensibly) focusing on the main product family to get back market share.

[chew*]

First things first AMD needed to get back on the map. I spent like 4 days at AMD pretesting BD way before launch. I ended up fooling with Llano since the BD potential world record list was short......

Llano offered the IPC AMD needed at the time....any gain over current offering would have been better IMO. I had even suggested fusing 2 llano dies together. Unfortunately there were some things that were not known about Llano and that I am still probably not at liberty to discuss.

Being a extreme ocer and knowing whats best for AMD ( the realworld market ) even knowing Llano oced poorly I still suggested a fused Llano. Its not what I wanted but after testing BD its what AMD needed.

What they appear to be offering now is what they needed back then. A chip that benefits the majority. Ok so maybe they cant hit high clocks. If ipc is on par for 90% of market they don't need to hit high clocks.

[Metroid]

I think people want 4.8-5.0 chips but the reality is we are getting to the point where higher base clock chips are harder to hit targets for manufacturers.

I recall reading some article about some electrical moore's law as we go lower in nm....

It basically states right about where we are now...companies reach a clock threshhold (wall) so at that point they will start adding more cores. And here we are today. 16 to 20 threads will be avg...

So true, well it can't be helped, they have to give a reason for the people to buy, add more cores, slightly improved single thread by 1% ~ 5% and it's a sell and here we are in 2017 from 2006 at most 50% or so on single thread.

https://www.google.com/url?sa=i&rct=...87806854212625

E6600 r15 3100Mhz 80 cb single core, 154 cb multithread.

[vario]

Who's to say that won't be eventually coming though? It's not like the Black Edition chips launched immediately with the main product family.

As far as we know a year down the road an 140W X1900 Black Edition at 3.7/4.2 (or whatever) could appear. Right now they are (sensibly) focusing on the main product family to get back market share.

Im not saying they wont , all im saying AMD cant do them now.
With AMD, we pretty much always get a new revision of the same architecture that clocks better.

[qcmadness]

First things first AMD needed to get back on the map. I spent like 4 days at AMD pretesting BD way before launch. I ended up fooling with Llano since the BD potential world record list was short......

Llano offered the IPC AMD needed at the time....any gain over current offering would have been better IMO. I had even suggested fusing 2 llano dies together. Unfortunately there were some things that were not known about Llano and that I am still probably not at liberty to discuss.

Being a extreme ocer and knowing whats best for AMD ( the realworld market ) even knowing Llano oced poorly I still suggested a fused Llano. Its not what I wanted but after testing BD its what AMD needed.

What they appear to be offering now is what they needed back then. A chip that benefits the majority. Ok so maybe they cant hit high clocks. If ipc is on par for 90% of market they don't need to hit high clocks.

Llano clocks so poorly that it cannot get past 3.5GHz easily.

[chew*]

Llano clocks so poorly that it cannot get past 3.5GHz easily.

Llano was capable of 5 gig plus.........and that was on early crappy samples.....



Llano also had a locked theoretical max multi which with the proper in house tools I used to flip the multiplier range to the high side.

So to clock LLano high without AMD in house tools it requires fsb clocking once reaching the max multi. also if you wanted to overclock llano....you needed an unlocked chip for max.

Llano was more than capable of 4 gig clock speeds and the ipc was beating deneb and thuban easily and BD's IPC in single threaded was worse than thuban and deneb.

I can not and could not control what AMD did I could only suggest. I'm quite sure a few refreshes maybe a respin could have gotten better yields.

Not to mention the idea of fusing 2 APU cores together could have opened up a possibility to possibly Xfire the igp's


The main point I'm trying to make is AMD needed ipc first, clock speed second.

They did this azz backwards with BD and failed miserably

Now they are on the right track. IPC first. I'm quite sure that yields will now be a priority while trying to slightly optimize ipc even more.

If they never hit 7-8 gig again on ln2.....ehh me personally its a small disappointment but the general topic at all the shows I ever did was.......my 24/7 rig does not run on ln2..

[StyM]

[zalbard]

I think that depends on whether we're looking at true 4-core or a cutdown 8-core.

Ryzen's CCX (Core Complex) is a quad core module.

[AliG]

Ryzen's CCX (Core Complex) is a quad core module.

