AMD Ryzen 3DMARK Physic score leaks

[WYP]

Alleged 3DMARK Physics scores have leaked for AMD's Ryzen CPUs, showcasing performance that is highly competitive with Intel's current offerings.



Read more on AMD's Leaked Ryzen 3DMARK Physics scores.

 

[Kaapstad]

Those scores look very bogus.

For example the single core scores are just the total score divided by the number of cores.

 

[Dicehunter]

IF these scores are real then Intel needs to start charging less, There will be no excuse to charge 2-3 times as much for the same performance as no sane person would buy it.

 

[Doomslayer]

IF these scores are real then Intel needs to start charging less, There will be no excuse to charge 2-3 times as much for the same performance as no sane person would buy it.

Yes and very big IF.

But Dice i'm not sane.....lol

 

[Zoot]

To be honest, the leaks the last few weeks have been producing a picture too good to be true I think.

For AMD to come from so far behind Intel to producing something extremely competitive with not only the i7-6900k but all of Intel's lineup too, and to do so at a dramatically reduced power envelope (140W down to 95W), and for them to do it at a much lower price point seems far too good to be true.

I want to believe, but I just can't given that. Although I do hope I'm wrong.

 

[AlienALX]

To be honest, the leaks the last few weeks have been producing a picture too good to be true I think.

For AMD to come from so far behind Intel to producing something extremely competitive with not only the i7-6900k but all of Intel's lineup too, and to do so at a dramatically reduced power envelope (140W down to 95W), and for them to do it at a much lower price point seems far too good to be true.

I want to believe, but I just can't given that. Although I do hope I'm wrong.

You do realise they have been working on Zen for years, right? they didn't just decide six months ago to make a new CPU tech and then bish bam release it within a year?

As for the prices? for AMD they are a lot. It's been over a decade since they could charge £500 for a CPU. Also when you say 140w down to 95w.. You do realise that Vishera and Bulldozer are a totally different core tech to Zen right? and were crap from the offset.

This is new tech, on a shrink, six years in the making or more.

Kaap - Well if they've done that for the Ryzen CPUs then they have also done exactly the same for the Intels so the whole chart is bogus in that case.

Also, whilst I'm here. AMD have themselves claimed that Ryzen is at least 40% faster than Vishera. If you consider that Vishera was basically almost as quick as Westmere (sans the HT per core) and then think about all of the 5% here and there from Intel? 40% is about right and where it needs to be.

And that is AMD claiming that, a claim which could be refuted within about ten seconds of the launch. So I really can't see them lying about it because they would only get caught out any way.

 

[AlienALX]

First of all apologies for the double post. It's important, though, and I have put some of my time into this.

OK, so Kaap says that apparently the scores are bogus because to calculate the single core score they have merely divided the overall score by the number of cores. TBH in a perfect world that would actually be a half reasonable way of calculating a CPU's power and allowing you to basically put it in a list and compare it with others. 3DMARK Firestrike is a benchmark, and was designed to scale with hardware no matter how much you add. So personally I don't think it was a terrible thing to do, and here is why.

OK, so I have a Ivybridge 8 core 16 thread Xeon. It is a chopped down version of the retail 2680 V2. Mine has 8 cores as opposed to 10 though, because it's an ES (Engineering Sample). It does have the 20mb cache, though.

OK so I decided to see if I could replicate my CPU's core score in Firestrike, by deliberately hobbling it (and thus making it act like a lower end CPU) and here is what I found.

Firstly I ran the CPU with four cores enabled and the rest disabled. I disabled any background apps etc. So here is Firestrike with 4c 8t.

OK, so our Physics score is 8320. So, using the method used here let's divide that by the core count of 4. 8320 / 4 = 2080. Note that I ran HWMON and monitored the clock speeds closely. Note that the CPU hit a maximum of 3300mhz.

OK, so for the next test I ran the CPU as a 6 core 12 thread. The reason I used 4 and 6 cores in the test is because they both boost to the same speed on my CPU. Here is how the CPU fared (physics score) with 6c 12t.

Let's ignore the overall score and focus instead on the physics score, which is 11,652. Let us then divide that number by 6 (because we have 6 cores active.) The score is 1942.

So, overall we have a four core score of 8320 that divides down to 2080 and a 6 core score of 11,652 that divides down to 1942. That's pretty damn close tbh and IMO well within a margin of error. I could also be scoring higher per core on the 4 core arrangement because it has extra cache, now that there are less cores.

Also, if we look back at the benchmark posted today.

OK so concentrate on the 5960x single core result. 2018 @ 3ghz. My CPU is very very close in layout and cores etc as the 5960x but obviously mine is Ivybridge, not Haswell. So I should be within 10% of the 5960x.

If these tests and results are bogus then somebody went to an awful lot of trouble making them break down quite well.

 

[Shingara]

Im in the 'ignore everything till ppl have them in there hands' stage, but just remember it was jim keller who designed these things, so anything is possible.

 

[Nogginthenog]

6 Core CPUs matching i7's performance for i5 money seems to be the real shining light in this release as far as I can see.

And at 65W power as well, they seem to be the perfect basis a for mainstream gaming/all purpose PC that most people could afford and use.

I must admit, if the numbers in the leaks all become reality, I'll be tempted to try a small form factor rig for 1080p gaming and media consumption built around an an R5, probably the 1500.

 

[AlienALX]

Oh BTW guys, something else you may find interesting..

