8700k benchmarks leaked.

[Guest lilbman]

I'm going to guess that their 8700HQ and 8820HK are going to be 6c/12t.

[ravenshrike]

1 minute ago, mr moose said:

tdp is tdp, it's an industry spec just like horsepower or watts.

What? No seriously, what? A better comparison would be to say Intel's TDP is akin to VWs emissions standards.

[mr moose]

2 minutes ago, ravenshrike said:

What? No seriously, what? A better comparison would be to say Intel's TDP is akin to VWs emissions standards.

come back when you have an education on the topic.

[Dash Lambda]

26 minutes ago, mr moose said:

TDP is TDP, it's an industry spec just like horsepower or watts. Intel do not "guess" the TDP, it is a measured and tested spec.

Yes, that is true, but two things are suspect:

-We don't know Intel's testing methodology, so we don't know what load the TDP is attributed to. It would be nice if it were peak power at stock settings, but it's not, so there's ambiguity.

-Intel gives common TDP ratings to vastly different chips. I mean, the value does come from testing, but tell me: Do you think 140W is a severe underestimate for their 12-core, or a severe overestimate for their 6-core? ... Or maybe it's both?

[ravenshrike]

41 minutes ago, mr moose said:

come back when you have an education on the topic.

So what is the testing spec in comparison to real world usage then? What programs exactly are used and how do they compare to real world usage patterns?

[Brooksie359]

52 minutes ago, mr moose said:

tdp is tdp, it's an industry spec just like horsepower or watts.  Intel do not "guess" the TDP, it is a measured and tested spec.

more or less,  there are too many parameters outside of Intel's control when enthusiasts overclock.   The idea of TDP is that pc designers can implement a suitable thermal solution for the conditions they expect, I.E if they are selling an overclocked machine they will know they need a cooler larger than the rated TDP to handle the output.

 

 

People seem to think that TDP is a concrete unchanging figure and that because the thermal output of an overclocked or overworked CPU is higher, then the TDP figure is wrong or "guessed", this is not the case, it is calculated, and if you are designing your PC to operate outside that definition then you need to understand how it is calculated.   I can only assume it is lucky aftermarket coolers are generally better than rated (marketed), because with the number of people who don't understand TDP we'd have a lot more people complaining about their CPU's unnecessarily throttling and not know knowing why.

TDP is a spec but if you are trying to tell me that all the Intel cpus all have exactly the rated TDP they state I would have to disagree. The TDP is likely a maximum that is allowed for that part and any cpu tested that are outside of that range are designated defective. You seem to be under the impression that the advertised TDP has to be 100% accurate when in reality that is basically impossible. How can you put a TDP rating that is going to be 100% accurate for all chip of a certain line of cpus? you can't which is why I said it is an estimate and while it is likely tested to make sure it doesn't exceed that TDP at stock clocks it is not an exact representation of an individual's cpu's TDP. 

[Lays]

JFC, people bitch and moan for years about wanting 6 cores on the mainstream platform, now it's going to happen and everyone cries about it still.

 

Remember how much headroom Ryzen chips have when overclocking, and how much headroom 7700k's have, now add 2 cores and a better process, think about the result when both are overclocked

 

[mr moose]

22 minutes ago, Dash Lambda said:

Yes, that is true, but two things are suspect:

-We don't know Intel's testing methodology, so we don't know what load the TDP is attributed to. It would be nice if it were peak power at stock settings, but it's not, so there's ambiguity.

-Intel gives common TDP ratings to vastly different chips. I mean, the value does come from testing, but tell me: Do you think 140W is a severe underestimate for their 12-core, or a severe overestimate for their 6-core? ... Or maybe it's both?

Not really

https://www.google.com.au/url?sa=t&rct=j&q=&esrc=s&source=web&cd=13&cad=rja&uact=8&ved=0ahUKEwiW06Su3PjVAhXFW7wKHXjxBDkQFghqMAw&url=https%3A%2F%2Fwww.intel.com%2Fcontent%2Fdam%2Fdoc%2Fwhite-paper%2Fresources-xeon-measuring-processor-power-paper.pdf&usg=AFQjCNHF5Ou8mUgHZlnUGnu0ez48BHMVIg

19 minutes ago, ravenshrike said:

So what is the testing spec in comparison to real world usage then? What programs exactly are used and how do they compare to real world usage patterns?

real world usage changes depending on the user.  That is why Intel have  TDP.  It is up to the systrem designer to determine from the TDP what cooling solution is necessary for the CPU in question. See link above.

