Core i7 7700k Confirmed Benchmarks- Zero IPC gains over Skylake

[Zodiark1593]

4 hours ago, imreloadin said:

I wonder how many of the Intel supporters ragged on AMD when they released the FX-9590? This is essentially the same thing, you know minus the millions of dollars spent in R&D that went into the "optimizing" of Kaby Lake...

Optimizing probably has little to do with performance. Fixing errata, tweaking the manufacturing process, improving yields of high clocked chips, etc. A Skylake Refresh would be an apt name. 

[done12many2]

Just got the z270 / 7700k up a running early evening yesterday, but initial impressions are very good.  I ran out of good thermal paste and was forced to use some EK Ecotherm that I found in a few of my water block boxes.    Unfortunately, nobody local stocks the stuff, so I ordered some Grizzly Kryonaut and Gelid Extreme and they're on the way. 

 

There is NO need to run AVX offset in order to hit 5 GHz or higher.  There's also no need to delid for 5 GHz.  According to Silicon Lottery, "As of 1/05/17, the top 62% of tested 7700Ks were able to hit 5.0GHz or greater."  My experience so far tells me that this is seems reasonable.  

 

5 GHz was literally as simple as setting XMP, changing the CPU multiplier and booting.  For the first run, I left voltage on Auto just to see what would happen.  It ran fine, but voltage was 1.4v + so I dialed it back for testing.  It can run 5.2 GHz, but temps are a bit too high for my tastes when I run RealBench, so I kept testing at 5 GHz and 5.1 GHz until better TIM arrives.

 

Anyways, here are some initial RealBench runs.  Like I said, temps will improve once I get a respectable TIM on the chip.  I bought two 7700k CPUs and once I'm done running through the first one, I'll throw the second one in.

 

 

 

 

[MageTank]

On 1/7/2017 at 11:39 AM, done12many2 said:

Just got the z270 / 7700k up a running early evening yesterday, but initial impressions are very good.  I ran out of good thermal paste and was forced to use some EK Ecotherm that I found in a few of my water block boxes.    Unfortunately, nobody local stocks the stuff, so I ordered some Grizzly Kryonaut and Gelid Extreme and they're on the way. 

 

There is NO need to run AVX offset in order to hit 5 GHz or higher.  There's also no need to delid for 5 GHz.  According to Silicon Lottery, "As of 1/05/17, the top 62% of tested 7700Ks were able to hit 5.0GHz or greater."  My experience so far tells me that this is seems reasonable.  

 

5 GHz was literally as simple as setting XMP, changing the CPU multiplier and booting.  For the first run, I left voltage on Auto just to see what would happen.  It ran fine, but voltage was 1.4v + so I dialed it back for testing.  It can run 5.2 GHz, but temps are a bit too high for my tastes when I run RealBench, so I kept testing at 5 GHz and 5.1 GHz until better TIM arrives.

 

Anyways, here are some initial RealBench runs.  Like I said, temps will improve once I get a respectable TIM on the chip.  I bought two 7700k CPUs and once I'm done running through the first one, I'll throw the second one in.

 

 

 

 

 

 

 

 

You cannot say the AVX offset was not required if you don't run AVX intensive stress tests though. Run XTU(Linpack), Prime95 (with AVX/FMA3), or Aida64's FPU test and see if that holds stable. I understand realbench's h.264 encoder uses AVX (at least, last I checked it did) but it's nowhere near as stressful as the others. The fact that it's capped at 16gb of ram makes it weak for my specific needs (need at least 24GB to test 75% of my capacity)

 

When using prime95, low FFT's (8-48k) will suffice for temperatures, while 1344k-2688k shows vcore stability. Linpack, it's heat scales with the amount of ram you throw at it. The more ram, the hotter it gets. I forget which size XTU defaults to, but if you want more control, just download a compiled linpack version of your choice and give it a go. 

 

Again, non-AVX overclocks being higher than AVX-stable overclocks does not surprise me (or anyone for that matter). I've seen plenty of 5ghz overclocks that work perfectly for gaming, but fails under more stressful workloads. Now, it doesn't mean it's not impressive to see high 5ghz yields from Kaby, it's just not as impressive if it requires neutering AVX performance to achieve it. We can do that already with Skylake. 

