Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[CatMerc]

Always has been, for AMD. Have you looked at the bandwidth numbers though? Also, have you considered the effects of IF speed on memory latency? I don't think any of us fully understand the effects of the IF on latency, since it isn't adjustable independently of RAM clock.

Just because it has always been this way doesn't mean it's fine or shouldn't get attention and improved upon.

 

[DrMrLordX]

Just because it has always been this way doesn't mean it's fine or shouldn't get attention and improved upon.

If you look at the horrific bandwidth and latency numbers for Bristol Ridge's IMC, you will see that they HAVE improved upon it. Hell I can get sub-70ns RAM latency times on my RAM no problem. It seems evident to me that IF frequency may be holding it back. Show me Pinnacle Ridge running DDR4-4000 and you will see something very interesting indeed.

 

[Dufus]

Well, I can't find a link right now, but that's ONE piece of software. Most others, SMT on Ryzen is more gain than HT on Intel.

Hope this will help. A quote from Agner Fog

"http://www.agner.org/optimize/blog/read.php?i=838&v=t

"The Ryzen has a micro-operation cache which can hold 2048 micro-operations or instructions. This is sufficient to hold the critical innermost loop in most programs. There has been discussions of whether the Ryzen would be able to run four instructions per clock cycle or six, because the documents published by AMD were unclear at this point. Well, my testing shows that it was not four, and not six, but five. As long as the code is running from the micro-operations cache, it can execute five instructions per clock, where Intel has only four. Code that doesn't fit into the micro-operations cache run from the traditional code cache at a maximum rate of four instructions per clock. However, the rate of fetching code from the code cache is not 32 bytes per clock, as some documents seem to indicate, but mostly around 16 bytes per clock. The maximum I have seen is 17.3 bytes per clock. This is a likely bottleneck since most instructions in vector code are more than four bytes long.

The combination of a compare instruction and a conditional jump can be fused together into a single micro-op. This makes it possible to execute a tiny loop with up to six instructions in one clock cycle per iteration. Except for tiny loops, the throughput for jumps is one jump per two clock cycles if the jump is taken, or two not-taken jumps per clock cycle.

256-bit vector instructions (AVX instructions) are split into two micro-ops handling 128 bits each. Such instructions take only one entry in the micro-operation cache. A few other instructions also generate two micro-ops. The maximum throughput of the micro-op queue after the decoders is six micro-ops per clock. The stream of micro-ops from this queue are distributed between ten pipelines: four pipes for integer operations on general purpose registers, four pipes for floating point and vector operations, and two for address calculation. This means that a throughput of six micro-ops per clock cycle can be obtained if there is a mixture of integer and vector instructions.

Let us compare the execution units of AMD's Ryzen with current Intel processors. AMD has four 128-bit units for floating point and vector operations. Two of these can do addition and two can do multiplication. Intel has two 256-bit units, both of which can do addition as well as multiplication. This means that floating point code with scalars or vectors of up to 128 bits will execute on the AMD processor at a maximum rate of four instructions per clock (two additions and two multiplications), while the Intel processor can do only two. With 256-bit vectors, AMD and Intel can both do two instructions per clock. Intel beats AMD on 256-bit fused multiply-and-add instructions, where AMD can do one while Intel can do two per clock. Intel is also better than AMD on 256-bit memory writes, where Intel has one 256-bit write port while the AMD processor has one 128-bit write port. We will soon see Intel processors with 512-bit vector support, while it might take a few more years before AMD supports 512-bit vectors. However, most of the software on the market lags several years behind the hardware. As long as the software uses only 128-bit vectors, we will see the performance of the Ryzen processor as quite competitive. The AMD can execute six micro-ops per clock while Intel can do only four. But there is a problem with doing so many operations per clock cycle. It is not possible to do two instructions simultaneously if the second instruction depends on the result of the first instruction, of course. The high throughput of the processor puts an increased burden on the programmer and the compiler to avoid long dependency chains. The maximum throughput can only be obtained if there are many independent instructions that can be executed simultaneously.

This is where simultaneous multithreading comes in. You can run two threads in the same CPU core (this is what Intel calls hyperthreading). Each thread will then get half of the resources. If the CPU core has a higher capacity than a single thread can utilize then it makes sense to run two threads in the same core. The gain in total performance that you get from running two threads per core is much higher in the Ryzen than in Intel processors because of the higher throughput of the AMD core (except for 256-bit vector code)."

The instructions per clock (IPC) will be somewhat dependent on the software optimization. Note:- per clock, per cycle and per clock cycle here all mean the same thing.

