AMD Ryzen HAS BEEN ANNOUNCED!!!

[Lays]

3 hours ago, arnavvr said:

Funny part is that Darren can't watch 4K Netflix anyways due to having a Z170 board.

 

3 hours ago, MageTank said:

That was the joke, lol. 

 

3 hours ago, djdwosk97 said:

Wait, not only do you need a kaby lake CPU to watch 4k Netflix, you also need a kaby lake motherboard? 

 

Of course you do.

The sad thing is, you don't need a kaby Lake board you need a special board with an add on chip that like upscales the display port signal to be hdcp2.2 compliant.  It's reeeetardedeed

[LAwLz]

8 hours ago, flipped_bit said:

Apples to oranges.  The 1700 (non-X) turbos at 3.7GHz.  The 7700K turbos at 4.5GHz.  

Are you saying that the benchmarks are apples vs oranges? Because that's like saying you shouldn't compare the 7700K vs the 1700 in Cinebench multithreading because "it's 4 cores vs 8".

Ryzen's 8 core chip does not clock as high as Intel's quad core, and on top of that it's behind in IPC. You can't just say the benchmark is invalid just because it measures the type of scenario where Ryzen appears to be behind, just like you can't say a benchmark is invalid when it measures the type of scenario where Ryzen is ahead.

 

The benchmark is not apples to oranges, and I am getting really sick and tired of people say it is.

 

7 hours ago, Sprawlie said:

Yes, so far it's looking like the i7-7700k is still the best gaming CPU. But what about gaming and twitch streaming? what are your in game FPS going to look like on the i7-7700k v the new Ryzen chips? What about if you're trying to game while your computer is running a flash video stream on one display, websites and other functionality on a 2nd display, and trying to game GTA @ 2k on Ultra!

Depending on what GPU you got, the i7 might still be better for streaming.

Reason: QuickSync

 

Streaming with an i7 does not add any extra strain on the CPU. Streaming with a Ryzen chip will, unless your discrete GPU can do it for you (like with an Nvidia card) and at that point, you don't really need a beefy CPU.

[Citadelen]

17 hours ago, Kloaked said:

-snip-

He's had me on ignore for months now.

[Gdourado]

Do Rysen CPU's come with a soldered heat-spreader like the previous AM3 processors?

Or will delliding be necessary?

[Lays]

15 minutes ago, Gdourado said:

Do Rysen CPU's come with a soldered heat-spreader like the previous AM3 processors?

Or will delliding be necessary?

Soldered yes, the die is still quite large so all of em should be soldered.

[Sprawlie]

6 hours ago, LAwLz said:

Are you saying that the benchmarks are apples vs oranges? Because that's like saying you shouldn't compare the 7700K vs the 1700 in Cinebench multithreading because "it's 4 cores vs 8".

Ryzen's 8 core chip does not clock as high as Intel's quad core, and on top of that it's behind in IPC. You can't just say the benchmark is invalid just because it measures the type of scenario where Ryzen appears to be behind, just like you can't say a benchmark is invalid when it measures the type of scenario where Ryzen is ahead.

 

The benchmark is not apples to oranges, and I am getting really sick and tired of people say it is.

 

 

Depending on what GPU you got, the i7 might still be better for streaming.

Reason: QuickSync

 

Streaming with an i7 does not add any extra strain on the CPU. Streaming with a Ryzen chip will, unless your discrete GPU can do it for you (like with an Nvidia card) and at that point, you don't really need a beefy CPU.

 

You're right on the streaming. Didn't realize how much better the new Nvidia GPU's are for this. Last time I was streaming I had been using a 290x which, did a horrible job, so i was software encoding. 

 

Just fired everything up and set it to Nvidia NVEC in OBS, and CPU usage was < 5%.

 

However, Pulling up my NHL.com hockey stream used 15% of my CPU and caused my GTA to dip < 30fps in some areas. But as I said earlier, I know I have CPU bottleneck right now with the 4670. I'm sure the 7700k will solve that issue for me. But, RyZen with it's more cores for the multi-tasking while gaming is promising

 

still won't buy anything until I see real world tests. Benchmarks don't tell you how well these things do always in real world scenarios like mine.

