Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[unseenmorbidity]

LIve stream Ryzen 1700X Benching + Gigabyte Aorus Setup and Features

He is accepting requests.

 

[looncraz]

Not sure where you said HBC and too lazy too look it up lol
Anyway with DDR4 they would have 30-50GB/s in laptop and Vega is supposed to have larger cache + more to use less BW. My expectation is that they can get to Polaris 11 level of perf in gaming. Maybe 11CUs at 1GHz or so for 1.5RFLOPs and greater utilization in gaming than Polaris.

This is what I said:

That means it makes more sense to use fewer CUs - perhaps only four (256 SPs) and include the high-bandwidth cache (or just a very large L2).

You probably just interpreted "high-bandwidth cache" as high bandwidth memory.

However, it seems AMD is, in fact, going with HBM on APUs - according to Fottemberg, Apple has ordered them: https://semiaccurate.com/forums/showpost.php?p=284628&postcount=7858

That would be a pretty useful means to get HBM into the realm of reasonable applicability for the masses.

 

[looncraz]

I am totally fine with this test as a datapoint. Where I have a problem is when it's not contrasted with the benefit of more cores.

The question isn't "[faster theoretical draw calls] or [slower theoretical draw calls]?" when you pit a 7700k vs Ryzen 7. The correct question is: "[faster theoretical draw calls] or [more cores / higher minimum frames / multithread future proof]?".

Many reviewers have failed to present this choice to the reader/viewer. And I find that disingenuous.

I think the fact that a Core 2 Quad holds back a GTX 970 in BF1 and an FX-8350 doesn't really even hold back a GTX 1070 says everything:

Core 2 Quad Q9300

FX-8300

To be clear: the FX-8300 has an IPC deficit in relation to the Q9300. It does have a frequency advantage, but by nowhere near enough to make up for the additional processing demand.

 

[looncraz]

https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx11-multiplayer-fps

too bad most reviewers don't do tests for more relevant CPU limited scenarios such as multiplayer in current games due to inability to do consistent tests

personally, I've been curious if someone can load up something like BF1 multiplayer on two PCs with different CPUs and just follow each other around and compare #s when they are by each other

I have logged hours of play time on BF4 and BF1 on my i7-2600k @ 4.5Ghz and my stock-clocked RX 480.

I max everything and run 144hz FreeSync. I *do* artificially limit my max-FPS to 120, so I'll need to make some unrestrained runs and turn off Radeon Chill

I bought a Fury so I could test with it as well.

I will then repeat, with both cards, on both Windows 10 and Windows 7, and report 1% lows, 0.1% lows, averages, and will graph frame times.

I will then do that with my Ryzen 7 1700X at stock clocks... and again overclocked.

LOTS of testing. I best get to it...

 

[unseenmorbidity]

I have logged hours of play time on BF4 and BF1 on my i7-2600k @ 4.5Ghz and my stock-clocked RX 480.

I max everything and run 144hz FreeSync. I *do* artificially limit my max-FPS to 120, so I'll need to make some unrestrained runs and turn off Radeon Chill

I bought a Fury so I could test with it as well.

I will then repeat, with both cards, on both Windows 10 and Windows 7, and report 1% lows, 0.1% lows, averages, and will graph frame times.

I will then do that with my Ryzen 7 1700X at stock clocks... and again overclocked.

LOTS of testing. I best get to it...

What software do you use to record those lows?

 

[Spartak]

No. It appears you are not up to the speed with information we know. Polaris 11 16 CU design in Radeon Pro 460 from Macbook Pro consumes under load around 18W of power(GPU only). It is clocked at 907 MHz, and TDP, or rather power gate, in BIOS of the GPU is 35W. The rest is consumed by memory, made from 4 memory cells for 128 bit memory bus. AdoredTV has tested how Polaris 11 with 14 CU design behaves declocked to 850 MHz, and it was around 15-18W of power under load, so there is a pattern here.

Raven Ridge APUs are not using Polaris GPUs, but Vega. Vega GPUs have been optimized to work with higher core clocks at the same thermal envelope compared to previous generations of GPUs.

And now we have Raven Ridge mobile Engineering Sample with 4C/8T design clocked at 3.0/3.3 GHz, and 12 CU's with 1 CU taken out totalling with 11 CU's for 35W TDP for whole APU package. About this engineering sample Canard PC Twitter reported some time ago, already.

Ok but that's basic math. CU units power envelope should scale linear, and the frequency should scale to roughly the third degree.
But nice that the APU is hitting these clocks, I was hoping for the 35W part to be 3.0/3.3 GHz so that's an instant buy! Well, probably this time next year as I'm waiting for 4x PCI3.0 NVMe equipped mini-ITX boards to become available. And slightly lower prices would also be welcome.

 

[guachi]

I own an 8350 and I'm impressed by how well it did at 1440 in those videos. CPU isn't particularly meaningful in most games.

And I was impressed by the low CPU usage of Ryzen. Should enable the chip to keep up with GPUs for some time to come.

I'm even more convinced I should get a Ryzen chip and keep it for 5+ years. I could then just buy a new GPU every other year.

