Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[Genx87]

This is about what I expected. It appears in applications that are not taking advantage of all the threads Ryzen can give it is not up to the 7700k. But when the application can take advantage this chip looks really good. Until games start taking advantage of more threads. This line will be in a deficit to the 7700ks higher clocks and better single thread performance.

But this is a really good start for AMD to get back in the game. Just need software to catch up. And I say this as somebody who has bashed AMD since the K10 failed miserably against the Core 2 launch. And really ramped it up after the Bulldozer debacle.

This looks really promising in the HPC and server world.

 

[lolfail9001]

Okay, The Stilt has officially salvaged my ITX build, i am actually going for it once ITX boards are out. I think i can settle for performance of stock 5960X for power consumption of my i5-6400.

 

[rbk123]

With no real Intel competition, game development has absorbed a LOT of Intel specific optimization. AMD performing well on Intel optimized code is a win, but everyone expects it to be as good or better than Intel on Intel optimized code. A little maturity on the AMD/Bios side and some from the game developers to take advantage of AMD abilities should bear a fair bit of fruit.

Gaming performance is still good enough that Intel fans should benefit by paying less for their chips now and going forward.

 

[alexruiz]

https://www.servethehome.com/amd-ryzen-7-1700x-linux-benchmarks/
2670v1 is a SB 8C/16T 2.6Ghz base, 3.3Ghz turbo.

So you mean a single Ryzen 7 1700X is almost as good as 2 of the E5-2670 v1?
That is what I thought.

 

[R0H1T]

Not sure if already posted, SMT impact on gaming both CPUs at 3GHz with DDR4-2400 15-15-15-35 http://www.hardware.fr/getgraphimg.php?id=439&n=1

So if I read that (french) chart correctly SMT gives you less fps?

 

[KTE]

@anandtech

This review format is really a big let down from your prestige.

Seriously, this isn't what AT was known for.

Within the first pages traditionally was IPC vs. the last 2-3 AMD archs. Quick, down and dirty.

And/or power at the 12v.

Then a clearcut manipulation of some parameters to show the difference.

Then like an OC page.

This has nothing.

No clearcut Phenom/Bulldozer compare to showcase AMD improvements.

No clearcut Haswell/Broadwell/Skylake IPC compare to show the same.

Just generic SKU vs SKU rollout.

 

[tamz_msc]

https://www.servethehome.com/amd-ryzen-7-1700x-linux-benchmarks/
2670v1 is a SB 8C/16T 2.6Ghz base, 3.3Ghz turbo.

Do you even read your own charts?

 

[USER8000]

LMAO,has anyone noticed these comments from Legitreviews:

http://www.legitreviews.com/amd-ryzen-7-1800x-1700x-and-1700-processor-review_191753/15

You hit the nail on the head on all your comments. I really wish the Windows drivers were ready before we were given the parts to review. Instead we got a statement 24 hours before launch from AMD saying that they'll be coming in 30 days if all goes well. Game optimizations will be hit or miss when they come, but they appear to be coming. That takes time though and we'll see what happens.

I put it in the conclusion on the last page. The quote came direct from AMD's John Taylor. There was talk of it coming with Ryzen 5 and then they said 'in the next month' in an official statement that was e-mailed out last night.

So you launch a CPU with:
1.)SMT issues in games which you keep quiet about which reduce gaming performance
2.)Launch the CPU one month before windows drivers
3.)Launch with motherboards which just about have stable BIOSes

A great way to sell your product,AMD. Such an own goal.

 

[french toast]

1700 vs 7700k:

well we need more games tested but my pre release prediction of " out of 15 games tested 7700k would win 10 at stock, whilst being reversed at 4ghz and 5ghz respectively " Didn't quite hold up, BUT it was practically a tie at both 1080p and 1440p, i feel pretty good right now about this processor being the best perf/$ along side the intel pentium which equally awesome value.

Judging by the fact 1800x seems to only get 200mhz overclock advantage over r7 1700 you are getting such a bargain right now with r7 1700.

 

[mxnerd]

AMD stock drops 5% today. Ryzen is very, very good, but real world performance did not meet original high expectation.