[done12many2]

1 hour ago, MageTank said:

You cannot say the AVX offset was not required if you don't run AVX intensive stress tests though. Run XTU(Linpack), Prime95 (with AVX/FMA3), or Aida64's FPU test and see if that holds stable. I understand realbench's h.264 encoder uses AVX (at least, last I checked it did) but it's nowhere near as stressful as the others. The fact that it's capped at 16gb of ram makes it weak for my specific needs (need at least 24GB to test 75% of my capacity)

 

When using prime95, low FFT's (8-48k) will suffice for temperatures, while 1344k-2688k shows vcore stability. Linpack, it's heat scales with the amount of ram you throw at it. The more ram, the hotter it gets. I forget which size XTU defaults to, but if you want more control, just download a compiled linpack version of your choice and give it a go. 

 

Again, non-AVX overclocks being higher than AVX-stable overclocks does not surprise me (or anyone for that matter). I've seen plenty of 5ghz overclocks that work perfectly for gaming, but fails under more stressful workloads. Now, it doesn't mean it's not impressive to see high 5ghz yields from Kaby, it's just not as impressive if it requires neutering AVX performance to achieve it. We can do that already with Skylake. 

 

You're correct that RealBench does use AVX and does so less aggressively, but you are mistaken on the RAM limitation.  You can use up to 64 GB with RealBench.  With the 64 GB on my x99/5960x rig, the most I've seen it actually use out of the 64 GB selected was in the area of 40%.

 

I completely understand your argument with regards to AVX, but what I'm trying to demonstrate is that given the same expected level of stability, averages for 7700k overclocks are going to be quite a bit higher than averages for 6700k overclocks.  AVX to AVX or non-AVX to non-AVX.  It just so happens that the majority of folks in this community use the non-AVX standard.  With that said, 5 GHz is a VERY possible average for 7700k CPUs.

 

Anyways, I've got a few more to test.  

 

 

@Lays

[MageTank]

12 minutes ago, done12many2 said:

 

You're correct that RealBench does use AVX and does so less aggressively, but you are mistaken on the RAM limitation.  You can use up to 64 GB with RealBench.  With the 64 GB on my x99/5960x rig, the most I've seen it actually use out of the 64 GB selected was in the area of 40%.

 

I completely understand your argument with regards to AVX, but what I'm trying to demonstrate is that given the same expected level of stability, averages for 7700k overclocks are going to be quite a bit higher than averages for 6700k overclocks.  AVX to AVX or non-AVX to non-AVX.  It just so happens that the majority of folks in this community use the non-AVX standard.  With that said, 5 GHz is a VERY possible average for 7700k CPUs.

 

Anyways, I've got a few more to test.  

 

 

 

If you'd be willing to adopt me, i'll help you test those, lol. 

[done12many2]

11 minutes ago, MageTank said:

If you'd be willing to adopt me, i'll help you test those, lol. 

 

Do you mow grass and do dishes?  If so, I'll have a lawyer start on the papers immediately.  

 

I'm binning them for myself and another member who shall remain unidentified (famous brand of potato chips ).   We ordered the der8auer Delid-Die-Mate so I should be scalping whichever proves best with the IHS intact.

[MageTank]

12 minutes ago, done12many2 said:

 

Do you mow grass and do dishes?  If so, I'll have a lawyer start on the papers immediately.  

 

I'm binning them for myself and another member who shall remain unidentified (famous brand of potato chips ).   We ordered the der8auer Delid-Die-Mate so I should be scalping whichever proves best with the IHS intact.

Grass? Yes, but I have not done dishes since I worked at a pizzeria. I actually assembled mowers and tractors for a living, so cutting grass came with the job. 

 

As for he who shall not be named (Mr Frito's), bin their IMC's, and give him the worst one out of the group. Can't have him exceeding my memory speeds by an even larger amount, or else i'll be sad. 

[VerticalDiscussions]

1 hour ago, done12many2 said:

O-o

[done12many2]

 

19 minutes ago, MageTank said:

Grass? Yes, but I have not done dishes since I worked at a pizzeria. I actually assembled mowers and tractors for a living, so cutting grass came with the job. 

 

As for he who shall not be named (Mr Frito's), bin their IMC's, and give him the worst one out of the group. Can't have him exceeding my memory speeds by an even larger amount, or else i'll be sad. 

 

Too funny on both accounts.  One shitty IMC coming up!

 

9 minutes ago, VerticalDiscussions said:

O-o

 

Weird.  I got the same reaction from my wife when she saw them, but I don't think it meant the same thing.  

[VerticalDiscussions]

25 minutes ago, done12many2 said:

-snip-

(づ◔ ͜ʖ◔)づ (づ◔ ͜ʖ◔)づ, you can do it!