7900X Coming In Hot?!

[Taf the Ghost]

41 minutes ago, Jumper118 said:

i disagree, i would say anyone doing stock tests is pointless because they would all be the same. overclocking bring in a variance, therefore it makes you have to read/watch multiple reviews. Also nobody spending that much on an unlocked cpu should not overclock it, if you are that bothered about stability you would go with a cheaper xeon with the same or high number of cores. . 

OC'ing is useful information, but we have a vast supply of consumer level results saying a very small amount actually overclock their systems.

 

Testing for consumer advice & system comparisons should be done at Stock. OC testing is useful information for enthusiasts. 

[Taf the Ghost]

31 minutes ago, NuclearKing said:

  Reveal hidden contents

IPC performance

One difficult thing to discern with most of these benchmarks is just how “efficient” each CPU microarchitecture is. One way to tease that out is by running a test using a single thread, with all the CPUs locked to the same clock speed. We locked most of the CPUs here to 2.5GHz and turned off any Turbo Boost. We then ran Cinebench R15.037 (which we used to generate scores for the older CPUs).

We can see that instructions per clock (IPC) has built itself up slowly from the days of Sandy Bridge. Skylake-X comes out in front of even Kaby Lake, surprisingly. 

Note: The FX-8370 is fast because we couldn’t underclock the CPU to 2.5GHz, so we just used its score running at 4GHz. Yes, the performance of Vishera at 4GHz is still below that of a Haswell CPU running at 2.5GHz. Woof.

 

IDG

We locked down almost all of the CPUs to 2.5GHz to test the IPC. No, the AMD FX isn’t holding its own: It’s at 4GHz vs. 2.5GHz for the other CPUs.

http://pcworld.com/article/3201187/computers/intel-core-i9-review.html?page=2

 

PCworld did something like this, but seeing more benchmarks and games as well at the same clock speeds would be quite fascinating, if not practical.

 

Quote

Note: The FX-8370 is fast because we couldn’t underclock the CPU to 2.5GHz, so we just used its score running at 4GHz. Yes, the performance of Vishera at 4GHz is still below that of a Haswell CPU running at 2.5GHz. Woof.

I've included their note, as I was really confused by the test otherwise.  

 

As for their test, either 3.0 or 3.5 Ghz is probably better test level.  The downscaling with Clock Speed isn't linear. Though it looks like Cinebench does like the cache hierarchy change more than other programs.  Corona is a program that the Skylake-X saw a solid Uplift in, while IPC is either flat or actually down across the board. (Sideways Upgrade is probably the best phrase; it'll be better for the stuff Server buyers like.)

[NuclearKing]

36 minutes ago, Taf the Ghost said:

The downscaling with Clock Speed isn't linear.

I hadn't known that, I assumed that the clockspeed downscaled pretty linearly. What is the reason that it doesn't downscale linearly?

 

In this case it does seem to, at least for Intel (don't know about Ryzen). My 6600K at 4.7 gets 199 CB. 199 divided by 4.7 is 42.3 CB/1 GHz. The 6700K gets 106 CB. Divide by 2.5, and you get 42 CB/1 GHz.

[2FA]

5 minutes ago, NuclearKing said:

I hadn't known that, I assumed that the clockspeed downscaled pretty linearly. What is the reason that it doesn't downscale linearly?

 

In this case it does seem to, at least for Intel (don't know about Ryzen). My 6600K at 4.7 gets 199 CB. 199 divided by 4.7 is 42.3 CB/0.1 GHz. The 6700K gets 106 CB. Divide by 2.5, and you get 42 CB/0.1 GHz.

Not only does clockspeed affect the speed of transistor gates switching, it also affects the latency. Hence why performance doesn't scale linearly with clockspeed since it affects more than one variable. (There are probably other variables as well but that's the main one I know of)

[NuclearKing]

4 minutes ago, DeadEyePsycho said:

Not only does clockspeed affect the speed of transistor gates switching, it also affects the latency. Hence why performance doesn't scale linearly with clockspeed since it affects more than one variable. (There are probably other variables as well but that's the main one I know of)

Ok, that makes sense. But as explained in the math on the second paragraph, Cinebench doesnt seem to care. Makes me wish Gordon Mah Ung had tested some other benchmarks as well.