 

[tamz_msc]

@anandtech

This review format is really a big let down from your prestige.

Seriously, this isn't what AT was known for.

Within the first pages traditionally was IPC vs. the last 2-3 AMD archs. Quick, down and dirty.

And/or power at the 12v.

Then a clearcut manipulation of some parameters to show the difference.

Then like an OC page.

This has nothing.

No clearcut Phenom/Bulldozer compare to showcase AMD improvements.

No clearcut Haswell/Broadwell/Skylake IPC compare to show the same.

Just generic SKU vs SKU rollout.

Give Ian a break. He has put out the most objective review, with a complete revamp of the benchmarking suite, minus gaming, all done for over a week in a friggin' hotel room. He's set the standards other outlets are only wish they'd match.

 

[Insomniator]

LMAO,has anyone noticed these comments from Legitreviews:

http://www.legitreviews.com/amd-ryzen-7-1800x-1700x-and-1700-processor-review_191753/15





So you launch a CPU with:
1.)SMT issues in games which you keep quiet about which reduce gaming performance
2.)Launch the CPU one month before windows drivers
3.)Launch with motherboards which just about have stable BIOSes

A great way to sell your product,AMD. Such an own goal.

That's been their MO going back to the 90's. Many people swore off AMD chips.

 

[scannall]

Hopefully by the time they start testing the R1600 and below, they re-do the R1800 to see if new bioses or whatever, fixes up some of these problems.

Also, is Ryzen suffering in all games, or just some select ones?

What I am also interested in seeing is the debate between 6-8 cores vs 4 cores in gaming, using 6-8 cores from both AMD & Intel.

A few days ago, many people seemed convinced that 4 core had seen its best days and 6-8 cores were poised to take over. Is this still the view amongst these people?

For my use anyway, doing a lot of rendering I need cores. And for gaming, it's 4k. I expect that with the Ryzen I'll stay GPU bound for quite a bit longer than a 4 core processor. I don't replace my computer very often, so I am looking at the long haul. Spending another 600-700 bucks to go Intel 8 core would be silly. I'd rather buy another GTX 1080 and run SLI with the difference and have a bit left over.

If a person doesn't do much of anything that requires a lot of cores, and replaces their computer every 2 years then I'm sure an i7 is plenty. Different people, different use cases.

 

[Topweasel]

This result is about where I expected it. So I will be building a Ryzen system. But will be looking hard at the 1700 but probably will get the 1700x. No use throwing more money at SP/Gaming performance if the difference is going to be so small and it won't improve it's standing with the 1800x. I do want to comment that it looks a little like what happened with Bulldozer or even the P4 with SMT. Windows didn't know how to efficiently assign threads and SMT is probably killing performance a little right now. A hot fix or update to the scheduler and I expect a nice boost in game performance.

 

[tamz_msc]

LMAO,has anyone noticed these comments from Legitreviews:

http://www.legitreviews.com/amd-ryzen-7-1800x-1700x-and-1700-processor-review_191753/15





So you launch a CPU with:
1.)SMT issues in games which you keep quiet about which reduce gaming performance
2.)Launch the CPU one month before windows drivers
3.)Launch with motherboards which just about have stable BIOSes

A great way to sell your product,AMD. Such an own goal.

While I understand the sentiment, always remember:
Damned if you do, damned if you don't.

 

[Det0x]

http://www.tomshardware.com/reviews/amd-ryzen-7-1800x-cpu,4951-5.html

The Cache Testing Dilemma

The Ryzen reviewer's guide contained an interesting note regarding L1, L2, and L3 cache measurement tools. AMD indicates that AIDA64 and SiSoftware Sandra, both commonly-used tools for cache testing, are "not yet equipped to accurately measure cache performance of the Zen architecture." AMD provided its own internally-measured reference values and noted that it is working with the FinalWire (AIDA) and SiSoftware teams to facilitate accurate Zen cache measurement methodology in the future.

We measured performance with the utilities and achieved similar results for Intel's Core i7-6900K, but we also noticed a large gap between the AMD-provided Ryzen measurements and our test results. Ryzen’s L3 cache latency measured 20 ~ 23ns, which is double the provided value. Due to some of the performance characteristics we noted during our game testing, we also tested with SMT enabled and disabled, but the results fell within expected variation. We also measured a ~10ns memory latency gap in favor of the Intel processor.

