AMD Ryzen R7 1800X performance review - TechPowerUp

[MageTank]

Just now, Morgan MLGman said:

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. If it's at Haswell now, it might be Broadwell in a month, right?

Why is it not a good idea at the moment? What information has been presented that directs us towards a potential fix in these tests? We simply cannot take AMD for their word that the IMC is going to be fixed (if you knew the exact nature of the IMC's deficiency, you would be skeptical too), and a lot of the non-memory bound tests still put it at or around Haswell. 

 

Understand this. We cannot look at every single product to ever be released, with the doubt in the back of our minds that the information won't be concrete, or that it won't matter until X amount of time passes. We do not do this when comparing new Intel releases against their older ones, and we cannot make an exception for AMD. We speak in the present at all times, because the present is what matters for consumers. Anyone buying a product under the pretenses that it "might" get better in the future, are setting themselves up for disappointment. 

 

Until AMD (or any third party benchmark/application) releases patches for Ryzen that improves it's performance, we shall use what we currently have as the basis for our information. With that in mind, claiming "Ryzen is on par with Haswell in terms of IPC average" is simply the most accurate way to describe it's relative performance in general. As I said before, this is not a bad thing. Haswell is still a very competent architecture, and is plenty fast for consumers.

[Kinda Bottlenecked]

Just now, Morgan MLGman said:

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

 

Oh no worries. Please do it at your own convenience. Stock would be nice. Since that would be the clock speed that I'm guaranteed to get on the i7s.

[PCGuy_5960]

2 minutes ago, Pohernori said:

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.

The 7700K is definitely faster than X99 in games. But X99 is also faster than Ryzen in games.

Haswell-E derps because of its low clock speed. I mean, the 5960X runs at 3.0GHz which is very misleading because you can EASILY overclock it to 4.4GHz+ and get much better performance. 

[Morgan MLGman]

2 minutes ago, MageTank said:

Why is it not a good idea at the moment? What information has been presented that directs us towards a potential fix in these tests? We simply cannot take AMD for their word that the IMC is going to be fixed (if you knew the exact nature of the IMC's deficiency, you would be skeptical too), and a lot of the non-memory bound tests still put it at or around Haswell. 

 

Understand this. We cannot look at every single product to ever be released, with the doubt in the back of our minds that the information won't be concrete, or that it won't matter until X amount of time passes. We do not do this when comparing new Intel releases against their older ones, and we cannot make an exception for AMD. We speak in the present at all times, because the present is what matters for consumers. Anyone buying a product under the pretenses that it "might" get better in the future, are setting themselves up for disappointment. 

 

Until AMD (or any third party benchmark/application) releases patches for Ryzen that improves it's performance, we shall use what we currently have as the basis for our information. With that in mind, claiming "Ryzen is on par with Haswell in terms of IPC average" is simply the most accurate way to describe it's relative performance in general. As I said before, this is not a bad thing. Haswell is still a very competent architecture, and is plenty fast for consumers.

Well, I made an exception for AMD because that's just how they release they products and I got used to it, especially looking at GPUs here, I know that's a bad thing as you don't get maximum performance @ launch but that's just how it is, so I'd give 'em a month or two (tops) to do what they need to and then re-visit the IPC matter again  

[PCGuy_5960]

1 minute ago, Morgan MLGman said:

Well, I made an exception for AMD because that's just how they release they products and I got used to it, especially looking at GPUs here

GPUs have drivers though, CPUs don't and most of the time performance will not change much...

[Vode]

3 minutes ago, PCGuy_5960 said:

Take a look at benchmarks...

Averaging benchmarks doesn't seem like the one final answer.

 

1) There are several use cases missing entirely -> Audio production, several other video applications, where is Linux performance? And don't say that's an oddly specific case, so is a single game.

 

2) How are the average scores created and how are they weighted? Is 1 game weighted the same as general x264 performance?

 

It's smarter to look at what applications you're using/need to perform well, look at specific benchmarks and buy accordingly. In the cases Ryzen is marketed towards it certainly isn't Haswell IPC. In games probably and below in some cases. 

 

[Kinda Bottlenecked]

1 minute ago, PCGuy_5960 said:

The 7700K is definitely faster than X99 in games. But X99 is also faster than Ryzen in games.