 

5 minutes ago, Brooksie359 said:

TDP is a spec but if you are trying to tell me that all the Intel cpus all have exactly the rated TDP they state I would have to disagree. The TDP is likely a maximum that is allowed for that part and any cpu tested that are outside of that range are designated defective. You seem to be under the impression that the advertised TDP has to be 100% accurate when in reality that is basically impossible. How can you put a TDP rating that is going to be 100% accurate for all chip of a certain line of cpus? you can't which is why I said it is an estimate and while it is likely tested to make sure it doesn't exceed that TDP at stock clocks it is not an exact representation of an individual's cpu's TDP. 

TDP as spec'd by Intel does have to be accurate.  Otherwise it is false advertising.   A TDP of 140Watts does not mean the chip will not go any higher.  It means that for a given load and clock that is the watts a thermal solution must handle to maintain Tj temp.  

 

It's not really that complicated.

[flibberdipper]

3 minutes ago, Lays said:

JFC, people bitch and moan for years about wanting 6 cores on the mainstream platform, now it's going to happen and everyone cries about it still.

 

Remember how much headroom Ryzen chips have when overclocking, and how much headroom 7700k's have, now add 2 cores and a better process, think about the result when both are overclocked

 

Only thing I can think to bitch about is that the pricing will probably be retarded and/or they'll pull an X299 and make you pay to unlock shit.

 

Also no soldered IHS.

[Brooksie359]

19 minutes ago, Dash Lambda said:

Yes, that is true, but two things are suspect:

-We don't know Intel's testing methodology, so we don't know what load the TDP is attributed to. It would be nice if it were peak power at stock settings, but it's not, so there's ambiguity.

-Intel gives common TDP ratings to vastly different chips. I mean, the value does come from testing, but tell me: Do you think 140W is a severe underestimate for their 12-core, or a severe overestimate for their 6-core? ... Or maybe it's both?

like with all engineering it is likely an underestimate for the 6 core and maybe a low factor of safety on the 12 core. I mean the fact that people think the advertised TDP is going to be the actual TDP is kinda silly. Its like people think the maximum capacity for an elevator is the actual max capacity. No, it is a conservative estimate based on the theoretical max capacity. 

[Lays]

1 minute ago, tmcclelland455 said:

Only thing I can think to bitch about is that the pricing will probably be retarded and/or they'll pull an X299 and make you pay to unlock shit.

 

Also no soldered IHS.

Delidding tools are so damn cheap now, IDK why people make it such a big deal, I delid every chip I buy and I don't have any problems with having to do that lol

[Brooksie359]

4 minutes ago, mr moose said:

Not really

https://www.google.com.au/url?sa=t&rct=j&q=&esrc=s&source=web&cd=13&cad=rja&uact=8&ved=0ahUKEwiW06Su3PjVAhXFW7wKHXjxBDkQFghqMAw&url=https%3A%2F%2Fwww.intel.com%2Fcontent%2Fdam%2Fdoc%2Fwhite-paper%2Fresources-xeon-measuring-processor-power-paper.pdf&usg=AFQjCNHF5Ou8mUgHZlnUGnu0ez48BHMVIg

real world usage changes depending on the user.  That is why Intel have  TDP.  It is up to the systrem designer to determine from the TDP what cooling solution is necessary for the CPU in question. See link above.

 

TDP as spec'd by Intel does have to be accurate.  Otherwise it is false advertising.   A TDP of 140Watts does not mean the chip will not go any higher.  It means that for a given load and clock that is the watts a thermal solution must handle to maintain Tj temp.  

 

It's not really that complicated.

You're joking right? the TDP can be higher than what is needed and it wont be false advertising. Again they are doing a conservative estimate based on testing like basically all other engineering. You test what it would actually take and you give a bigger number because that way you have some room for error. 

