7900X reviewed!

[MageTank]

2 minutes ago, PCGuy_5960 said:

LOL, yeah  But almost everyone is quick to say "Threadripper will be better" without knowing that there are a lot of instances in which this is simply not true...

BTW, in case you were interested,the article has memory benchmarks

I saw, I just wasn't impressed with it. Sure, it's bandwidth is amazing (and with AVX-512, it needs to be) but it's latency is still untamed. It's unlikely that the reviewers spent time tweaking it, but 74ns is simply out of the question for me. Especially since my latency is literally half that. If I didn't have a ton of money wrapped up in other platforms to test, I'd love to get my hands on a X299 platform just to OC it's ram as well. It seems my memory guide will never be completed, given all the rapid platform releases we've seen recently, lol.

[done12many2]

2 minutes ago, knightslugger said:

bought into the euro-diesel limited. just tack on another $4500 to that loan! still though, wow what a "truck". Find new stuff in it every week! found a button on the paddle shifter that does some weirdness with the radio. a button on a button. "...so we heard you like buttons..." --FCA.

 

You've got to love features bud.  Newer vehicle are awesome, but I've given up on spending too much money there.  I've learned over the years that I'm a drive from point A to point B type guy now.  Congrats though bud!

 

3 minutes ago, MageTank said:

@PCGuy_5960 Friend, I think you've reached the "how many times one can say AVX in a single thread" quota. Dial it back some. AVX, as good as it is, is not widely adopted, and likely won't be any time soon. Even with the extremely efficient AVX-512 on these upcoming chips, I can't imagine many people will adopt it due to the amount of heat it will add when used. Consumers are afraid of hot things, and anything that shakes their "pseudostable" overclocks stability. As I stated before, you have to be clear which AVX you speak of when comparing Ryzen and Intel*. Their AVX2 performance (after tweaked tertiary timings, specifically tRDWR and tWRRD) should be relatively close. AVX1 and AVX3 (512) should absolutely hammer Ryzen, but as I said before, the use cases outside of niche applications are few and far in between.

 

So more or less, with some work, you may be able to get Zen AVX2 close?  

 

Agreed that AVX and AVX-512 performance aren't even in the same ball park when comparing Ryzen and Skylake-X.  AVX-512 for more obvious reasons.  

 

I personally don't see anything wrong with saying that AVX performance is strategically crippled on current gen AMD chips.  

 

3 minutes ago, MageTank said:

I am interested to see why the 10c Skylake X chip is smashing the 10c Broadwell-E chip, but we will find out once it's released. My money is on the cache improvements that Intel announced, but it's hard to be certain at the moment.

 

Looking forward to learning more about this as well.  

 

 

[PCGuy_5960]

1 minute ago, MageTank said:

but 74ns is simply out of the question for me.

They probably didn't spend time tweaking it, this is the first review that has come out

[Princess Luna]

I wish I could justify buying one, gotta still wait to see what Intel does on Canon Lake mainstream CPUs

[MageTank]

Just now, done12many2 said:

 

You've got to love features bud.  Newer vehicle are awesome, but I've given up on spending too much money there.  I've learned over the years that I'm a drive from point A to point B type guy now.  Congrats though bud!

 

 

So more or less, with some work, you may be able to get Zen AVX2 close?  

 

Agreed that AVX and AVX-512 performance aren't even in the same ball park when comparing Ryzen and Skylake-X.  AVX-512 for more obvious reasons.  

 

I personally don't see anything wrong with saying that AVX performance is strategically crippled on current gen AMD chips.  

 

 

Looking forward to learning more about this as well.  

 

 

It's what I've been working on at the moment. I must say, the inverse ratio of these tertiary timings have thrown me off. It goes against everything I've learned, but I'm starting to pick up on it. Even discovered a possible formula for tRFC for Ryzen that might be worth sharing once I've 100% confirmed it. So far, it's working. 

 

I will admit, AVX performance on Ryzen seems to have been an afterthought, and for those that use AVX, it will certainly be an issue, but for your average consumer, it's almost a non-factor at the moment. Unless we see a huge shift in AVX adoption, it's likely going to be swept under the rug.

 

2 minutes ago, PCGuy_5960 said:

They probably didn't spend time tweaking it, this is the first review that has come out

I'll probably wait until some of the guys from hwbot get their hands on it. While their results are often unrealistic for 24/7 usage, they are always reliable when it comes to knowing the absolute limitations of the IMC itself. By knowing the ceiling, we know what can be obtained, and it sets goals for those of us that want something better for real-world overclocks. If they can tame it's latency (something they really couldn't do with Haswell-E) then I'll have hopes that I can. 

[SpaceGhostC2C]

1 hour ago, Sauron said:

1000$? Threadripper will offer 16 cores for less than that. Sure, in single threaded workloads it won't be nearly as good but... who cares? You aren't going to buy this for games anyway. 

I agree with you, but somehow Dell seems to think you are... 