Haswell-E derps because of its low clock speed. I mean, the 5960X runs at 3.0GHz which is very misleading because you can EASILY overclock it to 4.4GHz+ and get much better performance. 

 

3.5Ghz boost* 

 

That's the kind of varying stock speeds you get on the 1700 as well. One can argue that the Ryzen chips can overclock to 4.2Ghz "easily" 

 

A presumably "guaranteed" overclock of an additional 300Mhz is not worth the additional $656 USD I'd have to fork out for the 5960x + the cost of a better cooler. 

[PCGuy_5960]

11 minutes ago, Pohernori said:

3.5Ghz boost* 

On one core.... The 5960X when it is using 4 cores will boost to 3.1-3.2....

12 minutes ago, Pohernori said:

That's the kind of varying stock speeds you get on the 1700 as well.

With one big difference, Ryzen boosts ALL cores to the boost frequency, whereas Haswell-E boosts only one... So when using all cores, the 5960X will run at 3.0GHz and the 1700 will run at 3.5GHz.

14 minutes ago, Pohernori said:

One can argue that the Ryzen chips can overclock to 4.2Ghz "easily" 

4GHz*

14 minutes ago, Pohernori said:

A presumably "guaranteed" overclock of an additional 300Mhz is not worth the additional $656 USD I'd have to fork out for the 5960x + the cost of a better cooler. 

I didn't say that it is worth it, I simply said that it will perform better

[PCGuy_5960]

16 minutes ago, Vode said:

-snip-

IPC is Instructions Per Clock.... I think that you are confusing IPC with single core performance....

[Vode]

40 minutes ago, PCGuy_5960 said:

IPC is Instructions Per Clock.... I think that you are confusing IPC with single core performance....

Looking back I kind of derped..

...yes IPC (Anandtech review) of the 1700 seems to be the same as 5960x in most applications, which is line with your IPC statement.  

 

[Kinda Bottlenecked]

6 minutes ago, PCGuy_5960 said:

On one core.... The 5960X when it is using 4 cores will boost to 3.1-3.2....

 

Hmm any docs for that? 

 

I've only seen 

oc3d's 5960x review and because in his video he states that after enabling xmp. The cpu freq doesn't jump around anymore and sticks to 3.5Ghz. Hardware unboxed is also using high speed DDR4 ram as well. So presumably, xmp is enabled on their X99 system as well. 

28 minutes ago, PCGuy_5960 said:

With one big difference, Ryzen boosts ALL cores to the boost frequency, whereas Haswell-E boosts only one... So when using all cores, the 5960X will run at 3.0GHz and the 1700 will run at 3.5GHz.

 

I'm sorry that you thought that Ryzen boosts all cores to what ever the boost freq is. If (again) you'd check oc3d's video, he shows the 1800x with 3.6Ghz(3.6Ghz is base clock of the 1800x btw) all cores in a stress test. And only in Cinebench do we see the 4.0Ghz on only a single core, very much like the haswell e you've mentioned. 

 

37 minutes ago, PCGuy_5960 said:

4GHz*

 

I've seen many chips pushing voltages with 4Ghz+ 

Thought the hard limit for many is 4.2Ghz.

Most chips can hit their OC with low voltages. But when you want push further. The amount of vcore one would be pushing is alot.  

40 minutes ago, PCGuy_5960 said:

I didn't say that it is worth it, I simply said that it will perform better

 

No you said it will "destroy" Ryzen. And so far it doesn't look like that's the case. 

 

I don't hate you though, just to be clear  

[MageTank]

5 minutes ago, Pohernori said:

 

Hmm any docs for that? 

http://www.intel.com/content/www/us/en/support/processors/000006652.html

http://www.cpu-world.com/CPUs/Core_i7/Intel-Core i7-5960X Extreme Edition.html

 

Every CPU has a boost table that goes beyond base as long as thermals and power delivery are adequate. 

[Carclis]

52 minutes ago, PCGuy_5960 said:

With one big difference, Ryzen boosts ALL cores to the boost frequency, whereas Haswell-E boosts only one... So when using all cores, the 5960X will run at 3.0GHz and the 1700 will run at 3.5GHz.

I just noticed Pohernori made this same comment, but Ryzen turbo works in the same way as it does on Intel processors. The only all-core boosting that AMD does is through the XFR feature which adds 100Mhz on the 1800X.