[Taf the Ghost]

4 minutes ago, NuclearKing said:

I hadn't known that, I assumed that the clockspeed downscaled pretty linearly. What is the reason that it doesn't downscale linearly?

 

In this case it does seem to, at least for Intel (don't know about Ryzen). My 6600K at 4.7 gets 199 CB. 199 devided by 4.7 is 42.3 CB/0.1 GHz. The 6700K gets 106 CB. Divide by 2.5, and you get 42 CB/0.1 GHz.

In the bigger picture sense, no complex system can shift in a truly linear fashion, but in the case of CPUs you have the issue of CPU's calculation capacity isn't necessarily the limiting factor for the output. It's the same reason that bigger L2 cache helps certain calculations but can hurt other operations.

 

There could also be issues with parts being designed to run at certain frequencies/latencies that can crop up.  Why I suggest IPC testing within the operating range of most CPUs, so limit the effect some.

[Jumper118]

1 hour ago, Taf the Ghost said:

OC'ing is useful information, but we have a vast supply of consumer level results saying a very small amount actually overclock their systems.

 

Testing for consumer advice & system comparisons should be done at Stock. OC testing is useful information for enthusiasts. 

i'm pretty sure most people who buy unlocked chips oc them. people who game but buy prebuilt pc's or locked chips wont be wanting this information, so its not relevant.

[Taf the Ghost]

5 minutes ago, Jumper118 said:

i'm pretty sure most people who buy unlocked chips oc them. 

All of the system surveys we have say it's a very low amount.  Now, the k SKU buyers are far more likely to, simply because they are able to, but it still appears to be a fairly small minority.  RAM OC might actually be slightly more common, since anything over baseline JEDEC counts.

[Jumper118]

Just now, Taf the Ghost said:

All of the system surveys we have say it's a very low amount.  Now, the k SKU buyers are far more likely to, simply because they are able to, but it still appears to be a fairly small minority.  RAM OC might actually be slightly more common, since anything over baseline JEDEC counts.

this is like saying 50% of gamers are women, when 99% of that 50% of them play candy crush and other moblie games. 

 

they are not the target audience for this cpu or review or video, so they dont even count  why would you watch a 7740k or 7900x review if you are not interested in a cpu like that. they are enthusiast grade parts. 

[done12many2]

55 minutes ago, Jumper118 said:

this is like saying 50% of gamers are women, when 99% of that 50% of them play candy crush and other moblie games. 

 

they are not the target audience for this cpu or review or video, so they dont even count  why would you watch a 7740k or 7900x review if you are not interested in a cpu like that. they are enthusiast grade parts. 

 

Because you can talk shit about it even if it's something that you'd never personally use anyways.  It's the LTT way to complain about shit that never mattered to you to begin with.  Look closely in the fine print of the Community Standards.  

 

90% of the people on LTT actively talking shit about Threadripper or Skylake-X will never own either.  

[PCGuy_5960]

Just now, done12many2 said:

90% of the people on LTT actively talking shit about Threadripper or Skylake-X will never own either.  

Dood, no one is talking shit about Threadripper (because it's AMD)

[Taf the Ghost]

 

7 hours ago, PCGuy_5960 said:

Dood, no one is talking shit about Threadripper (because it's AMD)

Only because it's not out for a month.  

 

8 hours ago, Jumper118 said:

this is like saying 50% of gamers are women, when 99% of that 50% of them play candy crush and other moblie games. 

 

they are not the target audience for this cpu or review or video, so they dont even count  why would you watch a 7740k or 7900x review if you are not interested in a cpu like that. they are enthusiast grade parts. 

 

For as many 7700k + mobo combos are sold, which is a huge part of the non-pre-built segment, the OC results just don't line up.  Yes, the CPUs may do a good OC (if you have the cooling), but even of the people that buy the OC-able chips it isn't the a given very many are running OCs.

