AMD Ryzen HAS BEEN ANNOUNCED!!!

[MageTank]

4 minutes ago, Efilnikufesin said:

 

Well, to be honest. Nothing on any official basis that anyone can take with a grain of salt has been tested at all. We will still see.  Too much talking in absolutes without any verifiable data. BIOS updates, driver updates, all these things will happen. There will be bugs, are there ever not?  Why are so many people talking in such absolutes when we really haven't seen a damn thing worth mentioning yet?

 

I agree with this, which is why I've refrained from speaking in absolutes about things we've yet to see. My IPC estimates are simply that: estimates. I have always assumed Haswell IPC, and Haswell is within 5-8% of Skylake, so my "within 5-10%" number came from my original guess. As for the LN2 overclocking results, we've seen them, they are real, but even they are not indicative of Ryzen's real overclocking potential. It's just a "pseudo-ceiling" for that specific chip that may or may not even be a retail chip, on an unrefined BIOS. 

[Efilnikufesin]

30 minutes ago, MageTank said:

I agree with this, which is why I've refrained from speaking in absolutes about things we've yet to see. My IPC estimates are simply that: estimates. I have always assumed Haswell IPC, and Haswell is within 5-8% of Skylake, so my "within 5-10%" number came from my original guess. As for the LN2 overclocking results, we've seen them, they are real, but even they are not indicative of Ryzen's real overclocking potential. It's just a "pseudo-ceiling" for that specific chip that may or may not even be a retail chip, on an unrefined BIOS. 

 

Ceiling for that chip with initial BIOS revisions and drivers, but mostly I agree. Guess I will find out in about a week. Do you think I will get that nifty wood box? I feel like the 6-4 cores are going to be the gaming sweetspot, but I doubt with a little bit of OC and tweaking the 1700 I have coming is going to disappoint. Just waiting on a good cooling solution that isn't a fully custom water or an obnoxiously ugly Noctua fan air cooling.

[LAwLz]

4 hours ago, Colonel_Gerdauf said:

So, what these results are telling me is to use either tweaked x264 or QuickSync, depending on the computer configuration (overclocking helps with this) and the nature of the game you are playing. The QuickSync capture, from the images, looked pitiful, but as you said you are using an early revision of a QuickSync-supported CPU. You have also proven that the use of NVENC comes at a great loss, as you are making requests from an already-busy GPU. In terms on running them as video streams, however, the differences become negligible. I might do some benchmarks myself when I have the time.

 

I have tested 3500kbps with OBS myself, and I agree with you that it is an absolutely awful thing to see. If anybody out there is considering recording 1080P gaming footage for later use, 16mbps is the way to go.

Really? I personally thought NVENC looked the best.

Please note that NVENC does not use up any GPU core time. NVENC is its own separate chip that is on the GPU.

 

Comparison 1

Comparison 2

Comparison 3

Comparison 4

Comparison 5

 

Mouse outside = NVENC

Mouse over the picture = QuickSync

[MageTank]

9 hours ago, Efilnikufesin said:

 

Ceiling for that chip with initial BIOS revisions and drivers, but mostly I agree. Guess I will find out in about a week. Do you think I will get that nifty wood box? I feel like the 6-4 cores are going to be the gaming sweetspot, but I doubt with a little bit of OC and tweaking the 1700 I have coming is going to disappoint. Just waiting on a good cooling solution that isn't a fully custom water or an obnoxiously ugly Noctua fan air cooling.

I think the wooden box is for press release kits, but I could be wrong. They included motherboards, ram, and a CPU cooler so the press had everything they needed to review the platform as a whole. 

[Ryujin2003]

12 hours ago, MageTank said:

I agree with this, which is why I've refrained from speaking in absolutes about things we've yet to see. My IPC estimates are simply that: estimates. I have always assumed Haswell IPC, and Haswell is within 5-8% of Skylake, so my "within 5-10%" number came from my original guess. As for the LN2 overclocking results, we've seen them, they are real, but even they are not indicative of Ryzen's real overclocking potential. It's just a "pseudo-ceiling" for that specific chip that may or may not even be a retail chip, on an unrefined BIOS. 

Is a CPU's IPC like super top secret? Having that available would make calculations nicer.

 

Also, I as stated earlier AMD definitely appears to have an upper hand in the performance per price.