 

[tamz_msc]

Negative SMT scaling for AMD on Euler3D isn't something to be concerned about too much because all CFD algorithms have high inter-thread dependency and are extremely sensitive to the memory hierarchy. I wouldn't be surprised if that behavior is due to something else, considering the memory subsystem dependency on IF in Ryzen, with less than optimal performance being merely exaggerated by the presence of SMT.

 

[tamz_msc]

If you look at the horrific bandwidth and latency numbers for Bristol Ridge's IMC, you will see that they HAVE improved upon it. Hell I can get sub-70ns RAM latency times on my RAM no problem. It seems evident to me that IF frequency may be holding it back. Show me Pinnacle Ridge running DDR4-4000 and you will see something very interesting indeed.

I too am inclined to think that the biggest gains in Pinnacle Ridge will have to be made with the IMC, after all, we've seen with fast low-latency memory how "gaming IPC" pans out in some recent testing even with a 25% clock speed deficit.

 

[.vodka]

Well, those tuned timings sure have a great effect on memory performance as it's been shown lately.


From my own experience, left is The Stilt's DDR4-3466 profile (needs 1.42v), right is XMP profile (1.35v) for these.

Reviewers at most set the XMP profile and test away... of course there's going to be lackluster performance (especially with stuff sensitive to latency, like games) with that much being left on the table.


Yeah, hopefully Pinnacle Ridge's memory controller improves latency below the ~70ns best case for Summit Ridge and also extracts better performance from looser memory timings if possible... but I can't help but wonder if that will come with 7nm Ryzen instead of the 14nm+/12LP step until then...

On AMD's defense, this is their first DDR4 controller and there's probably lots of stuff to tweak, Intel has been polishing theirs in HSW-E, BDW, BDW-E, SKL/KBL/CFL, SKL-X and have been doing a magnificent job on it.

 

[Chicken76]

On AMD's defense, this is their first DDR4 controller ...

No it isn't. Excavator had DDR3 & DDR4 memory controllers.

 

[.vodka]

Yes it is. That DDR3/4 controller was made by Synopsys, with a pretty hard DDR4-2400 limitation that doesn't benefit these bandwidth starved APUs at all. Same thing for Kaveri, it had a pretty funky GDDR5/DDR3 controller that never had its GDDR5 capabilities used anywhere. Piledriver/Trinity were the last chips that had an AMD IMC.

The one in Summit Ridge is AMD's.

 

[krumme]

Yes it is. That DDR3/4 controller was made by Synopsys.

The one in Summit Ridge is AMD's.

Okey. Interesting. Then it might be quite different. Lets hope to the better. Lol

 

[.vodka]

AMD marketing doing it right for once, lol

https://twitter.com/AMDRyzen/status/915977053463343104

 

[psolord]

You instantly know that these are not the same guys that came up with "poor volta"! xD

 

[Asterox]

Microcenter, R5 1600 only 170$.

http://www.microcenter.com/product/..._AM4_Boxed_Processor_with_Wraith_Spire_Cooler

 

[CatMerc]

I'd pick it up in a heartbeat if I was a US citizen lmao
Cheapest one here is 290$

 

[repoman0]

I finally convinced myself to pay $185 for a 3200/14 16GB kit I was just reusing a cheap Gskill 2400MHz kit I bought with Skylake a couple years back, which ran 2400MHz for a while but with the latest 1.0.0.6b BIOS update is stuck at 2133. It is a little bit surprising to me how large the difference even between 2133 and 2400 feels in the only game I'm playing at the moment, PUBG ... stutters and minimums at the slower speed are much worse and more frequent.

I really wanted to grab a 32GB kit, but with these prices :O I paid under $50 for this cheap-o GSkill kit ...

Well, got it installed and stable on my cheapo AB350M Pro4 board ... took a bunch of tweaking but it works. Using the_stilt's b-die timings was key, the XMP profile was useless and just caused no boot. I could mess with it more to see if some of the other changes I made while I was trying to make it work allow the XMP profile to work but I'm happy with where it is. The kit is actually the TridentZ one linked right above.

I haven't done any benchmarks but the difference is obvious in PUBG. All the little stutters and hitches that I thought were caused by graphics settings being too high are gone, it's just totally smooth. I can probably turn a bunch of options back up

I ended up flashing back to the 1.0.0.6 BIOS on this board (from 1.0.0.6b) while trying to get it to boot. So the AGESA updates are not a magic fix, just need to be able to set the subtimings

 

[.vodka]

Oh, this is getting good

https://twitter.com/AMD_UK/status/916256740890443776

This new AMD is such a breath of fresh air over the previous era!

 

[rainy]

AMD marketing doing it right for once, lol

Oh, this is getting good

This new AMD is such a breath of fresh air over the previous era!

That make me curious: why marketing of CPU division is so much better lately than RTG?