 

It's like the AIO craze among reviewers.  From a practical stand-point of a reviewer, an AIO is a lot better to run with because it's a lot easier to change out and easier for their life.  In the market? The Hyper 212 EVO probably outsells all AIOs by 4:1 if not even higher, so if you're buying advice is "buy an AIO" you've mostly failed in that advice already.

 

As I said when I made the first comment, you should test OC potential because we're in an area that wants to know how far you can push your hardware and there is value in that information.  It doesn't mean CPU comparisons should be done on the basis of OC results, especially since the Silicon Lottery is real.

[Jumper118]

1 minute ago, Taf the Ghost said:

 

Only because it's not out for a month.  

 

 

For as many 7700k + mobo combos are sold, which is a huge part of the non-pre-built segment, the OC results just don't line up.  Yes, the CPUs may do a good OC (if you have the cooling), but even of the people that buy the OC-able chips it isn't the a given very many are running OCs.

 

It's like the AIO craze among reviewers.  From a practical stand-point of a reviewer, an AIO is a lot better to run with because it's a lot easier to change out and easier for their life.  In the market? The Hyper 212 EVO probably outsells all AIOs by 4:1 if not even higher, so if you're buying advice is "buy an AIO" you've mostly failed in that advice already.

 

As I said when I made the first comment, you should test OC potential because we're in an area that wants to know how far you can push your hardware and there is value in that information.  It doesn't mean CPU comparisons should be done on the basis of OC results, especially since the Silicon Lottery is real.

Where are you even getting this information from cos it sounds like you are making it up as you go along. 

[Taf the Ghost]

2 minutes ago, Jumper118 said:

Where are you even getting this information from cos it sounds like you are making it up as you go along. 

There was a Microsoft study that went around a month or two ago that the lack of OC'ing cropped up.  Steam Survey points to only a few CPUs being base clocked above 3.7 Ghz. Review totals on popular products gives you a fairly decent general overview of product sales ratios, and just little tidbits of data that are available or crops up.  Any mass-scale product is always going to be massively dominated by being used at the "normie" level. 

 

Why else is the 7700k selling better than the 6700k.

[Jumper118]

6 hours ago, Taf the Ghost said:

There was a Microsoft study that went around a month or two ago that the lack of OC'ing cropped up.  Steam Survey points to only a few CPUs being base clocked above 3.7 Ghz. Review totals on popular products gives you a fairly decent general overview of product sales ratios, and just little tidbits of data that are available or crops up.  Any mass-scale product is always going to be massively dominated by being used at the "normie" level. 

 

Why else is the 7700k selling better than the 6700k.

Microsoft make an xbox so their research does even count.  steam does detect overclocks, it only reads the stock speed of the cpu. 

[NuclearKing]

1 hour ago, Jumper118 said:

Microsoft make an xbox so their research does even count.  steam does detect overclocks, it only reads the stock speed of the cpu. 

I think you mean "doesn't" on both accounts, but I'm not sure.

[PCGuy_5960]

3 minutes ago, NuclearKing said:

I think you mean "doesn't" on both accounts, but I'm not sure.

Yep, @Jumper118 always makes typing mistakes because his keyboard is overclocked too much

[done12many2]

3 minutes ago, NuclearKing said:

I think you mean "doesn't" on both accounts, but I'm not sure.

 

Communication is not @Jumper118's strong point, but the guy knows how to overclock the shit out of a CPU.  

[NuclearKing]

Just now, done12many2 said:

 

Communication is not @Jumper118's strong point, but the guy knows how to overclock the shit out of a CPU.  

 

Just now, done12many2 said:

Communication is not @Jumper118's strong point, but the guy knows how to overclock the shit out of a CPU.  

And keyboards apparently :'D That keyboard OC doesn't seem entirely stable, might want to dial back the multiplier on the WPM

[Jumper118]

thanks for pointing that out. i did indeed mean doesn't

6 minutes ago, NuclearKing said:

I think you mean "doesn't" on both accounts, but I'm not sure.

 

[done12many2]

Just now, NuclearKing said:

 

And keyboards apparently :'D That keyboard OC doesn't seem entirely stable, might want to dial back the multiplier on the WPM

 

I think it's the LN2.  Too much is causing his finger to stick to the keys.  