 

[AtenRa]

Not sure where you said HBC and too lazy too look it up lol
Anyway with DDR4 they would have 30-50GB/s in laptop and Vega is supposed to have larger cache + more to use less BW. My expectation is that they can get to Polaris 11 level of perf in gaming. Maybe 11CUs at 1GHz or so for 1.5RFLOPs and greater utilization in gaming than Polaris.

RX 460 has 112GB/s memory Bandwidth, there is no way a 11CU Vega with 30-50GB/s bandwidth will have the same performance as Polaris 11.

But with 50GB/s it will have adequate performance (30fps or more) for 1080p gaming at low/medium IQ settings.

 

[imported_jjj]

However, it seems AMD is, in fact, going with HBM on APUs - according to Fottemberg, Apple has ordered them: https://semiaccurate.com/forums/showpost.php?p=284628&postcount=7858

That would be a pretty useful means to get HBM into the realm of reasonable applicability for the masses.

That could be true but not the way most think.
It is likely Intel CPU and AMD GPU using Intel's silicon bridge packaging technology.
So my point stands that Si interposer and HBM are not a solution and we won't see it in APUs.
If and when AMD has viable solutions, and Intel's solution is viable(although not ideal), AMD will use them.

 

[looncraz]

What software do you use to record those lows?

I'm just capturing frametimes with FRAPS.

DX12 pretty much is a bust, so I'm not bothering with it (outside of Ashes of Singularity). And, Doom, of course will be Vulkan.

 

[Zucker2k]

Apparently, Joker's "5ghz" 7700k was actually running at 4ghz...

https://www.youtube.com/watch?v=VWarC_Nygew

 

[imported_jjj]

RX 460 has 112GB/s memory Bandwidth, there is no way a 11CU Vega with 30-50GB/s bandwidth will have the same performance as Polaris 11.

But with 50GB/s it will have adequate performance (30fps or more) for 1080p gaming at low/medium IQ settings.

Assuming the upper range of memory BW and that Vega increases utilization in a very meaningful way (aside form larger cache and the other ticks that reduce the memory BW needs), plus the proximity to the CPU that's what i am expecting. Ofc Polaris 11 level is a range so if it's 5-10% slower, that's fine.
The biggest risk in that assumption is utilization so FLOPS vs gaming perf and it is a bit unwise to assume a large gain without any proof.

 

[Mockingbird]

That could be true but not the way most think.
It is likely Intel CPU and AMD GPU using Intel's silicon bridge packaging technology.
So my point stands that Si interposer and HBM are not a solution and we won't see it in APUs.
If and when AMD has viable solutions, and Intel's solution is viable(although not ideal), AMD will use them.

More likely, Apple puts Raven Ridge in its MacBook Pro

 

[french toast]

Apparently, Joker's "5ghz" 7700k was actually running at 4ghz...

https://www.youtube.com/watch?v=VWarC_Nygew

I want to see a response from joker, innocent untill proven guilty, although the case against him is starting to look compelling.

 

[imported_jjj]

More likely, Apple puts Raven Ridge in its MacBook Pro

I don't think that at this point based on all the rumors.

 

[moonbogg]

Ryzen CPU's in stock right now on Amazon. I think the bigger problem is motherboards really, but if you want a chip, go get one now.

 

[CatMerc]

Everything except the price of HBM.

Personally, I think it would be best if AMD put a single HBM1 stack on an interposer for a high-end APU SKU or two.

Even just 512MB or 1GB of 128GB/s is better than nothing. And HBM1 is proven and cheaper.

Actually, due to economics of scale, I believe HBM2 is cheaper.

 

[MongGrel]

Personally, I'm not buying into the Ryzen hype myself.

AMD dropped the ball decades ago IMHO.

Trolling is not allowed
Markfw
Anandtech Moderator

 

[unseenmorbidity]

Personally, I'm not buying into the Ryzen hype myself.

AMD dropped the ball long ago IMHO.

k?

 

[MongGrel]

k?

k what ?

 

[Tup3x]

k what ?

"Ok, so what?"

Ryzen CPU's in stock right now on Amazon. I think the bigger problem is motherboards really, but if you want a chip, go get one now.

There are lots of 1700 and 1700X CPUs available here in Finland but motherboards are the problem. There are some pretty crappy B350 boards but no X370. To be honest this is rather unusual situation.

 

[french toast]

Personally, I'm not buying into the Ryzen hype myself.

AMD dropped the ball decades ago IMHO.

If you cant elaborate on that/go into detail or offer balanced and objective arguments for your opinion, just "AMD is rubbish,always was" then i suggest you should head to the Kabylake thread.

 

[Mockingbird]

Ryzen CPU's in stock right now on Amazon. I think the bigger problem is motherboards really, but if you want a chip, go get one now.

My pre-ordered Ryzen 7 1700X would get here for another couple of weeks.

I guess I will be shopping for a motherboard in the meanwhile.

 

[Mockingbird]

An In-Depth Look at Ryzen's Gaming Performance: 16 Games Played at 1080p & 1440p

http://www.techspot.com/review/1348-amd-ryzen-gaming-performance/

 

[Head1985]

Apparently, Joker's "5ghz" 7700k was actually running at 4ghz...

https://www.youtube.com/watch?v=VWarC_Nygew

Thats what i said when i tested tomb raider on 6700k 4.5Ghz.I was getting much higher fps.
https://forums.anandtech.com/thread...and-discussion.2499879/page-142#post-38776846