Interesting, so that implies a native quadcore should be rather easy to produce?

[StAndrew]

Ryzen's CCX (Core Complex) is a quad core module.

I'm curious about the interconnecting bus between the CPU cores (High bandwidth low-latency data transfer?).

[Warboy]

So far do to the press releases and everything. These seem fairly accurate. I wonder how its going to OC.

[Ket]

Firstly, that is the funniest thing I've seen all week, and that includes a Hitler intel/Zen spoof on youtube

Second, its nice to see sanity return and the importance of IPC over raw frequency once again realised. You don't need raw speed if the architecture is good. Based on images I've seen for Zen with decent (but not crazy) voltage I'd stick my neck out and say the design of Zen allows for Quad CPUs to hit up to 4.6GHz, Hexes and Octas I'd hazard at 4.4GHz. These numbers are as ever subject to the silicon lottery, cooling, and binning processes. "X" versions will more than likely always clock better in one way or another than non-X versions wether that be requiring less voltage or simply just clocking higher. The exception, and I would not at all be surprised by this, would be the 1800X. It could well be a dark arts horse and manage 4.8GHz with the right cooling. Theres just enough hints here and there that suggest it might well manage it. Watch this space.

[AliG]

Second, its nice to see sanity return and the importance of IPC over raw frequency once again realised. You don't need raw speed if the architecture is good. Based on images I've seen for Zen with decent (but not crazy) voltage I'd stick my neck out and say the design of Zen allows for Quad CPUs to hit up to 4.6GHz, Hexes and Octas I'd hazard at 4.4GHz. These numbers are as ever subject to the silicon lottery, cooling, and binning processes. "X" versions will more than likely always clock better in one way or another than non-X versions wether that be requiring less voltage or simply just clocking higher. The exception, and I would not at all be surprised by this, would be the 1800X. It could well be a dark arts horse and manage 4.8GHz with the right cooling. Theres just enough hints here and there that suggest it might well manage it. Watch this space.

The interesting part is Zen's IPC may actually be a tick slower per core. The 1800X ties the 6900k in single threaded Cinebench despite being clocked higher - but is 9% faster in multi-thread. It shows that AMD found a way to optimize not just the way cores compute, but also the way they interact with each other.

I'm most curious about the power consumption. In the end IPC is somewhat irrelevant if the thermals are out of wack (think GTX 1060 vs RX 480). If AMD managed to produce an octo-core that is not only more powerful than the 6900k, but also more power efficient, I think they'll steal HUGE market share in the server world.

[Ket]

IMO you can't really compare architectures as different as intels and AMDs, you can measure both and come up with a yardstick, or "preferred performance" if you like but not directly compare. I get the temptation to do so though because it always happens. One thing is certain though, AMD are about to nastily slap intel off of their high horse. Its going to be interesting watching the fanboys on both sides

[StAndrew]

The interesting part is Zen's IPC may actually be a tick slower per core. The 1800X ties the 6900k in single threaded Cinebench despite being clocked higher - but is 9% faster in multi-thread. It shows that AMD found a way to optimize not just the way cores compute, but also the way they interact with each other.

Maybe this? >

I'm curious about the interconnecting bus between the CPU cores (High bandwidth low-latency data transfer?).

[Ket]

Maybe this? >

AMD did say the architecture surprised even them with no significant bottlenecks, so the interconnecting bus could be a large part of that. Bandwidth will be interesting to see with even modestly decent RAM speed (say 2800-3000MHz). I'll be taking a look at that for sure, got a kit of G.Skill 3200 C15 sticks at the ready (F4-3200C15D-16GVK).

The interesting part is Zen's IPC may actually be a tick slower per core. The 1800X ties the 6900k in single threaded Cinebench despite being clocked higher - but is 9% faster in multi-thread. It shows that AMD found a way to optimize not just the way cores compute, but also the way they interact with each other.

I'm most curious about the power consumption. In the end IPC is somewhat irrelevant if the thermals are out of wack (think GTX 1060 vs RX 480). If AMD managed to produce an octo-core that is not only more powerful than the 6900k, but also more power efficient, I think they'll steal HUGE market share in the server world.

