AMD Ryzen HAS BEEN ANNOUNCED!!!

[MageTank]

2 minutes ago, djdwosk97 said:

 

There have been open-source benchmarks (which can easily be manipulated to favor one architecture over another), there have been gaming benchmarks (BF1 @ 4k -- which even an FX8320 does well in), and now the cinebench stuff. So, I'm withholding judgement until I've seen a full suite showing performance in all use cases. I've seen rumors of lacking AVX/FPU performance -- which could explain the low tdp and low price. 

 

I understand fully that Intel has ridiculous margins, but what advantage is there to undercutting Intel by such large margins? Intel will price match them anyway, so why not just price them 10-20% lower, where Intel is far less likely to react? They could be trying to get people with recent-ish i5s/i7s to upgrade because it's just "such a good upgrade", but at the same time, a cheap 6c/12 and enthusiast priced 8c/16t would have the same effect.  

 

Cinebench is definitely a good benchmark, but it also isn't all encompassing. 

I don't know about the FPU part, but I can help clarify the AVX issue. Ryzen can do 8x 128-bit op/clock or 8x 256-bit op/clock. Modern Intel i7's on the other hand, can do 16x 128-bit op/clock, or 8x 256-bit op/clock. Basically, Ryzen will be half as good at AVX1 as Intel, but should be on par with Intel at AVX2. Take from that what you will. 

[79wjd]

5 minutes ago, MageTank said:

I don't know about the FPU part, but I can help clarify the AVX issue. Ryzen can do 8x 128-bit op/clock or 8x 256-bit op/clock. Modern Intel i7's on the other hand, can do 16x 128-bit op/clock, or 8x 256-bit op/clock. Basically, Ryzen will be half as good at AVX1 as Intel, but should be on par with Intel at AVX2. Take from that what you will. 

Like I said, I'm withholding judgement until after I've seen a slew of reliable benchmarks showing various types of performance. 

 

I'm cautiously hopeful for Ryzen.

[Drak3]

11 minutes ago, djdwosk97 said:

I understand fully that Intel has ridiculous margins, but what advantage is there to undercutting Intel by such large margins? Intel will price match them anyway, so why not just price them 10-20% lower, where Intel is far less likely to react? They could be trying to get people with recent-ish i5s/i7s to upgrade because it's just "such a good upgrade", but at the same time, a cheap 6c/12 and enthusiast priced 8c/16t would have the same effect.

AMD needs to recover reputation as a viable option. Not going to happen if they price just under Intel, especially after the 9590.

 

And both companies know this. Intel also knows that they need AMD to succeed so that Intel doesn't get considered a monopoly and split into multiple companies. They're probably not going to give a strong reaction. They're more likely to let AMD eat up some marketshare for a small while, and then, provided Intel is still top dog by 60+%, do all they can to push AMD down to where they're not a perceivable threat for a while.

[done12many2]

53 minutes ago, MageTank said:

That's not my motto. My motto is "stable until unstable", lol. If your 5ghz hasn't crashed yet, it's stable. When it crashes, it's no longer stable. It's that simple.

 

It's probably safe to assume that your custom loop out-performs any air coolers or AIO's on the market, and you likely run at a low ambient, so i'll give you the benefit of the doubt. However, I'll let a quote from a friend of mine rest here.

 

I appreciate that.  

 

Quote

The impact thermals have on your required voltage is pretty significant. I am sure you know this, and have likely tested this. I've seen people require less volts after delidding to maintain the exact same clock speeds pre-delid. I've also seen people maintain the same voltage, and achieve higher clock speeds. Improving ones thermal headroom, especially with superior conduction, is a very big deal when it comes to overclocking. I'll amend my previous statement to this: Nobody is going to hit 5ghz on any modern i7 without a delid OR a very decent custom loop. No air cooler/AIO on the market can handle high clocks AND high voltage without a delid. 

 

The majority of 7700k chips (a lot tested) that I've tested saw a decent drop in voltage from IHS intact to delidded, or like you said, an increase in clock speed at the same voltage (same thing).  

 

With my latest setup, I've been testing all of these chips on an open air bench with a Corsair H115i with fans maxed.  As you already know, AIOs like to heatsoak over time and I've found that the H115i is good at minimizing this (with fans at max) compared to something like a H100i with smaller/weaker fans.  If an AIO begins to heatsoak, it's all over.  I've delidded the ones that make the cut from that point.  Way too much hassle to throw all of these chips into my larger custom loop.  

 

Kaby Lake definitely runs a bit warmer than Skylake at the same voltages, but Kaby Lake also produces higher clock speeds at those same voltages.  This is both before and after deilidding. 

