7800x vs 7700k, hint: the 7800x is slower(even when overclocked)

[cj09beira]

The new cache system will be a heat or completely miss because if the program can find in l2 it will be pretty fast but the l3 will many times be too small so skylake x will be slower because it has to grab from ram more times

[HKZeroFive]

Not surprised that the i7-7700K remains the number one choice for gaming. However, I do find it bizarre that Broadwell-E seems to do significantly better than Skylake-X, especially in titles where performance scales better with more CPU cores (for example, Ashes of the Singularity and Watch Dogs 2). In theory, Skylake-X should be better due to its higher clockspeed and higher IPC.

 

Yup, I'm chalking this up to the different architectures. Could be a case of mesh bus vs ring bus.

[Valentyn]

14 minutes ago, PCGuy_5960 said:

Results are probably wrong. 

 

This review:

  Hide contents

TechSpot's launch review:

  Hide contents

What I find very interesting is that both Hardware Unboxed/TechSpot and Science Studio (the only 2 channels that got those gaming results for X299) were using the AsRock Taichi board. Which makes me believe that this is a board problem, because Gamers Nexus (using the ASUS X299 Deluxe) got these results:

  Reveal hidden contents

And Tom's Hardware (using the MSI Carbon board) got these results:

  Reveal hidden contents

 

As for the IPC debate, well even if there was a difference (there's no difference tho) Skylake-X is the same or better than Kabylake clock for clock according to @done12many2's and @MageTank's testing

Looks like a Ryzen launch all over again. Funky cache, and motherboard vendors being rushed in some cases.

 

Hardware Unboxed did say he'll be retesting and including more cpus, including Ryzen.

 

hopefully on a different motherboard, that seems like a rather large performance delta for a motherboard.

[PCGuy_5960]

Just now, Valentyn said:

hopefully on a different motherboard, that seems like a rather large performance delta for a motherboard.

I really hope that they use a different motherboard, because so far the AsRock Taichi looks like a really bad board

[tom_w141]

3 minutes ago, HKZeroFive said:

I'm chalking this up to the architecture

I'm going for less L3 cache per core. Vs Kaby lake and Broadwell-E

[tom_w141]

1 minute ago, PCGuy_5960 said:

I really hope that they use a different motherboard, because so far the AsRock Taichi looks like a really bad board

Which is strange because the taichi is often good (well the X99, Z270 and X370 versions are). You shouldn't have issues regardless of motherboard though. They need to sort this shit out and I can't help but feel having until august to work on their boards (original launch plan) would have yielded much better performance out of the gates, rather than bringing it forward 2 months and dropping more SKUs on them at Computex.

[tom_w141]

40 minutes ago, jjohnthedon1 said:

Do they have 2 moduals with cores on like ryzen ? Could be latency between cores ? 

No

[HKZeroFive]

Just now, tom_w141 said:

I'm going for less L3 cache per core.

It could be either the reworked L2/L3 cache design or the new mesh bus as opposed to the old ring bus of Broadwell-E. But it is rather clear that Intel botched Skylake-X somehow.

[PCGuy_5960]

Just now, HKZeroFive said:

It could be either the reworked L2/L3 cache design or the new mesh bus as opposed to the old ring bus of Broadwell-E.

Doubt it, something is wrong with TechSpot's testing. Gamers Nexus got these results:

Spoiler

 

[Valentyn]

16 minutes ago, PCGuy_5960 said:

I really hope that they use a different motherboard, because so far the AsRock Taichi looks like a really bad board

 

12 minutes ago, tom_w141 said:

Which is strange because the taichi is often good (well the X99, Z270 and X370 versions are). You shouldn't have issues regardless of motherboard though. They need to sort this shit out and I can't help but feel having until august to work on their boards (original launch plan) would have yielded much better performance out of the gates, rather than bringing it forward 2 months and dropping more SKUs on them at Computex.

Was just about to say that, the Taichi has been one of the best performance boards around. Interesting that they're having issues here. Could be a BIOS issue or there's something else. 

 

[Valentyn]

1 minute ago, PCGuy_5960 said:

Doubt it, something is wrong with TechSpot's testing. Gamers Nexus got these results:

  Reveal hidden contents

 

Actually looking at those I don't see the 6800x there at all.

 

Also much smaller suite of games.

 

Need more tests in genera though.

Nice motherboard round up testing round as well, like they also did with Ryzen.

[PCGuy_5960]

4 minutes ago, Valentyn said:

Actually looking at those I don't see the 6800x there at all.

Well, the 7900X shouldn't perform better than the 7800X in games, I mean, most games can't use more than 8 threads.