 

And I do agree that we have no idea what yhe overclock potential really is, but I don't know if AMD would immediately announce it before launch if they were able to smack the pants off Intel. Why give Intel an opportunity to attempt to cut into sales? If these thins are awesome, Intel will definitely have a delayed response that could really hurt. Also, AMD is leaving a lot to be desired, which is nice. They have the bar set high, but not too high, so I don't expect there to be any issues or failures of the past.

 

And as far as OC and heat, I would assume you could always deactivate a couple cores on the R7 to help get higher OC. Drop to a x6 core for example.

[dexT]

K grain of salt time but http://wccftech.com/amd-ryzen-7-1800x-overclocking-performance/

 

I guess it's beating the 6950X after a one click overclock and has better single-threaded performance per clock than Kaby. And from the comments:

 

 

[LAwLz]

2 hours ago, dexT said:

K grain of salt time but http://wccftech.com/amd-ryzen-7-1800x-overclocking-performance/

 

I guess it's beating the 6950X after a one click overclock and has better single-threaded performance per clock than Kaby. And from the comments:

An overclocked CPU for 500 dollars beats a 340 dollar CPU at stock settings? Wow, who would have thought? What happens when you overclock the 7700K? Does the 1800X still win in single core performance?

[dexT]

25 minutes ago, LAwLz said:

An overclocked CPU for 500 dollars beats a 340 dollar CPU at stock settings? Wow, who would have thought? What happens when you overclock the 7700K? Does the 1800X still win in single core performance?

It says it has better IPC than Kaby not that's it's faster in single-threaded workloads than the 7700K, can you read? It says it beat the 6950X.

 

I couldn't give a damn about a quad-core CPU.

[MageTank]

1 hour ago, dexT said:

It says it has better IPC than Kaby not that's it's faster in single-threaded workloads than the 7700K, can you read? It says it beat the 6950X.

 

I couldn't give a damn about a quad-core CPU.

It's WCCF, they are over-exaggerating Ryzen's capabilities. AMD themselves won't come out and say they hit Skylake/Kaby IPC, because it's extremely unlikely. Every single estimate put them close to Haswell. From what we saw at their event on the LTT videos, they might have achieved Broadwell IPC (problem is, limited array of benchmarks for us to be certain) but there is no certainty that they achieved the additional 3%(+/-) required to match Skylake/Kaby. That, and you can't exactly boast that an overclocked CPU is beating two other CPU's at their stock clocks, because it means absolutely nothing. The moment you overclock those other chips, even slightly, the advantage that the Ryzen CPU had immediately falls away.

 

Now, I am not saying it's not impressive, because for the price (and if this rumor is to be believed) you are getting a lot of multi-threaded power and apparently a decent amount of single threaded power, for a fraction of the price that Intel is asking for. That's indeed a great thing, but I don't want people to buy into this hype and be let down because they ignored all rational thought for the sake of a rumor. 

 

I keep seeing IPC get thrown around as if it's the be-all, end-all but it seems many people forget to factor the rest of the equation in. Hopefully the vast majority are conscious of that. 

[cj09beira]

at least on ln2 the max people have pushed an 5960x on all cores for cinebench r15 was 5.5ghz although for the 6900k the max reached on hwbot for cinebench was 5.1ghz, 

so i expect ryzen to overclock to around the same levels as broadwell-E, 

it being an 8 core its normal 

i do believe that for the future ryzen will be good, unless you want super high fps (144hz and up) because then cpu frequency helps a lot, for the rest it will be just fine, and be able to stretch its legs in future games.

 

[dexT]

40 minutes ago, MageTank said:

It's WCCF, they are over-exaggerating Ryzen's capabilities. AMD themselves won't come out and say they hit Skylake/Kaby IPC, because it's extremely unlikely. Every single estimate put them close to Haswell. From what we saw at their event on the LTT videos, they might have achieved Broadwell IPC (problem is, limited array of benchmarks for us to be certain) but there is no certainty that they achieved the additional 3%(+/-) required to match Skylake/Kaby. That, and you can't exactly boast that an overclocked CPU is beating two other CPU's at their stock clocks, because it means absolutely nothing. The moment you overclock those other chips, even slightly, the advantage that the Ryzen CPU had immediately falls away.

 

Now, I am not saying it's not impressive, because for the price (and if this rumor is to be believed) you are getting a lot of multi-threaded power and apparently a decent amount of single threaded power, for a fraction of the price that Intel is asking for. That's indeed a great thing, but I don't want people to buy into this hype and be let down because they ignored all rational thought for the sake of a rumor. 

