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

[vissarix]

Ryzen has better IPC then kaby lake

https://youtu.be/f-Kw4XFXB30

Better Ipc then Ivy Bridge...and not even by much!!

 

[inf64]

Better Ipc then Ivy Bridge...and not even by much!!

Do you mean in CPUz benchmark? If you meant CPUz just for reference my Haswell @ 4.4ghz scores 2000 pts in ST test while Zen @ 3.7Ghz scored 1888. At the same clock Zen would score 2245 or some 10% better. But as we discussed before CPUz is not a very good benchmark.

 

[vissarix]

Do you mean in CPUz benchmark? If you meant CPUz just for reference my Haswell @ 4.4ghz scores 2000 pts in ST test while Zen @ 3.7Ghz scored 1888. At the same clock Zen would score 2245 or some 10% better. But as we discussed before CPUz is not a very good benchmark.

I mean Cinebench R15..

 

[inf64]

I mean Cinebench R15..

As for R15 we have ST scores on the level of Haswell (ST IPC within 2% error margin) and throughput on the level of BDW-E. How is that barely faster than Ivy Bridge?

 

[hotstocks]

Everyone in here is guessing and worrying about whether Zen is Sandy, Ivy, Broadwell, or whatever IPC. NONE of that matters, all those are within 10% IPC of each other. The ONLY thing that matters for people looking for an upgrade if how fast each core is going to run stable 24/7 with air. I usually upgrade my cpu ever 2 years for the last 25 years, but I still can't believe to this day I am using a 6 year old processor. I have upgraded my ram, SSD, and video card to the latest and greatest of course. And I would LOVE to buy Ryzen 1800X. But the bottom line is this 6 year old obsolete shitty cpu that I have, well it is a sandybridge running at 4.7Ghz rock stable 24/7 for 6 years. And after seeing all your guys posting your passmarks and RYzen guess passmarks, well none of that shit impresses me. My passmark single threaded is over 2500, and you guys are getting all excited at 1800-2100? All I have to say is this thing better overclock well and get over 2500 single thread or you will see that not that many people are going to pay to do a lower or sideways upgrade unless they are doing video editing or running a server and need more cores. For home use and games it doesn't matter, there are very few games that will use more than 4 cores, and even those are faster if the 4 cores are significantly faster than the 8 cores in Ghz, period. 246 pages of arguing and guessing. Well guess what, we will know in a week when real reviews come out with real retail products, real games, and real non-canned benchmarks. I am no Intel fanboi, If Ryzen pulls through and really beats Intel I will be the first to buy one, but it really has to give me a reason to upgrade my 6 year old killer cpu.

 

[Lodix]

Isn't cinebench more optimzed for intel ? Not a good indicator of IPC in that case.

 

[rodserling]

if any of this true, shouldnt amd's stock be skyrocketing? should i be investing haha?

 

[richierich1212]

if any of this true, shouldnt amd's stock be skyrocketing? should i be investing haha?

Where have you been for the past year? AMD stock has gained something like 600%. Sure it's still under $14 as they have yet to fully recover but hey, they are almost out of that hole...

 

[richierich1212]

My passmark single threaded is over 2500, and you guys are getting all excited at 1800-2100?
I am no Intel fanboi, If Ryzen pulls through and really beats Intel I will be the first to buy one, but it really has to give me a reason to upgrade my 6 year old killer cpu.

The CPU market has become a snooze fest in the past several years since AMD released bulldozer. That is why enthusiasts like us are excited. Having more cores at a more affordable price will help push the market towards software and games using more than what we have today (quad mainstream, when just a couple of years ago dual cores were acceptable.)

 

[rodserling]

Well yes, but the last month has been slowed,

 

[lobz]

I mean Cinebench R15..

deciding IPC based solely on cinebench... that'll take you far

 

[The Stilt]

Isn't cinebench more optimzed for intel ? Not a good indicator of IPC in that case.

No.
It has been compiled with Intel Compiler, which according to some automatically makes it to favor Intel. Intel compiler automatically adds a dispatcher to each and every binary or library it produces, however Cinebench has not been compiled with settings which would generate a different code path for Intel CPUs. Because of that, despite the dispatcher is present in the binary removing it makes no difference. It can be tested by either modifying few instructions found in the binary, or running it in VM and spoofing the CPU information (vendor, model, etc).

Regardless, based on my own tests the microarchitectures which receive the largest gain in FP code from using Intel compiler (compared to MSVC or GCC) are AMD microarchitectures. Their gains are generally significantly larger than the gains for Intel's recent µarchs.

When comparing the last four (PD, SR, XV and Zen) µarchs against Haswell or newer, Cinebenches are highly favorable workloads for AMD (especially for Zen). All of the current versions are legacy scalar workloads (up to SSE3 in R15) so newer Intels can never utilize their larger resources. Despite that Cinebench represents pretty well the CPUs capabilities in the current consumer workloads.

Cinema 4D R15 on which Cinebench R15 is based on supports Embree, however Cinebench R15 uses Maxon's own rendered.
Cinebench R18 should be out shortly, so it will be interesting to see if they continue using their own renderer or swap to Embree.