 

[Doobeedoo]

7 hours ago, LAwLz said:

Are you saying that the benchmarks are apples vs oranges? Because that's like saying you shouldn't compare the 7700K vs the 1700 in Cinebench multithreading because "it's 4 cores vs 8".

Ryzen's 8 core chip does not clock as high as Intel's quad core, and on top of that it's behind in IPC. You can't just say the benchmark is invalid just because it measures the type of scenario where Ryzen appears to be behind, just like you can't say a benchmark is invalid when it measures the type of scenario where Ryzen is ahead.

 

The benchmark is not apples to oranges, and I am getting really sick and tired of people say it is.

 

 

Depending on what GPU you got, the i7 might still be better for streaming.

Reason: QuickSync

 

Streaming with an i7 does not add any extra strain on the CPU. Streaming with a Ryzen chip will, unless your discrete GPU can do it for you (like with an Nvidia card) and at that point, you don't really need a beefy CPU.

Using x264 over Quick Sync or NVENC produces better quality and some settings don't work on those two no?

[MageTank]

2 hours ago, Lays said:

Soldered yes, the die is still quite large so all of em should be soldered.

Maybe on the 8 core SKU's, but I doubt the smaller ones will be soldered. We will know in a week or two though. I plan on buying one just to test the memory controller, so I'll likely delid it just to check. Time to prepare the heatgun, so I don't rip it in half like that poor dude did with his 5960X on OCN, lol. 

[captain cactus]

14 minutes ago, MageTank said:

Maybe on the 8 core SKU's, but I doubt the smaller ones will be soldered. We will know in a week or two though. I plan on buying one just to test the memory controller, so I'll likely delid it just to check. Time to prepare the heatgun, so I don't rip it in half like that poor dude did with his 5960X on OCN, lol. 

So far all AMD CPUs, high-end or low-end, have been soldered making de-lidding impossible. There's no reason for AMD to solder one SKU of Ryzen and then use a TIM on another one.

[LAwLz]

2 hours ago, Sprawlie said:

However, Pulling up my NHL.com hockey stream used 15% of my CPU and caused my GTA to dip < 30fps in some areas. But as I said earlier, I know I have CPU bottleneck right now with the 4670. I'm sure the 7700k will solve that issue for me. But, RyZen with it's more cores for the multi-tasking while gaming is promising

That doesn't sound right. A 4670 should be more than enough for GTA 5. Hell, even my 2500K is.

 

25 minutes ago, Doobeedoo said:

Using x264 over Quick Sync or NVENC produces better quality and some settings don't work on those two no?

Depends on the settings. x264 has a greater range of speed:quality settings. The default is the "very fast" preset (if I recall correctly) and that should be pretty comparable to NVENC or Quicksync. Once you start recording at 60 FPS and/or HEVC though, I wouldn't be surprised if GPU encoders becomes better (if you need to constantly encode 60 frames per second).

 

For archival purposes when you can afford to spend like 4 hours encoding a single video, x264 and (probably) x265 will beat the crap out of GPU encoders in terms of size and quality.

[MageTank]

Just now, lots of unexplainable lag said:

So far all AMD CPUs, high-end or low-end, have been soldered making de-lidding impossible. There's no reason for AMD to solder one SKU of Ryzen and then use a TIM on another one.

There is a ton of reasons for them to do so. You are referring to their 28nm+ SKU's. They are 14nm now. The 8 core SKU's should, in theory, have twice the surface area to work with, so they will be fine. The smaller ones on the other hand, will risk cracking under the soldering process. Read up on it here: http://overclocking.guide/the-truth-about-cpu-soldering/

 

You cannot use 28nm and above being soldered as an excuse to assume 14nm and below will be soldered. Look at Intel. Their 6c SKU's and above are soldered, but 4c and below are not. There is a reason for this. 

[79wjd]

5 minutes ago, lots of unexplainable lag said:

There's no reason for AMD to solder one SKU of Ryzen and then use a TIM on another one.