 

I keep seeing IPC get thrown around as if it's the be-all, end-all but it seems many people forget to factor the rest of the equation in. Hopefully the vast majority are conscious of that. 

Yeah I said a grain of salt. All of the other leaks including the ones from them appear to be true though.

[DocSwag]

7 hours ago, Ryujin2003 said:

Is a CPU's IPC like super top secret? Having that available would make calculations nicer.

 

Also, I as stated earlier AMD definitely appears to have an upper hand in the performance per price.

 

And I do agree that we have no idea what yhe overclock potential really is, but I don't know if AMD would immediately announce it before launch if they were able to smack the pants off Intel. Why give Intel an opportunity to attempt to cut into sales? If these thins are awesome, Intel will definitely have a delayed response that could really hurt. Also, AMD is leaving a lot to be desired, which is nice. They have the bar set high, but not too high, so I don't expect there to be any issues or failures of the past.

 

And as far as OC and heat, I would assume you could always deactivate a couple cores on the R7 to help get higher OC. Drop to a x6 core for example.

No, because it's approximation can be calculated quite easily through benchmarks.

 

However, you can't just slap a number on a CPU architecture and say, "This is the IPC of -CPU architecture-." CPUs are very complicated things with many components. Take Ryzen, for example. The IPC of its integer performance is very close to Broadwell, but the floating point performance lags behind because it's FPUs aren't as large. As well, there's the 5775C which at stock can sometimes pull ahead of a higher clocked 6700k due to its L4 cache.

 

And that's why we need benchmarks. To sure in so and so scenario, what CPU does well. IPC is not a static number and never will be. And anyone that thinks it is is flat out wrong.

[MyName13]

Well that's interesting, it seems AMD said that all zen CPUs will be available on day 1.

[Dylanc1500]

On 2/23/2017 at 10:53 AM, JDMFTW said:

Sucks as the 6700k right before Christmas was 249.99 

Ya i know i am still kicking myself for not grabbing one since they decided to jack them back up to msrp right after new year.

[3DOSH]

 

http://ranker.sisoftware.net/show_run.php?q=c2ffcdfcdabbdae7d4edd5e4d6f082bf8fa9cca994a482f1ccfc&l=en

 

AMD Ryzen cache performance looks impressive to say the least:

 

L1 Cache Bandwidth 776.86GB/s

L2 Cache Bandwidth 590.10GB/s

L3 Cache Bandwidth 408.71GB/s

 

Thats with 16GB Corsair CMK16GX4M2B3000C15 DIMM PC24000 a 3000 mhz RAM.
 

[DocSwag]

7 hours ago, MyName13 said:

Well that's interesting, it seems AMD said that all zen CPUs will be available on day 1.

It only said, "When Zen debuts it’ll debut in multiple (still unknown) configurations." That's basically what we got. The multiple configurations were 3 8 core CPUs.

4 hours ago, 3DOSH said:

 

http://ranker.sisoftware.net/show_run.php?q=c2ffcdfcdabbdae7d4edd5e4d6f082bf8fa9cca994a482f1ccfc&l=en

 

AMD Ryzen cache performance looks impressive to say the least:

 

L1 Cache Bandwidth 776.86GB/s

L2 Cache Bandwidth 590.10GB/s

L3 Cache Bandwidth 408.71GB/s

 

Thats with 16GB Corsair CMK16GX4M2B3000C15 DIMM PC24000 a 3000 mhz RAM.
 

 

4 hours ago, 3DOSH said:

 

http://ranker.sisoftware.net/show_run.php?q=c2ffcdfcdabbdae7d4edd5e4d6f082bf8fa9cca994a482f1ccfc&l=en

 

AMD Ryzen cache performance looks impressive to say the least:

 

L1 Cache Bandwidth 776.86GB/s

L2 Cache Bandwidth 590.10GB/s

L3 Cache Bandwidth 408.71GB/s

 

Thats with 16GB Corsair CMK16GX4M2B3000C15 DIMM PC24000 a 3000 mhz RAM.
 