 

[Lodix]

×.×

Nice thanks.

 

[teejee]

Well yes, but the last month has been slowed,

yes, because to get big gains you need to buy early when there is still a big risk. So many investors bought AMD last year assuming AMD would succeed with Ryzen.
If you wait until proof of success you will always miss big stock gains.

 

[r.p]

No.
It has been compiled with Intel Compiler, which according to some automatically makes it to favor Intel. Intel compiler automatically adds a dispatcher to each and

... but AMD (Piledriver) isn't doing particularly well in CB. (and Blender^^)
e.g. with Mentral Ray and Vray a 8350 is faster that a 3770k - were as in CB a 8350 is even slower than a 2600k ...
but this may chance with Ryzen ....

http://www.hardware.fr/articles/946-5/performances-applicatives.html
http://www.anandtech.com/show/10968...w-review-the-new-stock-performance-champion/4

 

[AtenRa]

No.
It has been compiled with Intel Compiler, which according to some automatically makes it to favor Intel. Intel compiler automatically adds a dispatcher to each and every binary or library it produces, however Cinebench has not been compiled with settings which would generate a different code path for Intel CPUs. Because of that, despite the dispatcher is present in the binary removing it makes no difference. It can be tested by either modifying few instructions found in the binary, or running it in VM and spoofing the CPU information (vendor, model, etc).

Regardless, based on my own tests the microarchitectures which receive the largest gain in FP code from using Intel compiler (compared to MSVC or GCC) are AMD microarchitectures. Their gains are generally significantly larger than the gains for Intel's recent µarchs.

When comparing the last four (PD, SR, XV and Zen) µarchs against Haswell or newer, Cinebenches are highly favorable workloads for AMD (especially for Zen). All of the current versions are legacy scalar workloads (up to SSE3 in R15) so newer Intels can never utilize their larger resources. Despite that Cinebench represents pretty well the CPUs capabilities in the current consumer workloads.

Cinema 4D R15 on which Cinebench R15 is based on supports Embree, however Cinebench R15 uses Maxon's own rendered.
Cinebench R18 should be out shortly, so it will be interesting to see if they continue using their own renderer or swap to Embree.

 

[JDG1980]

yes, because to get big gains you need to buy early when there is still a big risk. So many investors bought AMD last year assuming AMD would succeed with Ryzen.
If you wait until proof of success you will always miss big stock gains.

I don't think that the current run-up is primarily due to Zen hype. Most stock buyers don't hang around enthusiast forums like this one reading leaks. Yes, there have been some mainstream articles (even in investor publications) saying Zen is probably going to be good, but the average stock investor can't distinguish between that and the 10,000 other (largely baseless) hype articles out there. How many times do you hear company X tout that its next big product is going to be the best yet?

The current price increase is largely due, IMO, to better-than-expected earnings. The last spike happened right on Jan. 31, which was right when they announced the earnings for Q4 2016.

I haven't had a stock trading account in a long time but I just opened one up yesterday so I can buy 100 shares of AMD on Monday. I expect to see a measurable jump when the actual Zen reviews hit, and probably an even bigger boost in mid-to-late April when the Q1 2017 earnings hit. If I'm wrong, not the end of the world.

 

[swilli89]

I don't think that the current run-up is primarily due to Zen hype. Most stock buyers don't hang around enthusiast forums like this one reading leaks. Yes, there have been some mainstream articles (even in investor publications) saying Zen is probably going to be good, but the average stock investor can't distinguish between that and the 10,000 other (largely baseless) hype articles out there. How many times do you hear company X tout that its next big product is going to be the best yet?

The current price increase is largely due, IMO, to better-than-expected earnings. The last spike happened right on Jan. 31, which was right when they announced the earnings for Q4 2016.

I haven't had a stock trading account in a long time but I just opened one up yesterday so I can buy 100 shares of AMD on Monday. I expect to see a measurable jump when the actual Zen reviews hit, and probably an even bigger boost in mid-to-late April when the Q1 2017 earnings hit. If I'm wrong, not the end of the world.

I've been an AMD investor since about a year ago and the run-up is due to several things. Its always very hard to say why a stock performed the way it did (poorly or well) but yes you are correct in saying the slightly better than predicted earnings is the reason for the last little spike. In 2016 AMD announced several deals that renewed investor confidence.. the deal with the Chinese to license IP, the deal with Google for supplying chips for Deep AI learning and things like that. There is definitely already some anticipation of Ryzen/Vega built into the stock price, however. I think AMD going until at least the end of 2018 is an extremely attractive position as they will only be increasing revenue and gross profit margin from here on out.

And we as enthusiasts now have a fairly clear picture of how Ryzen may do in the marketplace based on recent performance leaks. We do not know exactly how it will perform technically, but its safe to say it will sell much better than FX in recent years which is all it will take for AMD to have some great earnings calls moving forward. Now imagine if Vega really is something special and they continue to make headway in GPGPU.. they are firing on all cylinders and confidence in Lisa Su specifically is at an all time high. There were even stories of her personally in the labs helping validate Zen samples! That's incredible! Their dual PHD MIT CEO rolling up her sleeves and personally lending her engineering skills to the company's lifeblood future product! Whatever your opinion on the company, pretty much everyone on Wallstreet is starting to really buy in to her long term vision of AMD, a return to a focused competitive spirit in high performance CPU and GPU products...