[ravenshrike]

 

On 6/22/2017 at 9:45 AM, PCGuy_5960 said:

It will lose if you compare it to the Threadripper CPU that costs the same. If you compare it to the AMD CPU that has the same number of cores, it will destroy it

Which clearly isn't AMDs plan for Ryzen Gen 1 chips. Their plan is to dominate the low to mid end of mainstream, HEDT, and server market for x86 chips as their opening blow. 

 

On 6/23/2017 at 8:46 AM, PCGuy_5960 said:

What? At 4.8Ghz he was getting 90C....

 

Delidded with a TIM with at the least double the thermal conductivity of whatever he OC'd the previous processors with. 

 

On 6/23/2017 at 1:03 PM, Taf the Ghost said:

Unless you need the I/O, I'd really recommend waiting for Coffee Lake for a 6c/12t Intel CPU. It really looks like it'll be on the 2nd revision of the 14nm process.

This is just a theory with nothing behind it but my gut reading of Intel's launch strategy, but I'm pretty sure the plan for Coffee Lake sans AMD reentering the market was to remove overclocking from the mainstream chipset. That's the only thing that really explains putting an i5 on their HEDT platform. Course, even if that was the plan, the probability they go through with it given the rise of Ryzen is... unlikely.

On 6/23/2017 at 2:11 PM, Taf the Ghost said:

I outlined in another post a bit back that Intel hasn't made any incorrect moves.

 

No, they made one incorrect move which all their blunders stemmed from. They assumed they would remain King without anybody even making it to the castle walls. Unfortunately for them AMDs sappers have just lit the fuses of the mines laid underneath the castle walls. It remains to be seen just how much damage they will do.

On 6/23/2017 at 2:46 PM, done12many2 said:

 

Asus already has their manuals for some of their boards up on their website.   That's how I learned that the VROC key thing was just everyone overreacting.  

What purpose is served running an NVME raid through your DMI lanes given everything else in that channel? I mean, MAYBE you can successfully run RAID 1 without too much drop in throughput, but RAID 5 would saturate that connection so fast it's not even funny.
 

 

[XenosTech]

On 6/24/2017 at 6:03 PM, done12many2 said:

 

Because you can talk shit about it even if it's something that you'd never personally use anyways.  It's the LTT way to complain about shit that never mattered to you to begin with.  Look closely in the fine print of the Community Standards.  

 

90% of the people on LTT actively talking shit about Threadripper or Skylake-X will never own either.  

I might be able to actually afford TR over X299 funny enough and it might be the only enthusiast platform that might make it to retail stores here, but this is just a pipe dream for me

[done12many2]

9 minutes ago, ravenshrike said:

What purpose is served running an NVME raid through your DMI lanes given everything else in that channel? I mean, MAYBE you can successfully run RAID 1 without too much drop in throughput, but RAID 5 would saturate that connection so fast it's not even funny.

 

Redundancy, volume expansion, performance and so on.  

 

Agreed it will saturate the DMI.  My 950 Pro NVME RAID 0 already does so on z270.  That wasn't the point of my comment that you quoted.  

 

As it stands now, any NVMe RAID before x299 required RAID via the chipset.  All of the reviewers announced that there would be no RAID without the purchase of a VROC key.  This sent everyone into a bitch fest, when nothing actually changed.  NVMe RAID via the chipset is still intact if it's routed that way by the MB manufacturer as it was on z270.  VROC is optional and better then what was available on z270.  

 

4 minutes ago, XenosTech said:

I might be able to actually afford TR over X299 funny enough and it might be the only enthusiast platform that might make it to retail stores here, but this is just a pipe dream for me

 

I hope it does come true for you bud.  Either platform is going to be great on its own right. 

[XenosTech]

1 minute ago, done12many2 said:

I hope it does come true for you bud.  Either platform is going to be great on its own right. 

I know either will be great but I know for sure X299 will not make it to stores here, X99 nor none of the other X platforms have made it to stores unless special ordered. From what I heard from a friend who runs a pc store here they may get X399 if the price is right, I mean the 7900x is about $2500 just for the cpu alone