 

I forgot to address your comment about me running RealBench for a "limited amount of time".  Per the requirements for submission, both of the 5.3 GHz chips that I have were run for 8 hours despite the fact that I considered that way longer than necessary.  It also happens to be the max time you can select on the test.  I have run 1hr tests on other chips, but I'm not submitting those to the statistics thread and that happens to be what I'm more comfortable with.  I'm not a big fan of running long test just to do it.  

 

 

1 hour ago, MageTank said:

At what voltage and temps? I'd be very curious to see those results. From what I saw when we last talked, you ran realbench for a very limited amount of time per-chip. Not exactly what I would personally deem "stable", lol. Poor @Lays still can't do 48k.

50 minutes ago, leadeater said:

Also my stability requirement and stress for how I use my computer is different to other people. My computer never gets turned off, ever. Not matter what it needs to be running without crashing or small issues or require an odd reboot. It needs to be powered on and can live like that for years straight if I want it to. My CPU is overclocked to the point at which this is true, not what it can get and not what people would call stable based on some rules a collective came up with and now accept as the standard.

 

I only run those stability test when submitting to the statistics thread.  I find value in the centralized collection of that type of data and therefor contribute.  I as well don't necessarily base my on personal requirements on the someone else's standard.  However, when someone asks you about your overclock's stability as @MageTank did, you have to be able to articulate or quantify that in some way other than "trust me, it hasn't crashed yet."

 

 

42 minutes ago, MageTank said:

I dare anyone to try that with memory. You will learn quickly that ram instability is by far the worst, lol. 

 

I keep a fresh Windows image on hand just because of memory testing.  

[79wjd]

13 minutes ago, Drak3 said:

AMD needs to recover reputation as a viable option. Not going to happen if they price just under Intel, especially after the 9590.

 

And both companies know this. Intel also knows that they need AMD to succeed so that Intel doesn't get considered a monopoly and split into multiple companies. They're probably not going to give a strong reaction. They're more likely to let AMD eat up some marketshare for a small while, and then, provided Intel is still top dog by 60+%, do all they can to push AMD down to where they're not a perceivable threat for a while.

They also said they don't want to be the known as the cheaper option. And frankly, I don't think being cheaper is going to make anyone trust them more, just more willing to risk it. And again, a 6c/12t priced like a 7700k, 4c/8t priced like an i5, and a 4c/4t priced like an i3 would still do that just fine.

 

And also, Intel will follow suit and cutting prices negating any benefits of risking it with AMD.

[Drak3]

3 minutes ago, djdwosk97 said:

They also said they don't want to be the known as the cheaper option. And frankly, I don't think being cheaper is going to make anyone trust them more, just more willing to risk it. And again, a 6c/12t priced like a 7700k, 4c/8t priced like an i5, and a 4c/4t priced like an i3 would still do that just fine.

 

And also, Intel will follow suit and cutting prices negating any benefits of risking it with AMD.

AMD has no other option. As it stands, Intel has a long standing reputation of being the only viable x86 manufacturer for the vast majority of use cases and budgets. AMD's new product doesn't appeal on a pure performance standpoint, to it has to have major appeal to stand a chance. That would be price to performance. Even then, people will be wary until Ryzen has been used enough that it's realistically being recommended over Intel's offerings for more than the $100 or less build.

 

That's not going to be an instant thing. AMD knows this, Intel knows this. Intel has brand recognition, and they can maintain their current pricing and make money hand over fist. Plus, Ryzen fails, and it's more than likely AMD goes under. This happens, Intel becomes a monopoly (and forced to split into multiple companies), and they lose access to the AMD64 patents. Both are unfavorable for Intel.

[79wjd]

5 minutes ago, Drak3 said:

AMD has no other option. As it stands, Intel has a long standing reputation of being the only viable x86 manufacturer for the vast majority of use cases and budgets. AMD's new product doesn't appeal on a pure performance standpoint, to it has to have major appeal to stand a chance. That would be price to performance. Even then, people will be wary until Ryzen has been used enough that it's realistically being recommended over Intel's offerings for more than the $100 or less build.

 

That's not going to be an instant thing. AMD knows this, Intel knows this. Intel has brand recognition, and they can maintain their current pricing and make money hand over fist. Plus, Ryzen fails, and it's more than likely AMD goes under. This happens, Intel becomes a monopoly (and forced to split into multiple companies), and they lose access to the AMD64 patents. Both are unfavorable for Intel.

Again, undercutting Intel by such significant margins doesn't help since Intel will match them. Furthermore, AMD recently said they don't want to be known as the cheaper option because it doesnt work.

[leadeater]

@MageTank @djdwosk97 @done12many2

Nice to actually talk to people who know what overclocking means and how you can't create a rule book for it .

[Colonel_Gerdauf]

I'd like to be optimistic, but to be brutally honest, the information released thus far reeks of egregious cherry-picking. May I remind you guys that AMD has had a lengthy history of doing this, even with the recent 400 series Radeon GPUs?