Many common utilities write zeros to the cache to measure performance. AMD responded to our inquiries and stated that Intel coalesces incoming "zero" write traffic before it passes it to the cache, which could yield artificially high cache throughput measurements, particularly because those patterns don't exist in real-world usage. In our opinion, changing the access pattern would result in reduced performance measurements for the Intel processors, but it certainly shouldn't boost AMD's cache measurements. AMD responded that the current utilities are also not optimized for Zen's unique architecture, and optimizing the utilities' code paths will expose more performance.

We contacted both SiSoftware and FinalWire to request any updated or beta versions of their utilities that can facilitate accurate testing. There is still collaboration between the company's and AMD to resolve the issue, but unfortunately, we aren't at liberty to discuss the details of those conversations. Until that time, and until the vendors and AMD agree on what's happening, we've determined that drawing any conclusions from these results would be irresponsible, and ultimately misleading. When we have final, verified, and real numbers, we'll provide them.

 

[lolfail9001]

With no real Intel competition, game development has absorbed a LOT of Intel specific optimization.

Like usage of fast memory access. What a cheat!

Anyways, i will tease bjt2 with all due respect, because Zen looks to be clocked beyond it's efficiency curve on 1700x/1800x SKUs. Way beyond

 

[cytg111]

LMAO,has anyone noticed these comments from Legitreviews:

http://www.legitreviews.com/amd-ryzen-7-1800x-1700x-and-1700-processor-review_191753/15





So you launch a CPU with:
1.)SMT issues in games which you keep quiet about which reduce gaming performance
2.)Launch the CPU one month before windows drivers
3.)Launch with motherboards which just about have stable BIOSes

A great way to sell your product,AMD. Such an own goal.

I dont care, I like it .. of course I am not the one with a preorder .. But having it evolve in the hands of reviewers, sure, I want that.
Its has gone from like Beta to Release Candidate 1 today and early adopters will roll with RC1.
Even if gaming perf stays the same its still one hell of a start.. Bodes well going forward and the post quad core area is there to stay. AMD is back in business.

 

[hotstocks]

Thank god I woke up early enough to read 6 reviews and cancel my Amazon order. Saved me $1000. My 4.7ghz Sandy Bridge that is 6 years old still destroys Ryzen at games. On top of that all the fails depending on which mobo and ram you have purchased, possible software bugs, needing to disable parts of the chip, WTF? Well I am no Intel fanboi, I have owned most processors made by Amd and Intel over the last 20 years, but I sure as hell thought Ryzen would do better in games, or at least be able to overclock to 4.5ghz and then do better in games. Oh well, my next cpu upgrade still will be a 6 or 8 core cpu by either manufacturer once they are bug free, reasonably priced, AND CAN BEAT A 6 YEAR OLD CPU, LOL. EPIC FAIL. And I am not trolling, if you never play games and only do content creation AND are willing to deal with bugs and early adopter shit, then yeah, go ahead and build a Ryzen system.

 

[lolfail9001]

We measured performance with the utilities and achieved similar results for Intel's Core i7-6900K, but we also noticed a large gap between the AMD-provided Ryzen measurements and our test results. Ryzen’s L3 cache latency measured 20 ~ 23ns, which is double the provided value. Due to some of the performance characteristics we noted during our game testing, we also tested with SMT enabled and disabled, but the results fell within expected variation. We also measured a ~10ns memory latency gap in favor of the Intel processor.

Many common utilities write zeros to the cache to measure performance. AMD responded to our inquiries and stated that Intel coalesces incoming "zero" write traffic before it passes it to the cache, which could yield artificially high cache throughput measurements, particularly because those patterns don't exist in real-world usage. In our opinion, changing the access pattern would result in reduced performance measurements for the Intel processors, but it certainly shouldn't boost AMD's cache measurements. AMD responded that the current utilities are also not optimized for Zen's unique architecture, and optimizing the utilities' code paths will expose more performance.

Your red text is invalidated by paragraph beforehand, even AMD's own tools put Ryzen behind 6900k on these departments.