[dexT]

2 hours ago, Dabombinable said:

I'll just have to test it inside the rig I built for my friend, since its got everything that was replaced by the 4790K+Z97 Sabertooth MKII. Its got to be the CPU though as I could actually get the motherboard to give out a beep before it'd go back to not doing anything (apart from powering up for a second), the RAM is fine (in fact only 1 stick isn't brand new), and the PSU is powering my extremely temperamental LGA775 machine (its really picky when it comes to RAM+the slots where expansion cards/graphics card are installed)

Cool. The stipulation was going to be that I bench it sub-zero so that's better for you probably

[flibberdipper]

1 minute ago, Lays said:

Delidding tools are so damn cheap now, IDK why people make it such a big deal, I delid every chip I buy and I don't have any problems with having to do that lol

IDK Kinda irritating that you have to delid your 300+ chip just to get thermals that aren't total bullshit. It's not like it's impossible or too costly cuz AMD does it to Ryzen and yadda yadda.

[mr moose]

1 minute ago, Brooksie359 said:

Your joking right? the TDP can be higher than what is needed and it wont be false advertising. Again they are doing a conservative estimate based on testing like basically all other engineering. You test what it would actually take and you give a bigger number because that way you have some room for error. 

 

I'm not sure why this is so hard to understand,   TDP is for a pre-defined condition.  It's not an estimate, it is a measured output for said condition (ergo x load at base clock).

 

If you exceed the base clock or the load  used to determine TDP, then the chip power dissipation will be higher, Intel clearly states that. 

 

 

 

[Drak3]

1 hour ago, Dabombinable said:

Actually, wouldn't an old i7 5820K be better value?

Not really, if it turns out that the 8700K can hit near 5GHz, if not 5GHz even, on average. The 5820K, hitting around 4.5GHz on average, water, is going to be at around a 5% IPC disadvantage, 11% clock speed disadvantage, for a roughly 16.5% disadvantage, before figuring in IMC improvements and RAM overclocking potential.

 

Of course, assuming that the 14nm++ node can enable the hexacore to reach 5GHz on average.

[Drak3]

19 minutes ago, Lays said:

JFC, people bitch and moan for years about wanting 6 cores on the mainstream platform, now it's going to happen and everyone cries about it still.

Are, are you new here? Is that post count a bug?

 

20 minutes ago, Lays said:

Remember how much headroom Ryzen chips have when overclocking, and how much headroom 7700k's have, now add 2 cores and a better process, think about the result when both are overclocked

I haven't seen Tom reply yet, so I'll try my best to emulate the experience:

 

BUT THE VALUE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1111111111111111111111111111111111111111111111111ONEONEONEONEONEONE

 

But I like having 40PCIe lanes, so the 8700K has no value to me.

16 minutes ago, Brooksie359 said:

Its like people think the maximum capacity for an elevator is the actual max capacity. No, it is a conservative estimate based on the theoretical max capacity.

Difference being that people die if you exceed the actual capacity that an elevator can handle. An Intel chip just doesn't boost, or it throttles a bit. Plus, anyone getting an unlocked SKU with the intention of overclocking should be buying a cooler than can manage some level of overclocking. If they aren't, TDP is meaningless as they'll be lucky to see a clockspeed above 800MHz.

[Brooksie359]

Just now, mr moose said:

 

I'm not sure why this is so hard to understand,   TDP is for a pre-defined condition.  It's not an estimate, it is a measured output for said condition (ergo x load at base clock).

 

If you exceed the base clock or the load  used to determine TDP, then the chip power dissipation will be higher, Intel clearly states that. 

 

 

 

ok so lets say I buy a 7700k. Chances are the 7700k won't have the exact TDP tested on the chip or chips Intel used to determine TDP. 

[Lays]

4 minutes ago, Drak3 said:

Are, are you new here? Is that post count a bug?

 

I haven't seen Tom reply yet, so I'll try my best to emulate the experience:

 

BUT THE VALUE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1111111111111111111111111111111111111111111111111ONEONEONEONEONEONE

 

But I like having 40PCIe lanes, so the 8700K has no value to me.

Difference being that people die if you exceed the actual capacity that an elevator can handle. An Intel chip just doesn't boost, or it throttles a bit. Plus, anyone getting an unlocked SKU with the intention of overclocking should be buying a cooler than can manage some level of overclocking. If they aren't, TDP is meaningless as they'll be lucky to see a clockspeed above 800MHz.

And what do you use those 40 PCI-E lanes for?

[Brooksie359]

2 minutes ago, Drak3 said:

Are, are you new here? Is that post count a bug?