 

[RiverRicer]

AMD has beat expectations for the past few quarters and this has helped lift the stock. That, coupled with Ryzen/Vega expectations, has put it into the $13-14 range. I bought some in Feb 2016 primarily because of Zen speculation and purchased some additional this January. You're correct in saying that most stock buyers don't hang around enthusiast forums, however, in the case of AMD, you'd be surprised how many geeks are on the financial forums posting the same benchmark articles as here.
Q1 ER will most definitely show a pop, but I 'm much more interested in the boost coming maybe Q4 from Naples and Vega, both in the cloud and deep learning applications. Good luck with your shares (you're not wrong).

 

[mikk]

No.
It has been compiled with Intel Compiler, which according to some automatically makes it to favor Intel. Intel compiler automatically adds a dispatcher to each and every binary or library it produces, however Cinebench has not been compiled with settings which would generate a different code path for Intel CPUs. Because of that, despite the dispatcher is present in the binary removing it makes no difference. It can be tested by either modifying few instructions found in the binary, or running it in VM and spoofing the CPU information (vendor, model, etc).

Regardless, based on my own tests the microarchitectures which receive the largest gain in FP code from using Intel compiler (compared to MSVC or GCC) are AMD microarchitectures. Their gains are generally significantly larger than the gains for Intel's recent µarchs.

When comparing the last four (PD, SR, XV and Zen) µarchs against Haswell or newer, Cinebenches are highly favorable workloads for AMD (especially for Zen). All of the current versions are legacy scalar workloads (up to SSE3 in R15) so newer Intels can never utilize their larger resources. Despite that Cinebench represents pretty well the CPUs capabilities in the current consumer workloads.

You could repeat this over and over again but certain AMD fanboys won't accept that. Of course you are right and in the past Cinebench has proved that the results were close to results over a bunch of real world applications, so if Skylake is really 10% above Zen in Cinebench Singlethread at the same clock it wouldn't surprise me if this result is close to reality averaged over a mix of applications.

 

[vissarix]

As for R15 we have ST scores on the level of Haswell (ST IPC within 2% error margin) and throughput on the level of BDW-E. How is that barely faster than Ivy Bridge?

Im talking about IPC, and you can only see it on ST test like the cinebench one..

also it is not within the 2% margin of error, cinebench is extremely precise..

just did 3 runs in a row with my OC i7 2600k at 4.8ghz and got 173 first time and then two times 172...thats less then 1% margin of error...

Honestly i dont even see a point on upgrading my Cpu to Ryzen, im not into video editing and games who uses more then 4 cores are quite low....i play mainly CSGO and BF3 which works better with 4 cores with hyperthreading disabled GTA V needs more cores but then im bottlenecked by my GPU..

 

[lolfail9001]

just did 3 runs in a row with my OC i7 2600k at 4.8ghz and got 173 first time and then two times 172...thats less then 1% margin of error...

Uhm, from what i see 4.8Ghz 3930k does only 163. How come?

 

[inf64]

Im talking about IPC, and you can only see it on ST test like the cinebench one..

also it is not within the 2% margin of error, cinebench is extremely precise..

just did 3 runs in a row with my OC i7 2600k at 4.8ghz and got 173 first time and then two times 172...thats less then 1% margin of error...

Honestly i dont even see a point on upgrading my Cpu to Ryzen, im not into video editing and games who uses more then 4 cores are quite low....i play mainly CSGO and BF3 which works better with 4 cores with hyperthreading disabled GTA V needs more cores but then im bottlenecked by my GPU..

I have a Haswell 4690K @ 4.4Ghz and my ST results vary between 164 and 166 on Win10. AT bench shows 3.9Ghz 4690K scoring 154 pts so 4.4Ghz should be getting ~173 while I'm getting 166 at best (with super fast DDR3 2400Mhz memory at that!). That is 4% variance between online benchmark and what I can verify myself. Zen supposedly scores 146 @ 3.7Ghz making it 173pts @ 4.4Ghz, basically equal to Haswell.

 

[hotstocks]

This guy gets it. Like him I will only upgrade to Ryzen if it overclocks on air to 4.5Ghz or above, if not there is absolutely no point for mainstream users, power users, or gamers who already have a quad 4.7GHz chip to sidegrade. I get it if you are rendering all day or a server, but Ryzen will only get the bulk of computer builders (gamers) if it's overclockable way beyond stock. And from the look of those guys pictured with the shitty air heatsink/fan, I bet it will.

 

[Crumpet]

For someone like me on a non-k skylake I consider it worth upgrading..

My all core turbo is what, 3.5/3.6ghz.. Ryzen 8c comes at what 3.6 stock? Either way that'll make up for the lacking single thread and quadruple my threads...