 

And Salazar used the same board with the 7900X and got the same results as HUB (20FPS worse than the 6700K/7700K) so this is why I believe that the Taichi is the issue

[jjohnthedon1]

22 minutes ago, tom_w141 said:

No

Cool story bro

[Valentyn]

1 minute ago, PCGuy_5960 said:

Well, the 7900X shouldn't perform better than the 7800X in games, I mean, most games can't use more than 8 threads.

 

And Salazar used the same board with the 7900X and got the same results as HUB (20FPS worse than the 6700K/7700K) so this is why I believe that the Taichi is the issue

Depends on the games also, and nexus has a tiny gaming suite to test with. Some games also like lots of cache and the 7900x has loads compared to 7800x.

 

Could be motherboard and more. We have so little tests of the 7800x.

[PCGuy_5960]

2 minutes ago, Valentyn said:

Depends on the games also, and nexus has a tiny gaming suite to test with. Some games also like lots of cache and the 7900x has loads compared to 7800x.

This is why I used games that both reviews had:

 

I doubt that the cache size makes a huge difference and the 7900X has higher latency (because it has more cores)

[HKZeroFive]

13 minutes ago, PCGuy_5960 said:

Doubt it, something is wrong with TechSpot's testing. Gamers Nexus got these results:

  Reveal hidden contents

 

Can't be a coincidence when ComputerBase concluded the same thing: https://www.computerbase.de/2017-07/core-i7-7740x-7800x-i9-7900x-spiele-benchmark/

 

Intel even came out with a vague PR statement:

 

"While these compromises have a slightly negative impact on a handful of applications, the Skylake X processors deliver an excellent performance per core and significant performance gains across a variety of applications, thanks to the new architecture."

[Vode]

39 minutes ago, HKZeroFive said:

Yup, I'm chalking this up to the different architectures. Could be a case of mesh bus vs ring bus.

According to Anandtech the added latency for the L3 mesh has over the ring bus is only 10% at most. Which translates to 19.5ns instead of 18ns...

 

Meh I don' know but I have a feeling the lack of sheer L3 cache size (compared to Broadwell-E) is causing those performance discrepancies. It is almost half the size afterall.

[Valentyn]

15 minutes ago, PCGuy_5960 said:

This is why I used games that both reviews had:

 

I doubt that the cache size makes a huge difference and the 7900X has higher latency (because it has more cores)

Let's not forget the cache structure. Also I wish some reviewers would rank Processors based on minimums FPS instead of average

Bit-tech in the UK usually do some great motherboard performance roundups, and tests. Sadly nothing including the x299 Taichi yet.

Do keep an eye out there, as it's surprising how cheaper motherboards can often net better performance overall than the big brothers.

https://www.bit-tech.net/reviews/tech/motherboards/

 

 

[Drak3]

1 hour ago, Vode said:

is only 10% at most. Which translates to 19.5ns instead of 18ns...

Wouldn't that be 19.8 vs 18 though?

[Taf the Ghost]

Wait for Coffee Lake if you want the "best" gaming CPU. 

 

There's already a few games that can use 6c better than 4c. TechDeals has talked about favoring his Ryzen 7 system over his 7700k system because of this. It's why Coffee Lake on 6c with Kaby Lake speeds & ring bus should be the best gaming CPU by a chunk for a while.

[Taf the Ghost]

1 hour ago, Vode said:

According to Anandtech the added latency for the L3 mesh has over the ring bus is only 10% at most. Which translates to 19.5ns instead of 18ns...

 

Meh I don' know but I have a feeling the lack of sheer L3 cache size (compared to Broadwell-E) is causing those performance discrepancies. It is almost half the size afterall.

There's also the different Chipset, and as testing with the 7740X has shown, that matters for a lot of games. Enough that the 7740X is worse than the 7700k.

[Vode]

13 minutes ago, Drak3 said:

Wouldn't that be 19.8 vs 18 though?

Oh yes sure! Haha 

 

I read 19.5ns on the comparison chart. The 10% was the ballpark percentage.

[mynameisjuan]

2 less FPS on average at over 100+ FPS and the world loses its shit.

 

Cant we all just be happy that the HEDT are matching consumer i7s for gaming? 

[tom_w141]

1 hour ago, Valentyn said:

Let's not forget the cache structure. Also I wish some reviewers would rank Processors based on minimums FPS instead of average

Min FPS is a horrible metric as spikes can happen. 0.1%, 1% and avg are the best ways to display fps. But yes I focus on 0.1 and 1 more when comparing because a high avg is a high avg doesn't matter if its 100 or 140 you won't notice. What you will notice is shitty 0.1 and 1%s.

[tom_w141]

4 minutes ago, mynameisjuan said:

Cant we all just be happy that the HEDT are matching consumer i7s for gaming?

The reasoning is for the money spent they should be better not slightly worse. Enthusiasts don't want to be punished for being on the enthusiast platform.