AMD Ryzen 2 Review Mega Thread

[Razor01]

17 minutes ago, leadeater said:

Gamers Nexus info on this is good too.

 

 

 

https://www.gamersnexus.net/hwreviews/3287-amd-r7-2700-and-2700x-review-game-streaming-cpu-benchmarks-memory?showall=1

 

The high power 2700X is a fair decent amount less power draw compared to the previous low power 1700.

That is while overclocking the 1700x, stock settings quite different.  The process and any changes uarch wise gave the 2700x more clocks (10%) but the cost is still power consumption even in single threaded workloads.  they lost 20% efficiency in single threaded and 15% efficiency in multi threaded.  This shows that the power envelope  and uarch wise and node wise, is pretty much at its limit, and even more evidence points to that when we look GN's data, the same/similar voltages we see both chip hitting up against stability issues.

[Taf the Ghost]

@Razor01

 

In the single vs multi-thread power draw test, there looks to be a 10 kJ IF upkeep cost, which gets averaged down under a full multi-threaded load. AMD released information on the energy cost for Epyc's IF, but I don't think I've seen it for Ryzen.

[Razor01]

3 minutes ago, Taf the Ghost said:

@Razor01

 

In the single vs multi-thread power draw test, there looks to be a 10 kJ IF upkeep cost, which gets averaged down under a full multi-threaded load. AMD released information on the energy cost for Epyc's IF, but I don't think I've seen it for Ryzen.

Yeah there is, but AMD hasn't been too specific about why that it. Even for Eypc they don't.  IF shouldn't be costing that much in power consumption lol.  20% is heavy for an interconnect,  Looking at 30-40 watts or so

[leadeater]

26 minutes ago, Razor01 said:

That is while overclocking the 1700x, stock settings quite different.  The process and any changes uarch wise gave the 2700x more clocks (10%) but the cost is still power consumption even in single threaded workloads.  they lost 20% efficiency in single threaded and 15% efficiency in multi threaded.  This shows that the power envelope  and uarch wise and node wise, is pretty much at its limit, and even more evidence points to that when we look GN's data, the same/similar voltages we see both chip hitting up against stability issues.

1700 OC not 1700X but it doesn't make any difference when OC'ing them, XFR and TDP limits get disabled so same diff.

 

Anyway yea GN did mention this but that's only at the top end frequency limits, Ryzen 2 is way more power efficient at the same clocks up until voltage runaway which is at the same frequency point as Ryzen 1 limits. If you stay below runaway Ryzen 2 uses much less.

 

Edit:

Quote

Lower voltage for a given frequency also means lower power consumption for the same frequency. To some extent, this is binning – but most of that large delta is from improvement of the product’s clock efficiency at the “old” high clocks of 4.0GHz. To get to 4.2GHz and beyond, granted, does take over 1.42V on our chip. It’s a significant, nearly exponential curve to increase frequency by a couple hundred megahertz. We found 4.3 to be impossible to sustain on 3 of our CPUs that we’ve tested.

 

Edit 2:

At 4Ghz it's 1.162v vs 1.425v which is a massive difference.

[cj09beira]

11 minutes ago, Jack_of_all_Trades said:

Wasn't this supposed to be called Ryzen+ , or did the marketing team just slap them over the hands and chaged the plus to a "2".

well it is but it would have less marketing appeal if they called it something like 1X50(x), and the + was applied to Zen meaning its a refined architecture and not exactly to ryzen

[leadeater]

Also just as a reminder be really careful comparing reviews for power consumption, TL;DR you basically can't.

 

Quote

Single-threaded, the 2700X operated at about 46W on the X470 board, or 37W on the X370 board. The 1700X ran at 43W on X470 and 37W on X370. Remember that this is more a factor of the BIOS and board vendor.

Quote

Wherever “X370” is mentioned, that’s the Crosshair VI. When X470 is mentioned, in gaming and production benchmarks, that’s the Gigabyte X470 Gaming 7. For overclocking, memory scaling, and game streaming tests, that’d be the ASUS Crosshair VII motherboard.

 

Unless the tester has taken manual control over the voltage settings and states them no two results will be comparable enough across reviews, only within the same one.

[Razor01]

3 minutes ago, leadeater said:

Also just as a reminder be really careful comparing reviews for power consumption, TL;DR you basically can't.

