AMD Ryzen R7 1800X performance review - TechPowerUp

[Guest]

3 hours ago, MageTank said:

...

Also, I just found out that AMD is working with big game devs. for thier GPU and CPU to improve their performance in gaming, which also tells us that, games will have Ryzen support in future.

[MageTank]

7 hours ago, Zackbare said:

Also, I just found out that AMD is working with big game devs. for thier GPU and CPU to improve their performance in gaming, which also tells us that, games will have Ryzen support in future.

Well, games already "support" Ryzen now. This also doesn't mean much for older titles, which make up the vast majority of games. Sure, this is great for future titles, but not everyone plays newer titles. Some of the largest player-base genre's are MMO's, and they tend to use antiquated engines for years. 

 

The notion that there is some secret programming that will make Ryzen magically better than Intel at gaming, sounds silly to me. To be honest, it's not that far behind as is, and is good enough for most gamers. Unless you are trying to push high refresh rates while gaming (competitive gamers), you won't really need anything better for casual titles. 

 

If AMD can get additional performance for their consumers by working with developers, than that is great. Free performance for the consumer is always good as long as other consumers do not suffer when using rival brands (Cough... Nvidia Gameworks). I just don't expect much to come of it, given what we already have. 

[Jahramika]

On 3/17/2017 at 1:46 PM, PCGuy_5960 said:

Not really... The CCX issue is an issue with the architecture... It is impossible to fix, unless games are coded to use only one CCX.(And thus being limited to 8 threads)

There is nothing to fix it wrecks the whole X99 platform cpus as it was 7needed to do.

[Vode]

Man the power consumption just looks freaking sweet. What the hell man Ryzen has double the cores under the same/lower power envelope as/than 7700k... Maybe the Ryzen chip isn't fullly utilized in many cases?

 

Would like to see some further testing! Especially Mafia III 7700k vs 1800x perf/watt would be interesting.

[Morgan MLGman]

Just now, Vode said:

Man the power consumption just looks freaking sweet. What the hell man Ryzen has double the core under the same/lower power envelope than 7700k... Maybe the Ryzen chip isn't fullly utilized?

 

Would like to see some further testing!

If you want an even more thorough testing, you can only watch this, but the 1800X has around the same power consumption under load as the 7700K, and twice as low at idle according to DigitalFoundry.

 

I highly suggest you watch the video in its entirety, it's worth the time.

 

[MageTank]

1 minute ago, Morgan MLGman said:

If you want an even more thorough testing, you can only watch this, but the 1800X has around the same power consumption under load as the 7700K, and twice as low at idle according to DigitalFoundry.

 

I highly suggest you watch the video in its entirety, it's worth the time.

 

It's funny how I agree exactly with their conclusion. Almost as if they saw my posts on this forum, and put it in video-form, lol. 

 

Very few reviewers I trust, but DF and GN are at the top of my list.

[Morgan MLGman]

1 minute ago, MageTank said:

It's funny how I agree exactly with their conclusion. Almost as if they saw my posts on this forum, and put it in video-form, lol. 

 

Very few reviewers I trust, but DF and GN are at the top of my list.

Exactly, there is no better channel for me than DF for hardware reviews. Richard just knows his shit and more importantly, he knows how to talk about it so it's pleasant to listen/watch

I suggest you give HardwareUnboxed a shot, their testings are always complex and seem accurate.

 

 

I considered switching my 6700K over to a 1700 but decided it's just not worth the hassle. I'd just switch a few FPS in games for more CPU horsepower, I doubt the tradeoff is worth all the time that's associated with a platform switch, reinstalling of W10, overclocking, setting up the platform is a pain itself with Ryzen and I just don't find it compelling enough, with that being said if I were building a PC right now, I'd get the R7 1700 and ASRock X370 Taichi and overclock everything as far as I can... But switching my 4,7GHz 6700K... No point I guess

[Vode]

@MageTank

@Morgan MLGman

Yeah guys I agree GN and DF are very reliable sources. The data is always on point and explained very well; also fun to watch. The testing methodology is excellent. 

 

Only sometimes I find myself dissagreeing with some conclusions they make.

 

DF Retro is the most amazing thing ever BTW.  

[PCGuy_5960]

13 hours ago, Jahramika said:

There is nothing to fix it wrecks the whole X99 platform cpus as it was 7needed to do.

