Official AMD Ryzen Benchmarks, Reviews, Prices, and Discussion

[Rifter]

Sure assessing power consumption is darn difficult. And will always be dependant on the specific situation.
A way to look at it is this:

Prime autocad metro idle all consistent points the same direction.
Bd is very bad at the same assessments.

Toms usually is quite good at this power stuff. Unfortunately they dont give us the details.
But so what. Look at the results. We know the stilts results and asumptions for lower freq of zen that will be applied outside of desktop.
When we get to 2.8 that is server and notebook base like freq it looks to me this arch is just far superior to intel for efficiency for many loads.
Its a guess but the data we have mostly point this way.

I agree, in the 2-3Ghz 16+ core server parts AMD is going to put the hurt on intel.

 

[raghu78]

We are seeing 3 Ghz at 65W TDP for 8C/16T. AMD will have had time for another revision of the Zeppelin die by the time Naples launches in late Q2 which could improve clocks even more at a given TDP . I think a 2.5-2.6 Ghz Zeppelin could come at 45W TDP. Naples could easily clock 2.5 - 2.6 Ghz base clocks for the flagship 32C/64T model. Skylake is rumoured to be clocking at 2.3 Ghz for the flagship 32C/64T model. The higher clocks combined with the higher bandwidth could help offset some of the IPC deficit against Skylake core. AMD's Naples could turn out very competitive against Intel's best Skylake-EP SKUs while coming in at a disruptive price.

More importantly all of this is with a 1st gen Zen 14LPP design. As we saw from Carrizo to Bristol Ridge AMD can definitely improve clocks and efficiency significantly as GF 14 LPP process matures. Skylake-EP will probably take atleast 2 years to be replaced by 10nm + Icelake-EP. AMD has Zen 2 ready for 2018 and this core will make its way into desktops and servers next year. I think the pressure on Intel will really start to show next year as GF adds 14nm capacity in Q1 2018, improves the process and AMD launches Zen 2 .

 

[scannall]

Its a well known fact that software always lags hardware which is exemplified when considering how long multicore cpu's have been on the market and how many software applications still utilize single thread performance. Even so there's a waterfall effect in Ryzen's cache system that needs to be addressed to speed up the flow of data between them and the cpu and I hope that the next iteration of it is much better in this area. While its still a magnificent leap forward for AMD it cannot dethrone Intel's finest and no amount of fanning will make it happen.

I don't think all that many people were expecting Ryzen to bury Intel in everything. Ryzen does pretty much bury Intel in performance/$$, aside from some edge cases. Back in December people were expecting 40% over the last construction core, and AMD over delivered there. Overall, I think they hit a homerun. To be able to complain about such close performance is in itself a gift. And not something anyone would have thought possible just a few months ago.

 

[AnandThenMan]

The fact that we are even splitting hairs over performance in many games and applications is incredible. AMD has mounted one of the biggest comebacks ever, never thought I'd again see the day where I actually wanted AMD processors in my rigs.

 

[Dresdenboy]

The new version of AIDA64 with improved Ryzen support is available:
https://www.aida64.com/news/aida64-v590-amd-ryzen-benchmarks-latency-cache-speed

 

[psolord]

Has any review tested Forza Horizon 3 on Ryzen?

 

[KompuKare]

So many sites don't actually do a perf/watt calculation despite having the figures.
For instance PCPer used Cinebench load for their power measurements:

And they also of course have the Cinebench scores themselves:

And a bit of time with online spreadsheet gives this

Obviously this kind of chart will vary depending on load but at Cinebench Ryzen has very good perf/watt. The overclocked score isn't that good but for the most important markets (server and laptops) that is not an issue as both will be running well within the sweet spot of the process.
Actually judging from The Stilt's work, Hardware.fr and that Russian site imported_jjj linked to earlier, even stock Ryzen 1700 is still running above the process sweet spot.
I'm sure trawling through other reviews could yield different perf/watt metrics for other loads.

 

[ultima_trev]

Ryzen 1800X downcore to 6 (3+3) at 4.05 GHz versus 8 core at 3.92 GHz...

PassMark CPU

3+3: http://imgur.com/a/kSJ6H

4+4: http://imgur.com/a/yq2Eo

UserBenchmark

3+3: http://www.userbenchmark.com/UserRun/3242643

4+4: http://www.userbenchmark.com/UserRun/3220084

Cinebench

http://imgur.com/a/mUgbM

3DMark FireStrike

3+3: http://www.3dmark.com/fs/12139687

4+4: http://www.3dmark.com/fs/12140931

3DMark Time Spy

3+3: http://www.3dmark.com/spy/1456842

4+4: http://www.3dmark.com/spy/1454132

 

[thepaleobiker]

Ryzen 1800X downcore to 6 (3+3) at 4.05 GHz versus 8 core at 3.92 GHz...