[PCGuy_5960]

9 minutes ago, Pohernori said:

The cpu freq doesn't jump around anymore and sticks to 3.5Ghz. Hardware unboxed is also using high speed DDR4 ram as well. So presumably, xmp is enabled on their X99 system as well. 

Really? I think that they used stock settings, which means 2133MHz...

10 minutes ago, Pohernori said:

I'm sorry that you thought that Ryzen boosts all cores to what ever the boost freq is. If (again) you'd check oc3d's video, he shows the 1800x with 3.6Ghz(3.6Ghz is base clock of the 1800x btw) all cores in a stress test. And only in Cinebench do we see the 4.0Ghz on only a single core, very much like the haswell e you've mentioned. 

In Linus's launch video, the 1700 appeared to be boosting to its boost frequency across all 8 cores... Idk if that has to do with the Windows power plan though...

12 minutes ago, Pohernori said:

I've seen many chips pushing voltages with 4Ghz+ 

Thought the hard limit for many is 4.2Ghz.

Most chips can hit their OC with low voltages. But when you want push further. The amount of vcore one would be pushing is alot.  

With a voltage of 1.35V most Ryzen CPUs will go up to 4GHz...

13 minutes ago, Pohernori said:

No you said it will "destroy" Ryzen. And so far it doesn't look like that's the case. 

I know, it is not the case. He exaggerated, so I did the same

[Kinda Bottlenecked]

24 minutes ago, PCGuy_5960 said:

Really? I think that they used stock settings, which means 2133MHz...

 

Wouldn't make sense now if they specifically specified their high speed ram. And I'm even more sure now that the 5960x was running at 3.5Ghz(not all cores) since the documents thae magetank linked me states that 2 cores can boost to 3.5Ghz. Games still aren't quite heavy enough to be fully stressing 4 cores to the max. 

 

24 minutes ago, PCGuy_5960 said:

In Linus's launch video, the 1700 appeared to be boosting to its boost frequency across all 8 cores... Idk if that has to do with the Windows power plan though...

 

I'd need to request a link with a time stamp for that one. I can't find it. 

 

24 minutes ago, PCGuy_5960 said:

With a voltage of 1.35V most Ryzen CPUs will go up to 4GHz...

 

And going beyond that, users are hitting their chips with 1.4~1.5+ vcore. Though their temps can still be tamed. 

 

24 minutes ago, PCGuy_5960 said:

I know, it is not the case. He exaggerated, so I did the same

 

[MageTank]

For those of you saying Ryzen offers similar gaming performance to Broadwell-E, allow me to introduce this: 

 

 

 

Pay less attention to the current/average FPS being shown, and glue your eyes to the frametime chart. This is going to be your best metric for gauging a good gaming experience. It is here, that you understand the limitations of Ryzen when driving high refresh rates, and that even Broadwell-E does a better job of this (albeit, without showing a drastic difference in raw framerates). Even when comparing Broadwell-E to the 7700k, you will notice that it's frametime stays consistent, showing that those extra threads do in fact help with frametime/minimum framerates. 

 

So, what can we take away from this? Ryzen, while still being a very capable gaming CPU depending on your setup, is not faster at gaming than the X99 Broadwell or Z270 Kaby chips generally speaking. In extremely well-designed titles (Crysis 3 for example) we see frametime between the Broadwell and Ryzen chips to be nearly identical, and both being superior than the 7700k, but in other titles, Ryzen falls short of Broadwell-E in this aspect. I believe a balance can be found with the R5 chips, assuming they can overclock higher, as the extra threads with a little more clock speed might help alleviate frametime in titles that demand more threads, and other titles that simply demand faster cores. Only time will tell.

[PCGuy_5960]

3 hours ago, Pohernori said:

I'd need to request a link with a time stamp for that one. I can't find it. 

 

3 hours ago, Pohernori said:

And going beyond that, users are hitting their chips with 1.4~1.5+ vcore. Though their temps can still be tamed. 

AMD said that you shouldn't use more than 1.35V with Ryzen AFAIK....

3 hours ago, Pohernori said:

Because I can, LOL 

[ZackCodesAI]

@Morgan MLGman Looks like we have happy news to overclockers, the new Bios, Microcode and Stability updates are out (at least for ASUS MoBo), but yeah, it's quite smoother now.