Not even close. X99 destroys it in gaming and comes out on top in single core and multicore performance (in multicore, only when you compare it to the equivalent CPU)

[Morgan MLGman]

9 minutes ago, PCGuy_5960 said:

Not even close. X99 destroys it in gaming and comes out on top in single core and multicore performance (in multicore, only when you compare it to the equivalent CPU)

I wonder why do you keep saying that, Broadwell-E scores nearly exactly the same scores in singlethreaded Cinebench, not even mentioning Haswell-E parts it also wins in multithreaded tasks as the 5960X (Haswell 8C/16T) @6,044GHz lost to a 1800X (Zen 8C/16T) @5,2Ghz in multithreaded Cinebench...

 

[PCGuy_5960]

1 minute ago, Morgan MLGman said:

I wonder why do you keep saying that, Broadwell-E scores nearly exactly the same scores in singlethreaded Cinebench, not even mentioning Haswell-E parts it also wins in multithreaded tasks as the 5960X (Haswell 8C/16T) @6,044GHz lost to a 1800X (Zen 8C/16T) @5,2Ghz in multithreaded Cinebench...

If you overclock both a Ryzen and an X99 CPU on water (Not LN2) the X99 CPU will be faster, because it can clock higher. Ryzen's superior multicore performance (clock for clock) is a result of its CCX design. (Low latency between cores on the same CCX)

BTW, keep in mind, Haswell-E can overclock to 4.6GHz on water, while Ryzen can barely do 4.1 and Ryzen's IPC is on par with Haswell.

[Morgan MLGman]

7 minutes ago, PCGuy_5960 said:

Ryzen's IPC is on par with Haswell.

I just proved that to be incorrect, check the image out. 6850K (Broadwell-E) boosts to around the same value in singlethreaded tasks as the 1800X (around 4GHz) and the 1800X has 4 points more. And Broadwell was a bit over Haswell already in terms of IPC...

 

7 minutes ago, PCGuy_5960 said:

BTW, keep in mind, Haswell-E can overclock to 4.6GHz on water, while Ryzen can barely do 4.1

True, Ryzen 8-cores can't clock as high, but we don't know about the 4/6 core parts yet, you have a 5820K which may do 4,5GHz for example, that doesn't mean a 1600X won't be able to do the same thing and it doesn't mean that a 5960X can do it either... We need to wait to see the whole picture imo.

[PCGuy_5960]

8 minutes ago, Morgan MLGman said:

I just proved that to be incorrect, check the image out. 6850K (Broadwell-E) boosts to around the same value in singlethreaded tasks as the 1800X (around 4GHz) and the 1800X has 4 points more. And Broadwell was a bit over Haswell already in terms of IPC...

The 1800X boosts 1 core to 4.1GHz (XFR) while the 6850K boosts 1 core to 3.8GHz. Watch this video, you will see that clock for clock Ryzen is 5% slower than Broadwell-E in single core and this means that it is on par with Haswell

 

8 minutes ago, Morgan MLGman said:

True, Ryzen 8-cores can't clock as high, but we don't know about the 4/6 core parts yet, you have a 5820K which may do 4,5GHz for example, that doesn't mean a 1600X won't be able to do the same thing imo. We need to wait to see the whole picture imo.

True, but AFAIK, AMD's fab has trouble with overclocks higher than 4GHz, so I don't think that they will clock as high...

[Morgan MLGman]

6 minutes ago, PCGuy_5960 said:

The 1800X boosts 1 core to 4.1GHz (XFR) while the 6850K boosts 1 core to 3.8GHz. Watch this video, you will see that clock for clock Ryzen is 5% slower than Broadwell-E per core and this means that it is on par with Haswell

Check this out: https://ark.intel.com/products/94188/Intel-Core-i7-6850K-Processor-15M-Cache-up-to-3_80-GHz

Intel's own ark site states that a 6850K has a Turbo Boost 3.0 technology up to 4.0GHz (meaning 1-core)

 

XFR does around ~4050MHz-4100MHz, seeing as the 1800X wins with the 6850K by a tiny bit, imo it's safe to say they're around equal

 

6 minutes ago, PCGuy_5960 said:

True, but AFAIK, AMD's fab has trouble with overclocks higher than 4GHz, so I don't think that they will clock as high...

Yeah heard that GloFo has issues with getting their 14nm process higher, though we'll see when they launch, though I don't see why the fact that chips with higher core count (ex. 5960X) overclock worse than chips from the same architecture with a lower core count (ex. 4790K, 5820K, 5930K) which applies to Intel, wouldn't apply to AMD as well

[PlayStation 2]

I like how we're arguing that this is not on par with Kaby Lake when AMD themselves specified around Haswell/Broadwell performance.