There is, for the same reason that Intel HEDT chips are soldered while their consumer chips are not. Size. Soldering consumer Intel chips would severely damage the chips (due to thermal expansion or something like that -- small die = problems for soldering). Presumably, the R3 and possibly R5 dies are small enough that soldering isn't feasible. 

[done12many2]

1 minute ago, MageTank said:

There is a reason for this. 

 

Yeah, so der8auer, Rockit and Coollaboraty can get some love.  

 

Seriously though, great explanation.

[TheMissxu]

7 minutes ago, LAwLz said:

Depends on the settings. x264 has a greater range of speed:quality settings. The default is the "very fast" preset (if I recall correctly) and that should be pretty comparable to NVENC or Quicksync. Once you start recording at 60 FPS and/or HEVC though, I wouldn't be surprised if GPU encoders becomes better (if you need to constantly encode 60 frames per second).

 

For archival purposes when you can afford to spend like 4 hours encoding a single video, x264 and (probably) x265 will beat the crap out of GPU encoders in terms of size and quality.

My understanding is that CPU encoding simply outrivals NVENC, Quicksync, and VCE when it comes to streaming. This is due to the fact that twitch limits all streamers (aside from special cases) bitrate to 3,500, which GPU encoders handle very poorly. Youtube streaming does allow for a 10,000+ bitrate, but you'd need an upload plan to match,

[Sprawlie]

17 minutes ago, LAwLz said:

That doesn't sound right. A 4670 should be more than enough for GTA 5. Hell, even my 2500K is.

 

Depends on the settings. x264 has a greater range of speed:quality settings. The default is the "very fast" preset (if I recall correctly) and that should be pretty comparable to NVENC or Quicksync. Once you start recording at 60 FPS and/or HEVC though, I wouldn't be surprised if GPU encoders becomes better (if you need to constantly encode 60 frames per second).

 

For archival purposes when you can afford to spend like 4 hours encoding a single video, x264 and (probably) x265 will beat the crap out of GPU encoders in terms of size and quality.

the 4670, when dedicated 100% towards the game keeps up fine. with FPS in that ranges from 35-70, depending on area of the map and whats on screen.

 

It's as i said earlier. I tend not to just game when i'm playing though. I tend to have GTA 5, NHL.com flash based stream going in chrome, Discord, Slack, few other random tabs in Chrome open. I'm running a 2560x1080p display for gaming, and 2 1080p displays for other activities. 

 

So it's all that "other" stuff which hurts my game performance. As I'm sharing 4 cores amongst all of that.

 

[Space Reptile]

19 hours ago, LAwLz said:

Thanks to a certain reptile, I have the benchmarks here.

making a name for myself eh? 

 

[Guest Darth Revan]

I know this question can't really be answered until we actually have the results of how Ryzen actually performs, but I want to ask what would you guys do.

I had an AMD Phenom first generation until two weeks ago. My mobo died, so in the house I also had an old intel 775 mobo and a dual-core processor. So that's what I'm using now. Now initially I was planning on getting a new video card next month and later on this year or even beginning of 2018 to change the rest.

Now my budget is limited, so if I were to go for Ryzen, I would go for their R3 line-up and later on if I need upgrade the processor.

Thing is the R3 from what I understood comes out 2H, meaning second half of this year. So that could be in November.

However if I go for Intel, I would most likely go for the Pentium G4560, the one that's very close to the i3 which were I live is 3 times the price of the Pentium or maybe that's the price everywhere. The Pentium is a dual-core with hyper-threading. But I would get it next month and at the end of the year I'll get a video card.

 

So the question is. Wait for Ryzen and get a proper quad-core but once again I mention we really know how it's going to perform, or go for Intel next month and pray that Ryzen doesn't kick its arse?

What would you guys do?

 

 

 

 

 

[TVwazhere]

1 minute ago, Darth Revan said:

So the question is. Wait for Ryzen and get a proper quad-core but once again I mention we really know how it's going to perform, or go for Intel next month and pray that Ryzen doesn't kick its arse?

What would you guys do?