 

[Jahramika]

On 3/17/2017 at 4:36 PM, Valentyn said:

 

I'm sorry but stating "difference was because of the way the game was made" is a Perfect reason to test more at 4K, and 1440p.

Without those tests from other reviewers we have no idea if Ryzen performs similarly in 4K for other games. 

 

If we merely based all of our conclusions on 720P and 1080P we should assume that ryzen would give worse performance at 4K, or at best similar performance since the GPU takes the brunt of the work as time goes on.
As we can see in GTA that's not the case. 

Even if the 7700K gives 10FPS more on averages, but it's minimums are 50% under that of Ryzen with more drops under 30FPS, the gaming experience is significantly worse over all.

That's why a full data set is required, with as many games in the suite as possible.
There is absolutely NO harm in testing 1440p and 4K, all it does it highlight if there is any difference in some games.

People don't want to hear all that just trying toxins reason to say Ryzen is bad.

[ZothOmmog]

On 2017-03-20 at 5:45 PM, PCGuy_5960 said:

The 7700K is definitely faster than X99 in games. But X99 is also faster than Ryzen in games.

Haswell-E derps because of its low clock speed. I mean, the 5960X runs at 3.0GHz which is very misleading because you can EASILY overclock it to 4.4GHz+ and get much better performance. 

But on that picture of the paused YouTube video of Gears of War 2, you see that the 6900K is at 136 which is 6 behind the 7700K(142) but also 6 above the 1800X(130) and now my question or my wondering is that can the naked eye actually see that small difference of 12fps between the 1800X and 7700K?

 

And people might then just forget common sense and be like "score! now I have the requirements to make an argument!", maybe?

[PCGuy_5960]

2 minutes ago, ZothOmmog said:

now my question or my wondering is that can the naked eye actually see that small difference of 12fps between the 1800X and 7700K?

1. If you have a 144Hz display, yes.
2. If you are an enthusiast, those 12 frames matter, believe me.

[ZothOmmog]

48 minutes ago, PCGuy_5960 said:

1. If you have a 144Hz display, yes.
2. If you are an enthusiast, those 12 frames matter, believe me.

1. Yes, I understand that but all in all, what would that percentage be at this moment. Like what 15%, maybe?

2. Yes, if you are an enthusiast.

2.5 But then I can raise you with, why would an enthusiast then look at a game benchmark, and not at a score from, for example Cinebench 15? Isn't that what matters most at the end of the day, what scores are produced by enthusiast grade software?

[PCGuy_5960]

2 minutes ago, ZothOmmog said:

2.5 But then I can raise you with, why would an enthusiast then look at a game benchmark, and not at a score from, for example Cinebench 15? Isn't that what matters most at the end of the day, what scores are produced by enthusiast grade software?

Depends on the enthusiast

[Morgan MLGman]

9 hours ago, PCGuy_5960 said:

1. If you have a 144Hz display, yes.
2. If you are an enthusiast, those 12 frames matter, believe me.

Well, that's provided you're playing at 1080p and your GPU is actually able to push that many frames in AAA titles. Cause at 1440p the difference 7700K - 1800X is around 5%...

8 hours ago, ZothOmmog said:

1. Yes, I understand that but all in all, what would that percentage be at this moment. Like what 15%, maybe?

2. Yes, if you are an enthusiast.

2.5 But then I can raise you with, why would an enthusiast then look at a game benchmark, and not at a score from, for example Cinebench 15? Isn't that what matters most at the end of the day, what scores are produced by enthusiast grade software?

It just depends what you're doing... If you were to only game, then yeah, 7700K will win with the 1800X in average framerates (though frametimes are usually better on the 1800X).

But then again, why would you get the 1800X if the 1700 is just as good?

(proof)

 

 

[ZothOmmog]

19 hours ago, Morgan MLGman said:

Well, that's provided you're playing at 1080p and your GPU is actually able to push that many frames in AAA titles. Cause at 1440p the difference 7700K - 1800X is around 5%...

It just depends what you're doing... If you were to only game, then yeah, 7700K will win with the 1800X in average framerates (though frametimes are usually better on the 1800X).

But then again, why would you get the 1800X if the 1700 is just as good?

(proof)

 

 

Well I was only mentioning the 1800X because of the fact that the 1700 wasn't even in the photo of the paused video clip, and I just assumed that people would have started to nag about that.