[Morgan MLGman]

1 minute ago, Dan Castellaneta said:

I like how we're arguing that this is not on par with Kaby Lake when AMD themselves specified around Haswell/Broadwell performance.

True, the only Ryzen I'd compare to a 7700K would be the 1700 as they cost a similar amount of money and then calculated which offers better value (probably the 1700)

[PCGuy_5960]

3 minutes ago, Morgan MLGman said:

Intel's own ark site states that a 6850K has a Turbo Boost 3.0 technology up to 4.0GHz (meaning 1-core)

That's not how Turbo Boost 3.0 works You need to install a specific driver. Otherwise, the CPU won't boost higher than 3.8GHz...

And if a reviewer didn't say that he used the Turbo Boost 3.0 driver, the CPU didn't boost higher than its Turbo Boost 2.0 speed.

Check this out: (both CPUs were locked at 4GHz)

 

6 minutes ago, Morgan MLGman said:

Yeah heard that GloFo has issues with getting their 14nm process higher, though we'll see when they launch, though I don't see why the fact that chips with higher core count (ex. 5960X) overclock worse than chips from the same architecture with a lower core count (ex. 4790K, 5820K, 5930K) which applies to Intel, wouldn't apply to AMD as well

Not (exactly) true... Most 5820K, 5930K and 5960X CPUs can go up to 4.6GHz and most 4790Ks and 4690Ks can go up to 4.6GHz....

[dexT]

It's easy.

 

If you like videos games get the 7700K.

 

If you do work or want a powerful system get Ryzen 7.

[Vode]

39 minutes ago, PCGuy_5960 said:

If you overclock both a Ryzen and an X99 CPU on water (Not LN2) the X99 CPU will be faster, because it can clock higher. Ryzen's superior multicore performance (clock for clock) is a result of its CCX design. (Low latency between cores on the same CCX)

BTW, keep in mind, Haswell-E can overclock to 4.6GHz on water, while Ryzen can barely do 4.1 and Ryzen's IPC is on par with Haswell.

Need I remind you my CPU costs twice as much and gets wrecked by a 1800x @stock? Sure I can overclock but that headroom is not worth double the amount of money, nevermind the expensive X99 platform.

 

Ryzen is not on par with Haswell, it depends on the workload. You can't just make these generalized claims. 

[Morgan MLGman]

19 minutes ago, PCGuy_5960 said:

That's not how Turbo Boost 3.0 works You need to install a specific driver. Otherwise, the CPU won't boost higher than 3.8GHz...

And if a reviewer didn't say that he used the Turbo Boost 3.0 driver, the CPU didn't boost higher than its Turbo Boost 2.0 speed.

Huh, so the "Turbo Boost 3.0" is a retarded feature that needs a dedicated driver to work at all, gg wp IPC is not the issue of Ryzen, imo the biggest one right now is setting up the platform in a proper way... Even with the IPC between Haswell and Broadwell, it's still good-enough for any task  

19 minutes ago, PCGuy_5960 said:

Check this out: (both CPUs were locked at 4GHz)

 

To be fair I kind of look at all of those videos with a grain of salt, I need a reputable and good reviewer that I know will set the platform up properly (Such as the DigitalFoundry video I linked above), the earlier the benchmark released after launch = the less accurate it is, sadly... He even put a 1700X on the thumbnail instead of the 1800X

19 minutes ago, PCGuy_5960 said:

Not (exactly) true... Most 5820K, 5930K and 5960X CPUs can go up to 4.6GHz and most 4790Ks and 4690Ks can go up to 4.6GHz....

Well, the difference is that I've seen 5820Ks not go beyond 4GHz... The sillicon lottery plays a significant role with Intel's chips, with Ryzen R7 it's basically all around 4GHz, of course you can find a 5820K sample that'll do close to 5GHz

 

You need to remember that with those high-performance setups that have 8-cores such as the 6900K, not everyone overclocks. Those PC's are often used for computational purposes and they work 24/7, power consumption and heat are a huge issue there, one PC with a 6900K can heat up the entire room in a day if running at 100% by a lot...  It's not as simple as in the regular desktop market where overclocking is a common thing...

 

And yeah, maybe X99 wins for gaming, but loses hard in almost any other task when you consider value of the CPU and gaming isn't a primary point of buying X99...

 

$/performance is just incredible, and all of that with a LOT lower power consumption (1800X has lower power consumption overall than a 4C/8T 7700K that's on a newer architecture, let alone Haswell/Broadwell 6/8 cores with 140W TDP...)

