AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[lolfail9001]

Also this image from baidu topic lists 4Ghz Zen as scoring 165pts in R15 :

Looks too high compared to earlier Cinebench.
But well, it is sort of sad to look at this graph and realize that AMD sat at virtually exact same IPC since the first Phenom.

 

[cytg111]

IPC is clock independent. Both chips have Haswell cores and baring some drastic DDR3 memory subsystem differences these two would perform the same if clocked the same. It is simple.

Agreed. If clocks were the same. Were they?.. If in fact so, any deviation would be down to the subsystem.
But. If you are measuring how many instructions a 4Ghz haswell can pull through vs. a 3.5GHz "IPC" isnt gonna be same unless its running in uops or level1.

 

[inf64]

Agreed. If clocks were the same. Were they?.. If in fact so, any deviation would be down to the subsystem.
But. If you are measuring how many instructions a 4Ghz haswell can pull through vs. a 3.5GHz "IPC" isnt gonna be same unless its running in uops or level1.

You are confusing single threaded performance(IPC+clock) with IPC. IPC means instructions per clock cycle. On the same core IPC will remain the same but with clock the single threaded performance which is IPC+clock speed will change.

Looks too high compared to earlier Cinebench.
But well, it is sort of sad to look at this graph and realize that AMD sat at virtually exact same IPC since the first Phenom.

Not really. Check the image in my previous post, from the leaker himself. Seems that final/retail SKUs will score 140-150pts @ 3.4Ghz, so at 4Ghz it should get in the 164-176 range.

 

[formulav8]

There are really people expecting KL/SL IPC from AMD with this first gen core?

Honestly, in a single generational bump some of you would be upset and call AMD a failure if they didn't deliver SL/KL IPC?

AMD aimed for 40% IPC over XV and somewhat said they surpassed their goal.

Their IPC being IB/HW would really be a failure in the eyes of some of you?

Look at the expected prices. The bang for the buck is what you should be focusing on, not anything like getting the performance crown. IMO, myself, going from a 860k to an 4c/8t with X or 6c/12t nonX is better than Intel would give me for the money that I would want to spend when it comes to bang/buck. Stop overhyping peoples expectations already.

But BTW, I hope AMD pounds Intel into the sand in everything under the sun. But although I don't expect it to be, I hope it is true.

 

[Dresdenboy]

Looks too high compared to earlier Cinebench.
But well, it is sort of sad to look at this graph and realize that AMD sat at virtually exact same IPC since the first Phenom.

It doesn't reflect game performance, process improvements, or power efficiency (showing at a fixed clock, not what the chips were able to run at). Just today I dug up the PC Games Hardware review showing the effect of background tasks on games. The FX models started lower than i3 to i7, but then they kept the game's performance levels with heavy BG tasks, while the Intel CPUs went down.

Never mind.

That screen[ in OP] still makes little sense tbh.

We did some OPN guessing back then.

Now I'd read it as 2D3151.. (3.15GHz base clock).

 

[inf64]

Original thread :
http://tieba.baidu.com/p/4986046551?pn=2

Page 1 has this image:

Page 2 has R15 ST run and clock speed during the run.

Here is MT run with the score:
https://tieba.baidu.com/p/4987249977?pid=104274620087&cid=&red_tag=2485295660#104274620087

 

[lolfail9001]

We did some OPN guessing back then.

Basically the Canard's version?

Not really. Check the image in my previous post, from the leaker himself. Seems that final/retail SKUs will score 140-150pts @ 3.4Ghz, so at 4Ghz it should get in the 164-176 range.

Yeah, but look at image you have posted yourself. I, not for a second, believe the turbo was disabled in single threaded test with QS, so it is more like 140-150pts @ 3.7Ghz. At 4Ghz it is a just a little short of the mark. But close enough anyways to consider it mostly true.

 

[inf64]

Basically the Canard's version?

Yeah, but look at image you have posted yourself. I, not for a second, believe the turbo was disabled in single threaded test with QS, so it is more like 140-150pts @ 3.7Ghz. At 4Ghz it is a just a little short of the mark. But close enough anyways to consider it mostly true.