PassMark CPU

3+3: http://imgur.com/a/kSJ6H

4+4: http://imgur.com/a/yq2Eo

UserBenchmark

3+3: http://www.userbenchmark.com/UserRun/3242643

4+4: http://www.userbenchmark.com/UserRun/3220084

Cinebench

http://imgur.com/a/mUgbM

3DMark FireStrike

3+3: http://www.3dmark.com/fs/12139687

4+4: http://www.3dmark.com/fs/12140931

3DMark Time Spy

3+3: http://www.3dmark.com/spy/1456842

4+4: http://www.3dmark.com/spy/1454132

Interesting to see the Firestrike vs TimeSpy inversion of scores. Firestrike liked 3+3 better than 4+4, but TimeSpy went the other way.

Regards,
Vish

 

[WhoBeDaPlaya]

Ryzen 7 1700 for $249.99 shipped. That's used Xeon level of cheapness for the amount of compute power you get

https://hardforum.com/threads/ryzen-7-1700-249-99-after-promo-code-ebay-com.1928291/

 

[thepaleobiker]

Ryzen 7 1700 for $249.99 shipped. That's used Xeon level of cheapness for the amount of compute power you get

https://hardforum.com/threads/ryzen-7-1700-249-99-after-promo-code-ebay-com.1928291/

The offer was dead within 2 hours. It was on FP Deals over at Slickdeals and you can imagine the hordes rushing to order it off of Ebay

They also had the non-K i7 7700 for $219, which caught my eye. That's nearly the same price I paid for my i5 7500.

Regards,
Vishnu

 

[Magic Hate Ball]

Ryzen 1800X downcore to 6 (3+3) at 4.05 GHz versus 8 core at 3.92 GHz...

PassMark CPU

3+3: http://imgur.com/a/kSJ6H

4+4: http://imgur.com/a/yq2Eo

UserBenchmark

3+3: http://www.userbenchmark.com/UserRun/3242643

4+4: http://www.userbenchmark.com/UserRun/3220084

Cinebench

http://imgur.com/a/mUgbM

3DMark FireStrike

3+3: http://www.3dmark.com/fs/12139687

4+4: http://www.3dmark.com/fs/12140931

3DMark Time Spy

3+3: http://www.3dmark.com/spy/1456842

4+4: http://www.3dmark.com/spy/1454132

I noticed an oddity when my 4.05ghz OC was getting reliably lower scores than at 4.00ghz OC in Cinebench. Can you confirm that?

 

[KompuKare]

Okay, I also have some perf/watt figures for Handbrake.
Hexus used Handbrake for the power consumption test in their Ryzen 1700 review:
http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=7
And of course they also used it as CPU performance benchmark:
http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=3
Taken those two figures I came up with the following:

Again, I put entered them on EtherCalc
https://ethercalc.org/kiljnstwel5f

 

[ultima_trev]

I noticed an oddity when my 4.05ghz OC was getting reliably lower scores than at 4.00ghz OC in Cinebench. Can you confirm that?

I'll certainly look in to it. I've noticed that with 8 cores I almost consistently score lower at 4.02 and 4.05 in the combined test in Fire Strike than at 4.0. From what I see so far, RAM speed would help a bit going by the score Alexruiz posted a few pages ago.

I also though I might share my Ryzen CPU (4+4 and 3+3 configs) versus my 4.5 GHz i5 6600K...

Fire Strike

1800X at 4.0: http://www.3dmark.com/fs/12082722

1800X (3+3) at 4.05: http://www.3dmark.com/fs/12139687

6600K at 4.5: http://www.3dmark.com/fs/11666196

Test Details: http://www.3dmark.com/compare/fs/12082722/fs/12139687/fs/11666196

Time Spy

1800X at 4.0: http://www.3dmark.com/spy/1442262

1800X (3+3) at 4.05: http://www.3dmark.com/spy/1456842

6600K at 4.5: http://www.3dmark.com/spy/1193795

Test Details: http://www.3dmark.com/compare/spy/1442262/spy/1456842/spy/1193795

Ryzen 5 is looking like it's going to be beastmode.

 

[zinfamous]

Well the team stated very early they were looking at March for an initial release.

ah, not bad. I'm still waiting to upgrade from my 3GB 280X anyway, to give it a spin...so not until Vega + Ryzen 5 (or 1700), anyway. I am...patient.

 

[Zucker2k]

Okay, I also have some perf/watt figures for Handbrake.
Hexus used Handbrake for the power consumption test in their Ryzen 1700 review:
http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=7
And of course they also used it as CPU performance benchmark:
http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=3
Taken those two figures I came up with the following:

Again, I put entered them on EtherCalc
https://ethercalc.org/kiljnstwel5f

I was leery of the power-consumption numbers of Ryzen all the way from AMD's demo to clueless reviewers using P95 stress-test numbers to report power consumption when Ryzen wasn't running that code optimally. Finally, here's a more realistic result of the actual power Ryzen consumes under full load. The 65w chip is at 113, 95w at 161, while Intel's 140watter is at 132. No free lunch here, folks!

http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=7

 

[richierich1212]

TDP does not equal power consumption.