OK so a lot of people seem to be confused by the TDPs we have seen given. Some saying that a CPU with a 65w TDP within 300mhz of another one that is 95w does not seem to add up.

I was poking around in my bios today (obs, I was disabling cores etc) and I noticed a setting. I can't remember exactly what it was called. Power limit? maybe?

Any way, I know for certain that my CPU is 130w. I can set that wattage as I please in the bios, but when you highlight it it says the following..

"CPU will run at normal speeds. If limit is reached the CPU will lower clocks to meet power target".

So yeah, basically my motherboard has a set limit for my CPU (auto is 130w, spot on) that I can change if I want it to boost at full speed all of the time.

So AMD could well have used a similar method with Ryzen. And I would also imagine you can disable it and pee power up the wall

 

[AngryGoldfish]

That's some mighty fine diggin'.

 

[Zoot]

You do realise they have been working on Zen for years, right? they didn't just decide six months ago to make a new CPU tech and then bish bam release it within a year?

As for the prices? for AMD they are a lot. It's been over a decade since they could charge £500 for a CPU. Also when you say 140w down to 95w.. You do realise that Vishera and Bulldozer are a totally different core tech to Zen right? and were crap from the offset.

This is new tech, on a shrink, six years in the making or more.

Yes, I'm very much aware of how new product releases work in the Semiconductor industry that products are in the works for years and years before final release. I do after all work in it, albeit on the Analogue side of things.

There is a history of recent AMD products being underwhelming. The hype-train has hyped things up far too much in the past. What I find most curious now is how silent AMD themselves have been about it. We've next to no information (officially) and we're supposed to be like a month from launch.

On the other hand, undertaking the development of new x86 core is a huge costly R&D undertaking particularly for a company the size of AMD. So I'm inclined to think that it will be competitive with Intel for that very reason simply because it could totally ruin the (already pretty bad) finances of the company if it isn't. In fact if it isn't competitive, then they really shouldn't given the R&D cost.

I'm just skeptical - I'll remain skeptical until I see benchmarks from someone like Tom.

 

[AngryGoldfish]

Is it in the GPU sector that you have been disappointed? If that's the case then maybe the reason why we might not be disappointed by AMD in the CPU sector is because Intel have been such lazy sods while Nvidia have cranked out crackin' GPU's and have continued to advance even the most basic of topologies—Pascal is just Maxwell on speed which is Kepler on speed. In other words, competition in the GPU sector is hot while it's lukewarm in the CPU sector, leaving AMD to swoop in with Zen and blow us all away. AMD have might simply have tougher competition in the GPU sector.

 

[Kaapstad]

Here is my 6950X running @4.0ghz using 1 and 10 cores in the Firestrike physics test.

It is quite a bit different to the graphs in the article.

 

[Shingara]

Well underwhellming products can only really be leveled at vishera if we are honest, and even in the case of vishera when it was released yes it wasnt no where as good as they said, but back then the enviroment wasnt for multicore, it was single core, and thats where it fell on its arse.

But im sat here with an 8350 after 5 years and i lose what, around 5 fps to what intel have. So when it comes to getting my moneys worth with this cpu then god damn. And thats purly because amd got in on the console market and the way games are made is changing more to multicore, and vishera has benefited from that.

Now on the gpu side, the 480's and stuff are really solid cards, the market that they are built for is spot on, there cheap cards that you can throw in your system and forget about, there apus are in the same boat, cheap gaming apus that you throw in and forget about.

And its not like sides are playing fair, amd needed the money for r&d whilst intel were giving money to keep amd's out of pre built, then we have nvidia playing funny sods with gameworks. But here we are, amd are still holding on and they are coming out punching.

They got jim keller back to create new tech from the ground up, now that guy is a messiah when it comes to tech, hate amd, love amd, you cannot deny that guys brains. And really amd take chances, and they swing way above the weight.

So these benchmarks true or false, we have seen what ryzen can do with the new horizens event, we have seen that asus are willing to put forward good products for this that rival z270. So ye everyone could be talking rubbish about how good these are, everyone could be holding back on just how good these things are going tobe.

We just wont know until we get them in our hands.

 

[wozza365]

To be honest, the leaks the last few weeks have been producing a picture too good to be true I think.

For AMD to come from so far behind Intel to producing something extremely competitive with not only the i7-6900k but all of Intel's lineup too, and to do so at a dramatically reduced power envelope (140W down to 95W), and for them to do it at a much lower price point seems far too good to be true.

I want to believe, but I just can't given that. Although I do hope I'm wrong.

It's not like Intel have been maxing out. They've been rather lazy, AMD have been working on this for years. They started work around the time of Bulldozer, if I'm not mistaken.

 

[AlienALX]

It is quite a bit different to the graphs in the article.

Probably because the article is running your CPU at 3ghz, not 4.

 

[Kaapstad]

Probably because the article is running your CPU at 3ghz, not 4.

Nothing to do with clockspeed.

Using all 10 cores does not give 10x the score.

 

[AlienALX]

Nothing to do with clockspeed.

Using all 10 cores does not give 10x the score.

It won't. If you are only using one core it gets all the cache etc. However if you look at my results it was very close when running either 4 or 6. Within 100 points.

 

[Kaapstad]

It won't. If you are only using one core it gets all the cache etc. However if you look at my results it was very close when running either 4 or 6. Within 100 points.

Not the point

The graphs in the article are wrong.