 

 

Unless the tester has taken manual control over the voltage settings and states them no two results will be comparable enough across reviews, only within the same one.

What I would like to see is a wide range of tests that look at memory clocks vs power consumption that should give us a good idea of power consumption of IF.

[Brooksie359]

6 hours ago, Misanthrope said:

I was pretty clearly talking about the 10+ core chips not the entry level HEDTs because yes: I agree those are fucking stupid, both intel and AMD 8 core HEDTs.

Yeah at that point you might as well have 2 systems than spend all that money on the 10+ cores. 

[Brooksie359]

5 hours ago, Blademaster91 said:

Are they? I dunno the wording does seem odd, then again this is the first forum ive been on where spam bots are a occasional problem, requiring noobs to wait 5-10mins for a confirmation email before posting would maybe help but l'm getting OT.

There is kind of a point though, everyone praises AMD for the smallest gain while ryzen+ is rather like going from z170 to z270 an incremental upgrade and hyping it up to get you buying another board with really no new features. When Intel gives you a 2-7% incremental bump everyone screams ''OMG GREEDY SHINTEL WHERES THE REAL UPGRADE"?!?!!!ONE

I think the reason why people were mad at Intel for the small increases in performance wasn't about just from one generation to another generation. I think it is more about how they had small improvements generation over generation. Nobody expected a big boost in performance with this release but will be expecting one with Zen 2. One thing to point out is that the changes made with this release fixed alot of issues the first ryzen had so that is a huge plus. This is really good for making sure that Zen 2 doesn't have alot of the same issues.

[JoeCoke]

Damn, does Ryzen really beat Intel in gaming if you install Meltdown and Spectre patches? Is this actually true?

[The Benjamins]

10 minutes ago, 2Buck said:

Damn, does Ryzen really beat Intel in gaming if you install Meltdown and Spectre patches? Is this actually true?

Anandtech is looking into why they are the only review putting Ryzen a head.

 

as far as I see is that Ryzen 2 puts them very close to intel but still just behind. unless you care about 144z+  @ 1080p gaming on a 1080ti then Ryzen will be just fine.

[SpaceGhostC2C]

7 hours ago, Jack_of_all_Trades said:

Wasn't this supposed to be called Ryzen+ , or did the marketing team just slap them over the hands and chaged the plus to a "2".

Zen+, Ryzen 2 (for 2000 series). Zen 2 will be Ryzen 3.

Yes, crystal clear, I know.

[Doobeedoo]

So overall as expected, refined architecture on a better node giving few extra hundred MHz with updated chipset that brings better memory support and power delivery. Very solid for this gen.

[mr moose]

Why is Jayz2c not in the list? is there something wrong with his broken PC problem solving skills that makes it not worthy?

 

 

 

 

[WMGroomAK]

5 minutes ago, mr moose said:

Why is Jayz2c not in the list? is there something wrong with his broken PC problem solving skills that makes it not worthy?

 

 

 

 

He posted a video on an issue he had that killed his AMD SSD when he was getting ready to test...  Troubleshooting the problem, it appeared to be a bad SATA power cable.

 

 

[mr moose]

Just now, WMGroomAK said:

He posted a video on an issue he had that killed his AMD SSD when he was getting ready to test...  Troubleshooting the problem, it appeared to be a bad SATA power cable.

 

 

I know, hence my comment:

 

Quote

is there something wrong with his broken PC problem solving skills that makes it not worthy?

 

 

[Dylanc1500]

34 minutes ago, mr moose said:

Why is Jayz2c not in the list? is there something wrong with his broken PC problem solving skills that makes it not worthy?

 

 

 

 

Yes, not enough PC building simulator.

[SpaceGhostC2C]

@ImNotDeViLzzz you may also like this one:
 

Spoiler

 

 

[ravenshrike]

3 hours ago, Dylanc1500 said:

 

Yes, not enough PC building simulator.

Is bad SSD power cable causing a protection fault even a possible bug in PC building simulator?

[ravenshrike]

21 minutes ago, SpaceGhostC2C said:

@ImNotDeViLzzz you may also like this one:
 

  Hide contents

 

 

Arrrgh. He hits one of my pet peeves in that video. You don't mix positive and negative percentages, especially not when comparing relative performance of things. It's not consistent. 179/204 gets you 12.26% slower. However 204/179 gets you 13.97% faster. 