 

[krumme]

I was leery of the power-consumption numbers of Ryzen all the way from AMD's demo to clueless reviewers using P95 stress-test numbers to report power consumption when Ryzen wasn't running that code optimally. Finally, here's a more realistic result of the actual power Ryzen consumes under full load. The 65w chip is at 113, 95w at 161, while Intel's 140watter is at 132. No free lunch here, folks!

http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=7

Wake up dude. You still dont get the memo dont you?
You are referring to the handbrake test.
Did you notice the performance difference here for the same test?

A 1700 non x beats a 6900. Read again.

A 1700x beats a 6950. Yes the 350 usd cpu beats the 10 core 1800 usd intel. Read again.

Now go look at the power figures again.

http://m.hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=3

I know it wasnt you intention but you just proved both how insanely powerfull zen is and how efficient it is at the same time. Lol.

Free lunch for all that open their eyes to the new reality.

 

[PhonakV30]

I was leery of the power-consumption numbers of Ryzen all the way from AMD's demo to clueless reviewers using P95 stress-test numbers to report power consumption when Ryzen wasn't running that code optimally. Finally, here's a more realistic result of the actual power Ryzen consumes under full load. The 65w chip is at 113, 95w at 161, while Intel's 140watter is at 132. No free lunch here, folks!

http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=7

AMD didn't lie to you , Didn't lie to any consumers.AMD's TDP is the minimum amount of forced cooling needed to keep the CPU under a maximum temp.ofc It's completely different than Intel's definition or power consumption.no wonder someone liked your post.

 

[imported_jjj]

One thing about TDP that folks are failing to notice is that it excludes XFR.
XFR is "with better cooling" so can go above TDP.

 

[Ozzyrulez]

I think this has come up every time AMD has released a CPU.

 

[KompuKare]

I was leery of the power-consumption numbers of Ryzen all the way from AMD's demo to clueless reviewers using P95 stress-test numbers to report power consumption when Ryzen wasn't running that code optimally. Finally, here's a more realistic result of the actual power Ryzen consumes under full load. The 65w chip is at 113, 95w at 161, while Intel's 140watter is at 132. No free lunch here, folks!

http://hexus.net/tech/reviews/cpu/103270-amd-ryzen-7-1700-14nm-zen/?page=7

You know that Hexus like most reviewers take their power readings of the whole system at the wall?
And that none of the system have a 0W idle?

So what you really want is the power deltas between idle and load.
Something like this :

Although that still doesn't tell the full story as what the chip draws is only after the PSU's and the VRM's inefficiencies. Now we don't know the VRMs but the PSU they used is a be quiet Dark Power Pro 11 (1,000W). Which is a Platinum rated supply and we can do better than that since TPU reviewed it:
https://www.techpowerup.com/reviews/beQuiet/DarkPowerPro11_1000W/6.html

Right, not going to account for that whole curve as what the FX9590 used at load (295W) is a lot different than what the i3-7350K did (68W), but if we said they are all on average 90%, then we get something like this:

Okay, for Handbrake the i7-6950K doesn't use 140W, but the 1700 and 1700X are well within their 65W and 95W TPU. The 1800X exceeds it a bit.

Mind you, still no idea how much heat the VRMs waste - rather suspect it's more than 7.6W though.

 

[guachi]

So the 1700 is 23% faster and uses 2% less watts for the system than the 7700k in handbrake. It's also cheaper.

$330 CPU + $0 heatsink
Compared to:
$350 CPU + $30 heatsink.

That's $50 or 13% cheaper.

As someone who uses handbrake a lot (like, a lot a lot) the 1700 seems like a no-brainer. Cheaper AND faster. I could probably undervolt and save even more total energy. Or just have an OC profile and then set it stock for encoding. Since it clocks so low at stock it's actually BETTER than the 1700x or 1800x as it will use less total electricity to get the job done.

Am I supposed to be upset about this?

 

[IEC]

Better in outright performance, better in perf/W and better in $/perf... only loses in max OC (which is irrelevant for 99.9%+ of production users).

It's why I bought two of them.

 

[beginner99]

Am I supposed to be upset about this?

Absolutely not. If you encode a lot a lot, then this is the chip for you to get. And if it actually is commerical or really, really a lot a lot you might want to wait fro the rumored HEDT line based on Naples (server) die. Current leaks indicate about similar speed to 1700 at 16c/32t and 180w TDP. What we see her with power use and multi-threading AMD seems to have a bright future in the server space.

However for high FPS gaming, the 7700k is currently still the better choice if that is the most important factor. I never encode. For what? I can dl the stuff already encoded and free. I also don't render. Where it would be useful would be for work.