 

[dfk7677]

What are you looking at that I'm not able to see?

Half the cpu usage at the same framerate.

 

[formulav8]

But this is a really good start for AMD to get back in the game. Just need software to catch up. And I say this as somebody who has bashed AMD since the K10 failed miserably against the Core 2 launch. And really ramped it up after the Bulldozer debacle.

This looks really promising in the HPC and server world.

Someone who gets it.

 

[IntelUser2000]

Yea, so the performance outlier in games and other benchmarks are entirely because of poor latency results for cache and memory. It's like a GPU in the aspect that it has lot of memory bandwidth but probably not good latency. I was little too hasty in judging it had a very good memory controller.

Comparisons:
http://www.guru3d.com/articles_pages/core_i7_6950x_6900k_6850k_and_6800k_processor_review,12.html
http://www.guru3d.com/articles_pages/amd_ryzen_7_1800x_processor_review,13.html

(The tested chip is 10 core on the 6950X so you'll have to compensate that on bandwidth front. The focus is latency however)

I am going to use the 3000MHz memory results as the L3 latency test on Ryzen is wonky at 2133MHz.

(Lower score is better)

6950X
Memory: 57.9ns(~70%)
L1 cache: 1.1ns(roughly equal)
L2 cache: 4.6ns(~80%)
L3 cache: 16.6ns(78%)

Ryzen 7 1800X
Memory: 83.9ns
L1 cache: 1.1ns
L2 cache: 5.8ns
L3 cache: 21.4ns

The results on LGA115x platforms are even lower than Broadwell-E. Tomshardware is saying AMD told them current benchmarks aren't accurate in showing cache/memory results for Ryzen, but low performance in memory reflects some benchmarks so, I have to say they are pretty accurate. It's to be seen if it genuinely sucks or there's like a bug or something.

 

[Topweasel]

That's been their MO going back to the 90's. Many people swore off AMD chips.

Or how every piece of new architecture has been released. The Pentium 4 was released with Memory sockets disabled because of issues 3-4 sticks of memory that they never got over and when they released Hyperthreading it stopped thread locking and smoothed game performance. But it actually hurt several tasks because the Scheduler didn't know how to assign threads correctly to the CPU. Disabling SMT actually gave you about 10-15% jump in performance back in the day. Bios's on a new chip and new CPU will always be under developed. They need all sorts of tests per board, per chip, per memory stick. Always expect 4-5 bios updates a month for 3-4 months and then it level off. The only reason we haven't seen that lately has been because Intel has been massaging the same arch and same chipset for 6 years.

All new Windows devices have these issues. They find bugs and either try to work it out with drivers and FW updates or work with Microsoft to release an update. Or if you Creative Labs you just shrug your shoulders and tell your customers that they can suck it.

 

[R0H1T]

http://www.tomshardware.com/reviews/amd-ryzen-7-1800x-cpu,4951-5.html

Many common utilities write zeros to the cache to measure performance. AMD responded to our inquiries and stated that Intel coalesces incoming "zero" write traffic before it passes it to the cache, which could yield artificially high cache throughput measurements, particularly because those patterns don't exist in real-world usage. In our opinion, changing the access pattern would result in reduced performance measurements for the Intel processors, but it certainly shouldn't boost AMD's cache measurements. AMD responded that the current utilities are also not optimized for Zen's unique architecture, and optimizing the utilities' code paths will expose more performance.

The last version of AIDA64 perhaps addressed this ~

New in AIDA64 Extreme 5.80.4081 Beta (February 21st, 2017)

Cache & Mem Bench / latency method selection via right-click menu
EastRising ER-OLEDM032 (SSD1322) OLED support
preliminary GPU information for nVIDIA GV100
Intel Processor Number detection for Xeon E3-1265L
motherboard specific sensor info for Dell XPS 15 9560
motherboard specific sensor info for EVGA E279
fixed: concurrent disk activity measurement (aquasuite 2017 issue)
fixed: motherboard specific sensor info for Apple Mac mini 6.x (Late 2012)
fixed: motherboard specific sensor info for MSI 200-Series
fixed: motherboard specific sensor info for MSI AM4 Series