Forgot to say, Cinebench also seems to lean toward intel CPUs in its tests which could suggest the code isn't as optimised for AMD as it is for intel CPUs so I wouldn't take Cinebench results as gospel for Zen IPC being slightly lower than intels. I've seen a lot of results putting Zen single core perf around 146 with intels at the 150 mark although the latest shots I saw showed Zen at 154, swings and roundabouts.

[dasa]

AMD did say the architecture surprised even them with no significant bottlenecks, so the interconnecting bus could be a large part of that. Bandwidth will be interesting to see with even modestly decent RAM speed (say 2800-3000MHz). I'll be taking a look at that for sure, got a kit of G.Skill 3200 C15 sticks at the ready (F4-3200C15D-16GVK).

i would love if you could expand on that with some cpu limited game tests comparing the gains from overclocking the cpu vs higher ram speeds
on my 6700k most games and even lightroom can scale higher from overclocking the ram than the cpu to 4.7ghz
for example http://www.overclock.net/t/1611359/3...#post_25514506


its a shame it looks like ryzen may not be able to handle much over 3200mhz going by supported memory lists currently available on the mb
so it would be interesting to see just how dependent on ram speed\latency amd is
that may even give us an idea if the l4 cache on future amd quad cores could benefit cpu performance as it did for the 5775c

[Ket]

i would love if you could expand on that with some cpu limited game tests comparing the gains from overclocking the cpu vs higher ram speeds
on my 6700k most games and even lightroom can scale higher from overclocking the ram than the cpu to 4.7ghz
for example http://www.overclock.net/t/1611359/3...#post_25514506

its a shame it looks like ryzen may not be able to handle much over 3200mhz going by supported memory lists currently available on the mb
so it would be interesting to see just how dependent on ram speed\latency amd is
that may even give us an idea if the l4 cache on future amd quad cores could benefit cpu performance as it did for the 5775c

I'll throw what tests I can at the system but time limitations will prevent me from being as thorough as I would like to be. I think, assuming you use the right memory kit, 3400-3600MHz will be the upper limit for Zen, at least with the first version. Refresh or two down the line you would think that "limit" will be fixed. Not that anything over 3600MHz really nets you anything other than epeen and higher synthetic numbers anyway.

[AliG]

I don't think it'll be an order of magnitude difference in either case. And by that I mean Ryzen + slower RAM will not provide any difference in your gaming experience than a 6900k based build (despite costing less than half as much).

[Ket]

I think dasa is just interested to find out what benefits Zen most, higher CPU speed or higher memory speed. Given that according to AMDs own words there are no significant bottlenecks with Zen my inclination would be that you want to clock Zen as high as possible to try and limit any possible bottlenecks Zen may have.

[dasa]

Thanks for taking it into consideration Ket i understand time is tight
Maybe even in the future when more time is available


This could be of interest to some

Originally Posted by Gibbo View Post

A comment from Asus:

"I?ve decided to provide some recommendations on DDR4 limitations concerning AM4 currently.

As it stands the AMD code has restricted RAM tuning options which means many RAM kits at launch will not be compatible. This is the same for our competitors also.
What we recommend is the following:
If fully populating a system with 4 DIMMs (2DPC), use memory up to a max of 2400MHz.
If using 1DPC (2 DIMMs) ensure they are installed in A2/B2 and use memory up to max of 3200MHz.

The indication I have received from HQ is that AMD has focused all their efforts on CPU performance so far and will release updated code in 1~2 months when we expect improved DDR4 compatibility and performance."

In short if filling all 4 DIMM's set your speed to 2400MHz and work up from there.
If using 2 DIMM's put them in the A2/B2 slots and a max of 3200MHz should be possible.

In our testing only the Crosshair board achieved 3000-3200MHz, the others were in the 2400-2666MHz range.

BIOS updates will come!

[chew*]

Well in there testing. They would be dead wrong.

[ZenEffect]

so much speculation, so few samples.

[dasa]

Well in there testing. They would be dead wrong.

i believe they were only referring to asus mb or are you as well?
interesting that the b150 mb from all other manufacturers say they support up to 3200 while only the x370 crosshair mb from asus does
it will be interesting to see how many hit what they claim or higher