The guy who leaked it is using some Chinese version of twitter-like website where he posted it all. I cannot find it unfortunately.

PS You missed my passmark reply it seems . The application is just fine.

 

[cytg111]

You are confusing single threaded performance(IPC+clock) with IPC. IPC means instructions per clock cycle. On the same core IPC will remain the same but with clock the single threaded performance which is IPC+clock speed will change..

I understand that. How is a piece of software gonna tell the difference?

 

[inf64]

I understand that. How is a piece of software gonna tell the difference?

It detects the clock speed? It has a database full of entries for the same CPUs and has algorithms that detect deviations from the mean/standard results?

Back to Zen.
Has this been posted before?
https://www.nordichardware.se/nyheter/ryzen-presterar-battre-an-vantat.html

NordicHardware, after conversations with several Taiwanese motherboard manufacturers to confirm that the performance of AMD's new processors is almost surprisingly good. This according to internal testing of motherboard manufacturers showing that AMD can beat Intel to the fingers in several tests, not only specially selected and AMD-optimized scenario. Sources at NordicHardware points out that the results are positively surprising but at the same time that there are some storm clouds around performance per dollar.

 

[dogen1]

Okay, MOVBE sounds like the big deal here.

But none of those, except probably MOVBE could explain such a ridiculous jump in performance, could they?

Well, the 4th ALU raises the theoretical integer throughput by 33%. And powerpc already had bit manipulation instructions like the ones that haswell added, so now dolphin can execute code like that in fewer instructions.

I'm sure all of these things contribute.

Dolphin doesn't have any MOVBE instructions in it, and the only intruction belonging to BMI(1) is TZCNT.

No 256-bit instructions either.

Or not.

 

[zinfamous]

Do they really have the best or so they have the best marketing, which in turn created their reputation as the best?

When your brand is perceived to be the best, then you can ask more for it. Even, if it's perceived superior quality isn't born out in reality.

I am not saying you are wrong. I am genuinely curious.

Asus is good, but they typically have shit CSR (so I've read--personally, haven't had to deal with them, but YMMV).

I've been Gigabyte for my most recent 2-3 builds, though. Because I got a UD4 something back in 2009 as part of a Fry's bundle deal, and it was solid. Next mobo was solid, then the one after that.

Gigabyte has more coppers, or something. Anyway, I now tend to think of Gigabyte on the same tier as Asus, and also a bit cheaper, so that's where I'll by looking for Mobo. I'll look at MSI, too, as I had at least one good experience with them.

 

[unseenmorbidity]

There are really people expecting KL/SL IPC from AMD with this first gen core?

Honestly, in a single generational bump some of you would be upset and call AMD a failure if they didn't deliver SL/KL IPC?

AMD aimed for 40% IPC over XV and somewhat said they surpassed their goal.

Their IPC being IB/HW would really be a failure in the eyes of some of you?

Look at the expected prices. The bang for the buck is what you should be focusing on, not anything like getting the performance crown. IMO, myself, going from a 860k to an 4c/8t with X or 6c/12t nonX is better than Intel would give me for the money that I would want to spend when it comes to bang/buck. Stop overhyping peoples expectations already.

But BTW, I hope AMD pounds Intel into the sand in everything under the sun. But although I don't expect it to be, I hope it is true.

IB to HW would be a failure

Good thing it's hw to bw.

Asus is good, but they typically have shit CSR (so I've read--personally, haven't had to deal with them, but YMMV).

I've been Gigabyte for my most recent 2-3 builds, though. Because I got a UD4 something back in 2009 as part of a Fry's bundle deal, and it was solid. Next mobo was solid, then the one after that.

Gigabyte has more coppers, or something. Anyway, I now tend to think of Gigabyte on the same tier as Asus, and also a bit cheaper, so that's where I'll by looking for Mobo. I'll look at MSI, too, as I had at least one good experience with them.

I have had good experiences with gigabyte as well.