[ivan134]

2 minutes ago, Colonel_Gerdauf said:

I'd like to be optimistic, but to be brutally honest, the information released thus far reeks of egregious cherry-picking. May I remind you guys that AMD has had a lengthy history of doing this, even with the recent 400 series Radeon GPUs?

What specifically that's making you use such hyperbolic language? And what did they do with the 400 series?

[Lengran]

Finally, after so many years, AMD makes great CPU again

[MageTank]

1 minute ago, leadeater said:

@MageTank @djdwosk97 @done12many2

Nice to actually talk to people who know what overclocking means and how you can't create a rule book for it .

Many people believe we have secrets to overclocking, but I am just as clueless as anyone else. People seem to perceive me as some sort of memory guru, but truth me told, i just touch random settings and see what happens after I post, lol. 

 

People also pull random numbers and information out of thin air. "X CPU is safe until X voltage", "X memory voltage will kill X IMC", "Intel only recommends X speeds/voltages, any higher will cause damage". All of this is simply wrong, or impossible to prove. Nobody knows what the max safe voltage is for 24/7 use on any CPU, because it differs per user and per piece of silicon. Not all chips are equal, and some will see the effects of running higher volts before others. Some can tolerate it for long periods of time, while others choke immediately. Intel themselves can't really have the answer, unless you honestly expect them to spend years testing many different voltages simultaneously.

 

Intel's whitesheet recommendations follow a specific formula. Current/voltage/clocks will give them their maximum possible voltage before immediate failure. That is what they list. The "max supported memory speed" is simply the JEDEC standard that they personally tested with. They didn't go out and buy several different kits of different speeds/IC's to test and validate for official support. Their "official support/limitations" are simply guidelines. Remain within them, and you basically have a guarantee that your chip will function fine under normal circumstances. Go beyond these specifications, and you are on your own. They are by no means a real limitation.

 

As for individual overclockers and their "rules of overclocking", it's all nonsense. I don't care how great of an overclocker you are, or how great your hardware is, your methodology will never be flawless due to one simple reason. Time. It impacts all overclocks eventually. I've been running my ram at 3600 C14-14-14-28-2 for almost a year now, and it just recently started showing signs of instability. Granted, I know exactly why (it's not degradation, but ambient temps increasing combined with high tREFI) but it goes to show you that no matter how extensive I tested this memory, it eventually failed. Stable until unstable, lol.

 

When it comes to overclocking, you just have to do the best you can, with whatever you've got. Try to keep thermals low, try to keep realistic expectations (don't go into it demanding X clock speed straight out of the gate) and be prepared to compromise. That last part was the hardest for me to understand early on, but eventually led to me being much happier with my hardware. I tortured a PANRAM DDR4 kit because I absolutely refused to stop until I made 3200 C15 stable (this was before Samsung D/B die, and before any 3000+ kits were sold). Needless to say, I lost a lot of sleep over something so pointless. In the end, I discovered tertiary timings, and how great 3000 C15 + tight terts is when compared to 3200 C15 with absolutely no tertiary headroom due to a completely exhausted IMC. 

 

I've ranted on for too long now, time to go do something productive. 

[arnavvr]

On 2/22/2017 at 6:13 AM, Ekst4zy said:

I still want to see this game benchmarks

 

[79wjd]

17 minutes ago, Colonel_Gerdauf said:

I'd like to be optimistic, but to be brutally honest, the information released thus far reeks of egregious cherry-picking. May I remind you guys that AMD has had a lengthy history of doing this, even with the recent 400 series Radeon GPUs?

I don't know, cinebench is a pretty good and relevant metric. It's not indicative of complete performance in many tasks, but it's a good sign nonetheless. 

[done12many2]

6 minutes ago, MageTank said:

Spoiler

 

Many people believe we have secrets to overclocking, but I am just as clueless as anyone else. People seem to perceive me as some sort of memory guru, but truth me told, i just touch random settings and see what happens after I post, lol. 

 

People also pull random numbers and information out of thin air. "X CPU is safe until X voltage", "X memory voltage will kill X IMC", "Intel only recommends X speeds/voltages, any higher will cause damage". All of this is simply wrong, or impossible to prove. Nobody knows what the max safe voltage is for 24/7 use on any CPU, because it differs per user and per piece of silicon. Not all chips are equal, and some will see the effects of running higher volts before others. Some can tolerate it for long periods of time, while others choke immediately. Intel themselves can't really have the answer, unless you honestly expect them to spend years testing many different voltages simultaneously.

 

Intel's whitesheet recommendations follow a specific formula. Current/voltage/clocks will give them their maximum possible voltage before immediate failure. That is what they list. The "max supported memory speed" is simply the JEDEC standard that they personally tested with. They didn't go out and buy several different kits of different speeds/IC's to test and validate for official support. Their "official support/limitations" are simply guidelines. Remain within them, and you basically have a guarantee that your chip will function fine under normal circumstances. Go beyond these specifications, and you are on your own. They are by no means a real limitation.