 

I haven't seen Tom reply yet, so I'll try my best to emulate the experience:

 

BUT THE VALUE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1111111111111111111111111111111111111111111111111ONEONEONEONEONEONE

 

But I like having 40PCIe lanes, so the 8700K has no value to me.

Difference being that people die if you exceed the actual capacity that an elevator can handle. An Intel chip just doesn't boost, or it throttles a bit. Plus, anyone getting an unlocked SKU with the intention of overclocking should be buying a cooler than can manage some level of overclocking. If they aren't, TDP is meaningless as they'll be lucky to see a clockspeed above 800MHz.

yes you are correct that you use a more conservative estimate when dealing with something that will kill if it breaks but you also use conservative estimates when creating products that are being used at a certain spec. the reason being that their is variation in a production so if you want more of your chips produced to meet the TDP requirement then you want the TDP to be at least a little bit conservative. 

[Dabombinable]

14 minutes ago, Drak3 said:

Not really, if it turns out that the 8700K can hit near 5GHz, if not 5GHz even, on average. The 5820K, hitting around 4.5GHz on average, water, is going to be at around a 5% IPC disadvantage, 11% clock speed disadvantage, for a roughly 16.5% disadvantage, before figuring in IMC improvements and RAM overclocking potential.

 

Of course, assuming that the 14nm++ node can enable the hexacore to reach 5GHz on average.

Quad channel DDR4+more PCIe lanes+more motherboard features.....

[mr moose]

Just now, Brooksie359 said:

ok so lets say I buy a 7700k. Chances are the 7700k won't have the exact TDP tested on the chip or chips Intel used to determine TDP. 

 

It should be within margin of error,  If you buy a 7700K and run it at base clock and moderate workload and it dissipates more than 95watts, then you have a faulty chip. and should be entitled to return it.   The problem here is people aren't talking about variation in manufacturing, they are outright claiming the TDP value is a guess or just plain  BS.  Which it isn't, they just don't understand what it is exactly and why it exists.

 

 

 

 

[Brooksie359]

Just now, mr moose said:

 

It should be within margin of error,  If you buy a 7700K and run it at base clock and moderate workload and it dissipates more than 95watts, then you have a faulty chip. and should be entitled to return it.   The problem here is people aren't talking about variation in manufacturing, they are outright claiming the TDP value is a guess or just plain  BS.  Which it isn't, they just don't understand what it is exactly and why it exists.

 

 

 

 

I mean if Intel wanted they could put a higher TDP than what is needed but obviously the converse would be a problem. I suspect that Intel puts higher TDP than what is actually necessary for some of their cpus. This may be because they test some of the skews together. 

[mr moose]

1 minute ago, Brooksie359 said:

I mean if Intel wanted they could put a higher TDP than what is needed but obviously the converse would be a problem. I suspect that Intel puts higher TDP than what is actually necessary for some of their cpus. This may be because they test some of the skews together. 

They could, but then they would be essentially telling laptop designers they need to use a bigger coolers than necessary, which would result in products being more bulky and probably not selling as well.   I am pretty sure getting as close to accurate with the usual qualifiers is standard practice. 

 

 

[Drak3]

4 minutes ago, Lays said:

And what do you use those 40 PCI-E lanes for?

As of right now, gawking, dual GPUs at x8 each and 1 NVMe. In the future, either 2x Radeons in Crossfire with a GTX x80 or x80Ti and some NVMe, or add in 4 more 950 Pro/ 960 Evos on a x16 riser. Depending on whether I buy a new mainboard or someone releases a x16 riser first.

 

Or I buy into SLX, and the 5930K becomes an overkill NAS.

 

2 minutes ago, Dabombinable said:

Quad channel DDR4+more PCIe lanes+more motherboard features.....

Quad channel is nice, but the only benefit occurs in situations where you don't have enough bandwidth or enough capacity. Neither are going to happen for many users looking at the 8700K or 5820K.

28+8(2.0) vs 16+24. Neither are driving as many PCIe devices natively as a 5930K or 7900X, and some devices outright don't care or benefit enough from CPU lanes over chipset lanes, with few consumers caring about sniping the gap between 8700K and something like the 5930K.

Mainboard features also don't really vary that much going from top end, to top end. And if you got someone looking at those features, chances are, they're looking to pair it with a chip worth tinkering with, in the respective platforms.