 

P.S.: I also have an i7-5820K myself in a workstation-oriented build at home, 6700K is in my main gaming rig

[MageTank]

7 minutes ago, Vode said:

Need I remind you my CPU costs twice as much and gets wrecked by a 1800x @stock? Sure I can overclock but that headroom is not worth double the amount of money, nevermind the expensive X99 platform.

 

Ryzen is not on par with Haswell, it depends on the workload. You can't just make these generalized claims. 

If you take most of the tests we've seen, and average them all out, then yes, it is on par with Haswell. I (and many others) expected this to be the case long before we even saw benchmarks. 

 

IPC isn't a static thing. It varies per test, and it's hard to specifically say "X and Y have Z IPC" unless you are factoring in a total average. The total average puts Ryzen and Haswell on equal footing.

 

EDIT: Before anyone hits me with that "YOU ARE AN INTEL FANBOY" claim again, let me clarify this. Haswell IPC is NOT a bad thing. Broadwell is only 3% faster than Haswell in average IPC at best, and Skylake is only another 3-5% on top of that. The biggest benefit Skylake got over Haswell was DDR4. Sadly, with Ryzens IMC being lackluster, it's hard to see it's real benefits until it gets patched in a BIOS update (If AMD is to be believed that they can do it). I say give it some time, or wait a couple weeks for me to get my hands on it to test for myself.

[Morgan MLGman]

5 minutes ago, MageTank said:

If you take most of the tests we've seen, and average them all out, then yes, it is on par with Haswell. I (and many others) expected this to be the case long before we even saw benchmarks. 

 

IPC isn't a static thing. It varies per test, and it's hard to specifically say "X and Y have Z IPC" unless you are factoring in a total average. The total average puts Ryzen and Haswell on equal footing.

 

EDIT: Before anyone hits me with that "YOU ARE AN INTEL FANBOY" claim again, let me clarify this. Haswell IPC is NOT a bad thing. Broadwell is only 3% faster than Haswell in average IPC at best, and Skylake is only another 3-5% on top of that. The biggest benefit Skylake got over Haswell was DDR4. Sadly, with Ryzens IMC being lackluster, it's hard to see it's real benefits until it gets patched in a BIOS (If AMD is to be believed that they can do it). I say give it some time, or wait a couple weeks for me to get my hands on it to test for myself.

I disagree with the bolded claim because: averaging all scores is not a good idea atm, there are still issues with architecture that are yet to be fixed and the "average" IPC is lower because of that as it scores well in some tests and bad in other tests. If it's at Haswell now, it might be Broadwell in a month, right?

[Kinda Bottlenecked]

4 minutes ago, PCGuy_5960 said:

Not (exactly) true... Most 5820K, 5930K and 5960X CPUs can go up to 4.6GHz and most 4790Ks and 4690Ks can go up to 4.6GHz....

 

Yea I'd believe you if my haswell K chip wasn't such a dog that couldn't even overclock. Every chip has the chance at being dogshit. I wouldn't exclude Ryzen as well. 

 

52 minutes ago, PCGuy_5960 said:

Not even close. X99 destroys it in gaming and comes out on top in single core and multicore performance (in multicore, only when you compare it to the equivalent CPU)

 

I don't see that though. I see the  7700K wrecking everything in game benchmarks. 6950X has got nothing on it. 

 

I mean sometimes haswell e derps and ends up below. 

 

Ryzen is giving me a 5960x for cheap. I don't mind that. That old dog is still competitive till this day.

 

Still missing my multibox bench mark though. 

 

[PCGuy_5960]

5 minutes ago, Morgan MLGman said:

To be fair I kind of look at all of those videos with a grain of salt, I need a reputable and good reviewer that I know will set the platform up properly (Such as the DigitalFoundry video I linked above), the earlier the benchmark released after launch = the less accurate it is, sadly... He even put a 1700X on the thumbnail instead of the 1800X

This guy works with Hardware Unbowed, so his results were very accurate

The 1700X on the thumbnail is there because he probably couldn't find an 1800X image...

12 minutes ago, Vode said:

Ryzen is not on par with Haswell, it depends on the workload. You can't just make these generalized claims. 

It is on par... Take a look at benchmarks...

[Morgan MLGman]

1 minute ago, Pohernori said:

Still missing my multibox bench mark though.

I should have time today when I get home from work, should I downclock to stock settings or is 4,7GHz fine?