No not really. If you compare it to this, which is a 6900k:

http://ranker.sisoftware.net/show_run.php?q=c2ffcdfcdabbdae7d4e2daecd8fe8cb181a7c2a79aaa8cffc2f3c7&l=en

 

Ryzen SIGNIFCANTLY lags behind in l1 cache bandwidth. I'm not sure why; this seems a bit strange. In l2 cache bandwidth it's about the same, and in l3 cache bandwidth it seems to pull ahead quite significantly too. Though if I were you I'd wait for official benchmarks to see.

[leadeater]

24 minutes ago, DocSwag said:

Ryzen SIGNIFCANTLY lags behind in l1 cache bandwidth. I'm not sure why; this seems a bit strange. In l2 cache bandwidth it's about the same, and in l3 cache bandwidth it seems to pull ahead quite significantly too. Though if I were you I'd wait for official benchmarks to see.

Basically that's because currently Intel is the best fab for SRAM by a significant amount so when it comes to L1 cache Intel will be much faster.

[DocSwag]

4 minutes ago, leadeater said:

Basically that's because currently Intel is the best fab for SRAM by a significant amount so when it comes to L1 cache Intel will be much faster.

Isn't L2 cache also SRAM though? So why isn't AMD behind by a lot there too? Is it because AMD has double the L2 cache per core?

[leadeater]

1 minute ago, DocSwag said:

Isn't L2 cache also SRAM though? So why isn't AMD behind by a lot there too? Is it because AMD has double the L2 cache per core?

Not sure, L2 cache is supposed to be significantly cheaper and cache logic between levels likely plays a big part. Not expert enough to say without giving misinformation, I'd have to read up on it a bit before going in to that kind of discussion. Last time I bothered reading in to CPU cache was way back in 2000's comparing the then AMD Opterons to Intel Xeons and how AMD were absolutely crushing Intel in the L1 and L2 department and above dual socket efficiency, how things have changed heh.

[Ryujin2003]

23 minutes ago, leadeater said:

Not sure, L2 cache is supposed to be significantly cheaper and cache logic between levels likely plays a big part. Not expert enough to say without giving misinformation, I'd have to read up on it a bit before going in to that kind of discussion. Last time I bothered reading in to CPU cache was way back in 2000's comparing the then AMD Opterons to Intel Xeons and how AMD were absolutely crushing Intel in the L1 and L2 department and above dual socket efficiency, how things have changed heh.

It could also be that Intel owns their own fab facilities, and AMD uses 3rd party now don't they? I'm pretty sure AMD sold off their plants a while back, and that could potentially be part of it?

[leadeater]

1 minute ago, Ryujin2003 said:

It could also be that Intel owns their own fab facilities, and AMD uses 3rd party now don't they? I'm pretty sure AMD sold off their plants a while back, and that could potentially be part of it?

Yea AMD is fabless, Intel does have the advantage of being able to completely tailor the entire process for themselves which in theory should have some advantages.

[DocSwag]

1 hour ago, leadeater said:

Yea AMD is fabless, Intel does have the advantage of being able to completely tailor the entire process for themselves which in theory should have some advantages.

I've been thinking too if Ryzen is profitable enough would it be feasible for AMD or buy GloFo? Afterall, I think AMD is their main customer and it probably would be cheaper for AMD to own them if they have enough demand. As a matter of fact, I can't think of anyone else other than AMD who buys chips from GloFo  

[captain cactus]

30 minutes ago, DocSwag said:

I've been thinking too if Ryzen is profitable enough would it be feasible for AMD or buy GloFo? Afterall, I think AMD is their main customer and it probably would be cheaper for AMD to own them if they have enough demand. As a matter of fact, I can't think of anyone else other than AMD who buys chips from GloFo  

Psst, GloFo was the manufacturing arm of AMD before it spun off into its own company in 2009. GloFo isn't just making AMD parts. They also make automotive parts and high-voltage power management stuff.

[DocSwag]

Just now, lots of unexplainable lag said:

Psst, GloFo was the manufacturing arm of AMD before it spun off into its own company in 2009. GloFo isn't just making AMD parts. They also make automotive parts and high-voltage power management stuff.

I know they were spun off from AMD, but I thought they only made CPUs and stuff, definitely not auto parts?

[captain cactus]

Just now, DocSwag said:

I know they were spun off from AMD, but I thought they only made CPUs and stuff, definitely not auto parts?

Nope: https://en.wikipedia.org/wiki/GlobalFoundries

 

Quote

The firm manufactures integrated circuits in high volume mostly for semiconductor companies such as AMD, Broadcom, Qualcomm, and STMicroelectronics.