 

As for individual overclockers and their "rules of overclocking", it's all nonsense. I don't care how great of an overclocker you are, or how great your hardware is, your methodology will never be flawless due to one simple reason. Time. It impacts all overclocks eventually. I've been running my ram at 3600 C14-14-14-28-2 for almost a year now, and it just recently started showing signs of instability. Granted, I know exactly why (it's not degradation, but ambient temps increasing combined with high tREFI) but it goes to show you that no matter how extensive I tested this memory, it eventually failed. Stable until unstable, lol.

 

When it comes to overclocking, you just have to do the best you can, with whatever you've got. Try to keep thermals low, try to keep realistic expectations (don't go into it demanding X clock speed straight out of the gate) and be prepared to compromise. That last part was the hardest for me to understand early on, but eventually led to me being much happier with my hardware. I tortured a PANRAM DDR4 kit because I absolutely refused to stop until I made 3200 C15 stable (this was before Samsung D/B die, and before any 3000+ kits were sold). Needless to say, I lost a lot of sleep over something so pointless. In the end, I discovered tertiary timings, and how great 3000 C15 + tight terts is when compared to 3200 C15 with absolutely no tertiary headroom due to a completely exhausted IMC. 

 

I've ranted on for too long now, time to go do something productive. 

 

 

 

Very well said.  

 

The "did I win?" shit is tiring too.  There's no way of knowing with any certainty that one guy's chip in Canada is better than another's in Mexico.  I guess they could both fly to Vegas (hell yeah) and run them on the same system, but short of that, there are far too many variables for comparing via the internet.  

 

By the way @Lays I do know who's chip is indeed better.  

[MageTank]

1 minute ago, done12many2 said:

By the way @Lays I do know who's chip is indeed better.  

The answer? Anyone that can watch 4k Netflix! Take that, potatochips.

[done12many2]

1 minute ago, MageTank said:

The answer? Anyone that can watch 4k Netflix! Take that, potatochips.

 

I read that right in the middle of taking a drink.  Thanks!!!

[arnavvr]

7 minutes ago, MageTank said:

The answer? Anyone that can watch 4k Netflix! Take that, potatochips.

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

[MageTank]

1 minute ago, arnavvr said:

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

That was the joke, lol. 

[79wjd]

18 minutes ago, arnavvr said:

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

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.

[done12many2]

2 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.

 

Everything in the pipeline to include the screen has to be HDCP 1.4 2.2 compliant and can't be piped through dGPU.  This means that you have to use a 7700k, output video via onboard video with a 1.4 2.2 compliant controller to a 1.4 2.2 compliant screen.  DRM is AWESOME!!  It pretty much renders any high end system useless for watching 4k Netflix without redoing all your wiring just to watch it.  Don't you love how Intel just makes it sound so simple?

 

I guess technically you could run a monitor with two inputs and just switch between onboard video and discrete GPUs as needed, but it still stupid.

[Vash.]

Pre-ordered 1800x yesterday.. fingers crossed AUS gets enough. Decided to make this little bit of humour to mark the occasion 

 

[Lays]

1 hour ago, done12many2 said:

 

Everything in the pipeline to include the screen has to be HDCP 1.4 compliant and can't be piped through dGPU.  This means that you have to use a 7700k, output video via onboard video with a 1.4 compliant controller to a 1.4 compliant screen.  DRM is AWESOME!!  It pretty much renders any high end system useless for watching 4k Netflix without redoing all your wiring just to watch it.  Don't you love how Intel just makes it sound so simple?

 

I guess technically you could run a monitor with two inputs and just switch between onboard video and discrete GPUs as needed, but it still stupid.

Hdcp 2.2 compliant you mean. 

[done12many2]

8 minutes ago, Lays said:

Hdcp 2.2 compliant you mean. 

 

Yeah, that's it.  About time you woke up.

 

**I fixed my post.  Fix your quote so I don't look so dumb.  

[ivan134]

2 hours ago, done12many2 said:

 

Everything in the pipeline to include the screen has to be HDCP 1.4 2.2 compliant and can't be piped through dGPU.  This means that you have to use a 7700k, output video via onboard video with a 1.4 2.2 compliant controller to a 1.4 2.2 compliant screen.  DRM is AWESOME!!  It pretty much renders any high end system useless for watching 4k Netflix without redoing all your wiring just to watch it.  Don't you love how Intel just makes it sound so simple?

 

I guess technically you could run a monitor with two inputs and just switch between onboard video and discrete GPUs as needed, but it still stupid.

Of course I loves me some DRM. Now I'll go back to playing that game that was exclusive to i7s.