[SpaceGhostC2C]

30 minutes ago, ravenshrike said:

Is bad SSD power cable causing a protection fault even a possible bug in PC building simulator?

Maybe that was the problem

 

7 minutes ago, ravenshrike said:

Arrrgh. He hits one of my pet peeves in that video. You don't mix positive and negative percentages, especially not when comparing relative performance of things. It's not consistent. 179/204 gets you 12.26% slower. However 204/179 gets you 13.97% faster. 

I don't understand what you mean. In order to be consistent, you need to keep the same base. That is, if processor A is the denominator in one measure, it has to be the denominator in all percentages. Therefore, if you want to express AMD performance as a fraction of Intel performance, it is correct and consistent to say for example that AMD performed X% better in multithreaded and Y% slower in single threaded Cinebench. It is equivalent to say that AMD gives 128% of Intel's MT performance and 83% of Intel's ST performance. You are just subtracting 1 to express it in differences, whether positive (+28%) or negative (-17%).

Your example is correct, but it has nothing to do with showing negative and positive percentages on the same chart; instead, it has to do with changing the reference unit (i.e., the denominator). Funny enough, he did made that mistake once, when saying that "the 2700X has 25% more cores than Intel", since technically by saying "more cores than Intel" he's using Intel as the reference, and therefore it should have been "33% more cores". But the charts themselves are done as they should.

[ravenshrike]

32 minutes ago, SpaceGhostC2C said:

Maybe that was the problem

 

I don't understand what you mean. In order to be consistent, you need to keep the same base. That is, if processor A is the denominator in one measure, it has to be the denominator in all percentages. Therefore, if you want to express AMD performance as a fraction of Intel performance, it is correct and consistent to say for example that AMD performed X% better in multithreaded and Y% slower in single threaded Cinebench. It is equivalent to say that AMD gives 128% of Intel's MT performance and 83% of Intel's ST performance. You are just subtracting 1 to express it in differences, whether positive (+28%) or negative (-17%).

Your example is correct, but it has nothing to do with showing negative and positive percentages on the same chart; instead, it has to do with changing the reference unit (i.e., the denominator). Funny enough, he did made that mistake once, when saying that "the 2700X has 25% more cores than Intel", since technically by saying "more cores than Intel" he's using Intel as the reference, and therefore it should have been "33% more cores". But the charts themselves are done as they should.

Except the vast majority of people don't think like that. Anyone good at doing math in their head will probably subconsciously go 'wait a tic'. Others who regularly work with percentages can because they're used to it. But the average joe reading or watching something will assume that AMD being 12% slower means that Intel is 12% faster. This is a problem because the difference between the two percentages magnifies as the numbers themselves get farther apart. So if the numbers had been 150 and 200, then AMD would have been 25 percent slower but Intel would have been 33 percent faster. 

[SpaceGhostC2C]

2 minutes ago, ravenshrike said:

Except the vast majority of people don't think like that. Anyone good at doing math in their head will probably subconsciously go 'wait a tic'. Others who regularly work with percentages can because they're used to it. But the average joe reading or watching something will assume that AMD being 12% slower means that Intel is 12% faster. This is a problem because the difference between the two percentages magnifies as the numbers themselves get farther apart. So if the numbers had been 150 and 200, then AMD would have been 25 percent slower but Intel would have been 33 percent faster. 

I still don't understand your point. There is no better way to go about it, and it doesn't have anything to do with positive and negative. If both results were better for Intel, you would still have the choice of saying 25% slower or 33% faster. On the contrary, switching the reference point depending on which one is higher would be a misleading mess. Precisely what he does avoids the 25 vs 33 problem. It's all in the same units.

[pas008]

43 minutes ago, SpaceGhostC2C said:

I still don't understand your point. There is no better way to go about it, and it doesn't have anything to do with positive and negative. If both results were better for Intel, you would still have the choice of saying 25% slower or 33% faster. On the contrary, switching the reference point depending on which one is higher would be a misleading mess. Precisely what he does avoids the 25 vs 33 problem. It's all in the same units.

It's a point on baseline and view  point of reviewers

But imho is we are going into silicone limits war

If amd keeps this up Intel will be forced to use emib and use every inch of their silicone like a fake breasted stripper like amd is already is doing