 

[R0H1T]

I'd pick it up in a heartbeat if I was a US citizen lmao
Cheapest one here is 290$

US citizen, not deported to Mexico (or Canada?) & living near a Microcenter. That's a whole lot of conditions you need to fulfill.

That make me curious: why marketing of CPU division is so much better lately than RTG?

It generally goes with the product, RTG was a star as late as the days of 390/x & then it all went downhill.

 

[Despoiler]

That make me curious: why marketing of CPU division is so much better lately than RTG?

Probably because they gutted the GPU division to stay afloat enough to get Ryzen out the door. AMD listened to the analysts that said GPU sales were waning. Their budget was cut because of that. Those analysts were wrong of course. Also, AMD didn't have the money to fund both sides. Intel is the weaker of the two opponents because they sat on their hands for the entire run of Bulldozer. Nvidia kept executing like clockwork. Easy business call on where to focus resources and hire/retain talent.

 

[raghu78]

Probably because they gutted the GPU division to stay afloat enough to get Ryzen out the door. AMD listened to the analysts that said GPU sales were waning. Their budget was cut because of that. Those analysts were wrong of course. Also, AMD didn't have the money to fund both sides. Intel is the weaker of the two opponents because they sat on their hands for the entire run of Bulldozer. Nvidia kept executing like clockwork. Easy business call on where to focus resources and hire/retain talent.

You are wrong. Intel is a very strong company both financially and technically. I do not deny that they got complacent because of AMD's failures with Bulldozer. But the real reason for AMD to divert resources to Zen is simple. the Total Accessible Market for the x86 desktop and notebook CPU/APU market combined with the x86 server CPU market is an order of magnitude more than the GPU market. Thats clearly obvious from the size of Intel and their revenues. AMD as a company can succeed and thrive only if they have a competitive CPU business. The GPU business is simply not enough to sustain AMD.

 

[Despoiler]

You are wrong. Intel is a very strong company both financially and technically. I do not deny that they got complacent because of AMD's failures with Bulldozer. But the real reason for AMD to divert resources to Zen is simple. the Total Accessible Market for the x86 desktop and notebook CPU/APU market combined with the x86 server CPU market is an order of magnitude more than the GPU market. Thats clearly obvious from the size of Intel and their revenues. AMD as a company can succeed and thrive only if they have a competitive CPU business. The GPU business is simply not enough to sustain AMD.

I would say that is more of a third consideration than being wrong. If you have little to no chance to compete in a market with a larger TAM would you still attempt it? The CPU is actually something they can take significant chunks of because Intel got sloppy. TAM only matters based on how competitive you are. GPUs....not happening without gobs more money. They are squeezing blood from a rock with GPUs right now.

 

[raghu78]

I would say that is more of a third consideration than being wrong. If you have little to no chance to compete in a market with a larger TAM would you still attempt it? The CPU is actually something they can take significant chunks of because Intel got sloppy. TAM only matters based on how competitive you are. GPUs....not happening without gobs more money. They are squeezing blood from a rock with GPUs right now.

AMD can design competitive CPUs and GPUs as we have seen in the past with K7, K8, Jaguar and Radeon HD 4000/5000/6000/7000 series. AMD historically had 30-35% of the GPU market but still that was not enough to bring sustained profitability as the CPU business was shrinking rapidly after the Bulldozer disaster. With a competitive CPU architecture and a strong roadmap AMD can very well get to a long term PC market share of >=30% in desktops, notebooks and servers. That will bring consistent profitability and will allow AMD to spend sufficiently for CPU and GPU R&D.

 

[Mopetar]

You instantly know that these are not the same guys that came up with "poor volta"! xD

That make me curious: why marketing of CPU division is so much better lately than RTG?

Having a good product goes a long way. If Vega had the same kind of performance gains as Ryzen and took everyone by surprise, then "Poor Volta" would have been a great piece of marketing. There just isn't a whole lot of exciting things to say about a product that currently can be described as mediocre at best.

Meanwhile, even some quick throwaway jabs seem brilliant because Ryzen is quite competitive with Intel in a large number of segments and even comes out looking better in a few. How many memorable marketing campaigns can you think of for products that sucked?

 

[Puffnstuff]

Somehow with the release of the new intel chips I don't believe that AMD will be able to find a place in my next build.

 

[Schmide]

Somehow with the release of the new intel chips I don't believe that AMD will be able to find a place in my next build.

Then why are you posting in an AMD thread!!!

 

[Puffnstuff]

Then why are you posting in an AMD thread!!!

Because I was interested in the ryzen but things have since changed and I will leave the fan boys to their chips.

Trolling is not allowed
nathanddrews
Anandtech Moderator