I personally am buying the I3-7100 (the g4560 equivalent ish) ONLY because i have a deadline in May and cant wait. Otherwise I would wait, because 4 cores are better than 2 cores with hyper threading. (it can also OC so)

[MageTank]

6 minutes ago, Darth Revan said:

I know this question can't really be answered until we actually have the results of how Ryzen actually performs, but I want to ask what would you guys do.

I had an AMD Phenom first generation until two weeks ago. My mobo died, so in the house I also had an old intel 775 mobo and a dual-core processor. So that's what I'm using now. Now initially I was planning on getting a new video card next month and later on this year or even beginning of 2018 to change the rest.

Now my budget is limited, so if I were to go for Ryzen, I would go for their R3 line-up and later on if I need upgrade the processor.

Thing is the R3 from what I understood comes out 2H, meaning second half of this year. So that could be in November.

However if I go for Intel, I would most likely go for the Pentium G4560, the one that's very close to the i3 which were I live is 3 times the price of the Pentium or maybe that's the price everywhere. The Pentium is a dual-core with hyper-threading. But I would get it next month and at the end of the year I'll get a video card.

 

So the question is. Wait for Ryzen and get a proper quad-core but once again I mention we really know how it's going to perform, or go for Intel next month and pray that Ryzen doesn't kick its arse?

What would you guys do?

 

 

 

 

 

I say go with the pentium as a stopgap. Use it until you can afford a real Intel quadcore, and upgrade later on. Sure, AMD's quad core might be cheaper, but it's gonna be several months before we even see it on the market. In that amount of time, you should be able to save up enough cash to buy a real Intel quad core. Let's be real, it's not like AMD's quad core is going to out-perform Intel's, so you will still end up with better performance, it will just come at a higher cost.

[DocSwag]

I know this is a known fact, but why exactly do larger dies make solder more feasible? I can't wrap my mind around the physics of this, do any of you mind explaining?

[MageTank]

Just now, DocSwag said:

I know this is a known fact, but why exactly do larger dies make solder more feasible? I can't wrap my mind around the physics of this, do any of you mind explaining?

Thermal cycling damages solder via voids forming within the solder itself. As these voids form, thermal conductivity suffers, and thermal resistance is increased. Eventually, this will cause cracks to form in the solder (normally starting from the corners of the die) which will eventually lead to catastrophic failure for the CPU. The reason smaller dies cannot handle the soldering process, is because a smaller surface heats faster without spreading it out enough, resulting in the cracks. We are talking about thin glass after all. With a bigger surface area (specifically on the IHS side of the joint), the solder will cool faster. Not only that, but a bigger chip means more silicon which means less hotspots. The end result is less cracking.

[Vode]

6 minutes ago, MageTank said:

Thermal cycling damages solder via voids forming within the solder itself. As these voids form, thermal conductivity suffers, and thermal resistance is increased. Eventually, this will cause cracks to form in the solder (normally starting from the corners of the die) which will eventually lead to catastrophic failure for the CPU. The reason smaller dies cannot handle the soldering process, is because a smaller surface heats faster without spreading it out enough, resulting in the cracks. We are talking about thin glass after all. With a bigger surface area (specifically on the IHS side of the joint), the solder will cool faster. Not only that, but a bigger chip means more silicon which means less hotspots. The end result is less cracking.

But why was a soldered IHS used until Sandy Bridge?

[79wjd]

5 minutes ago, Vode said:

But why was a soldered IHS used until Sandy Bridge?

Sandy Bridge quad core dies are 216mm^2

Skylake quad cores dies are 122mm^2

[MageTank]

4 minutes ago, Vode said:

But why was a soldered IHS used until Sandy Bridge?

The die size was physically larger back then. Nehalem was 45nm. We are currently at 14nm. Quite a drastic difference in size if you ask me.

[Vode]

30 minutes ago, MageTank said:

The die size was physically larger back then. Nehalem was 45nm. We are currently at 14nm. Quite a drastic difference in size if you ask me.

Yeah, but wasn't Penryn around 110mm2? 

 

I'm aware of the difference in nodes die size is what matters.