From my experience, Gigabyte Asus and MSI seem like the best manufactures, then probably asrock, then biostar at the very bottom.

http://www.shopblt.com/search/order_id=805920186&s_max=25&t_all=1&s_all=AMD+AM4+&search=Search

MSI prices are the same as asus it seems.

 

[zinfamous]

I have had good experiences with gigabyte as well.

From my experience, Gigabyte Asus and MSI seem like the best manufactures, then probably asrock, then biostar at the very bottom.

http://www.shopblt.com/search/order_id=805920186&s_max=25&t_all=1&s_all=AMD+AM4+&search=Search

MSI prices are the same as asus it seems.

that's how I see it.

actually, the bundle that got me my first gigabyte board was actually a biostar deal: I picked that one up, (with PhenomII 950, I think), then the next week, Fry's listed a Gigabyte bundle with the 965 BE for like...$10 more. I drove back out, returned my unopened biostar/940 bundle and picked up the 965/Gigabyte, heh. ...pretty good for about $170 total (IIRC)

 

[formulav8]

"unseenmorbidity" Couldn't quote properly sorry.

So going from current XV to IB/HW area IPC would be a failure to you?

I may be not properly understanding..... which would be fairly normal for me.

Edit again: Didn't get an actual quote sorry.

 

[GroundZero7]

that's how I see it.

actually, the bundle that got me my first gigabyte board was actually a biostar deal: I picked that one up, (with PhenomII 950, I think), then the next week, Fry's listed a Gigabyte bundle with the 965 BE for like...$10 more. I drove back out, returned my unopened biostar/940 bundle and picked up the 965/Gigabyte, heh. ...pretty good for about $170 total (IIRC)

You lucked out, that Biostar board is a POS

 

[unseenmorbidity]

Yes, if their 6 core cannot beat a 5820k from what 3 years ago? Then that's a failure!

That's where I draw the line in the sand. A passing grade, 70%, is Haswell.

 

[lobz]

Ryzen looks like cheap Haswell-E, not anything even closely resembling 7700k.

Nothing wrong with just cheap Haswell-E in my book though.

you're gonna be so surprised

 

[lobz]

Yes, if their 6 core cannot beat a 5820k from what 3 years ago? Then that's a failure!

That's where I draw the line in the sand. A passing grade, 70%, is Haswell.

making an entire lineup of mighty intel HEDT processors obsolete does not count as a failure in my books

 

[formulav8]

Sorry, IMO, if AMD can get near IB/HW level IPC, and close to around 3.2k-4.2k clocks with the supposed prices, while you can get a $70 overclocking mb, what's the failure?

 

[csbin]

 

[lolfail9001]

you're gonna be so surprised

About what? I mean, sure, there's a rumor over amd subreddit of 4.8 on water, but it is still cheap Haswell-E level. If anything, i can be smug knowing that it would be Haswell-Broadwell in more than just Blender IPC-wise since last August, the only question being the clocks.

Sorry, IMO, if AMD can get near IB/HW level IPC, and close to around 3.2k-4.2k clocks with the supposed prices, while you can get a $70 overclocking mb, what's the failure?

The failure would be in them sitting 10 years on same IPC level [around Conroe, if not under] and then ending up 3.5 years behind in IPC anyways.

 

[formulav8]

Buddy, their IPC is already at least around Conroe or more overall, what's the deal? I must have missed something....

Edit: AMD is looking pull something like a P4 to Conroe IPC improvement. Which looks to be quite true. Everyone worshipped Intel when that happened.

 

[lobz]

The 1600X XFR enabled run ACT at just 3.45Ghz? i dont buy it, i dont belive 3.7Ghz ST either.

that's no problem, there will be lost of other people buying it, many of them possibly on launch day

 

[lobz]

About what? I mean, sure, there's a rumor over amd subreddit of 4.8 on water, but it is still cheap Haswell-E level. If anything, i can be smug knowing that it would be Haswell-Broadwell in more than just Blender IPC-wise since last August, the only question being the clocks.

about real world performance