[PCGuy_5960]

4 minutes ago, ONOTech said:

Another review is live. They bought the CPU themself (ie not a review sample), meaning that this review could be closer to real world benchmarks.

https://www.bit-tech.net/hardware/2017/06/16/intel-core-i9-7900x-and-x299-chipset-revie/1

Their results are a bit weird... I mean, the 4.4GHz 6950X beats the 4.6GHz 7900X in some tests

[PCGuy_5960]

Just now, ONOTech said:

My first guess is the cache - the 6950X has nearly double.

But the 7900X has quadruple (L2) cache...

[done12many2]

9 minutes ago, ONOTech said:

Another review is live. They bought the CPU themself (ie not a review sample), meaning that this review could be closer to real world benchmarks.

https://www.bit-tech.net/hardware/2017/06/16/intel-core-i9-7900x-and-x299-chipset-revie/1

 

They actually stated at the opening that they didn't buy it themselves.  It's not available for purchase anyways, so that wouldn't make sense. 

 

Quote

Please note that we have sourced multiple CPUs from multiple sources for the purpose of this review. No hardware was supplied by Intel.

 

[done12many2]

5 minutes ago, PCGuy_5960 said:

But the 7900X has quadruple (L2) cache...

 

I like the following quote from that article.

 

Quote

The fact that you can get a 10-core CPU to 4.6GHz with relative ease at less than 1.25V with a decent cooler is remarkable, and we have no doubt that it's the lack of solder between the heatspreader and core that's holding things back. Thankfully, the Core i9-7900X still represents a big leap in performance in many real-world tests, especially video transcoding and rendering both at stock speed and when overclocked despite thermals limiting overclocking.

 

[MageTank]

6 minutes ago, PCGuy_5960 said:

Their results are a bit weird... I mean, the 4.4GHz 6950X beats the 4.6GHz 7900X in some tests

A bit weird in what regard? We only have one other review to compare it to, lol. If more reviews come out showing vastly different results, then maybe I'd see it, but for now, I'd take both reviews with a fair bit of skepticism. This falls under the "get multiple sources before forming a final opinion" category. I am personally holding off until I see some GamersNexus/DigitalFoundry reviews, as I find them to be a bit more thorough, but that's my own opinion.

[Spectrez]

Quote

However, temperatures were definitely a concern with Cinebench and Terragen pushing 100°C with our 240mm AIO liquid cooler. As a result, while stable and potentially tameable under custom water-cooling, we decided to go for 4.6GHz for benchmarking, which required a super-low 1.22V. Interestingly our Core i7-6950X ran much cooler despite using a significantly higher voltage, albeit at 4.4GHz. This could well be due to thermal paste having been used between the heatspreader and CPU core with the new Skylake-X CPUs, in which case delidding could potentially yield significant benefits given the high heat density.

Oh i wish i had the money and the balls to delid this kind of product...

 

but for my daily programming use its so overkill its beyond me

[SpaceGhostC2C]

6 minutes ago, PCGuy_5960 said:

Their results are a bit weird... I mean, the 4.4GHz 6950X beats the 4.6GHz 7900X in some tests

Yes. Whatever one may say about the 7900X, it's hard to think of an explanation for not being better than the 6950X at everything.

 

Except...

5 minutes ago, ONOTech said:

My first guess is the cache - the 6950X has nearly double.

3 minutes ago, PCGuy_5960 said:

But the 7900X has quadruple (L2) cache...

That should be better, until you run out of all forms of cache and have to resort to RAM. Then it can still be at a disadvantage...

It's an hypothesis that would require true testing, don't think benchmarks will give us a definitive answer.

[PCGuy_5960]

4 minutes ago, MageTank said:

A bit weird in what regard?

Well, the 6950X should not be able to beat the 7900X

5 minutes ago, done12many2 said:

I like the following quote from that article.

Yep, the 7820X and the 7800X will be beasts!

[PCGuy_5960]

1 minute ago, ONOTech said:

I'm sure the extra L3 cache + early BIOS revisions could effect performance.

Yeah, the situation could be similar to Ryzen (early BIOS revisions perform much worse than newer ones)

[AnonymousGuy]

I wonder if the hammer vise method of delidding is coming back.  Someone find out whether it has pre-Skylake substrate thickness.

[PCGuy_5960]

Just now, AnonymousGuy said:

I wonder if the hammer vise method of delidding is coming back.  Someone find out whether it has pre-Skylake substrate thickness.

Nope, the substrate is thin AF (same as Kabylake) and it has a second substrate and you can't use the Vice method if you don't want to fuck up the CPU

[Spectrez]

Just now, PCGuy_5960 said:

Nope, the substrate is thin AF (same as Kabylake) and it has a second substrate and you can't use the Vice method if you don't want to fuck up the CPU

That glue seems super close to the surface components aswell... 

[done12many2]

1 minute ago, AnonymousGuy said:

I wonder if the hammer vise method of delidding is coming back.  Someone find out whether it has pre-Skylake substrate thickness.

 

I'm not sure that we're going to want to pursue that method of delidding with these chips.  Der8bauer was quick to point out how busy the PCB was surrounding the die.  The area adjacent the IHS border was just filled with shit waiting to get broken.  He further advise that delidding, even with his soon to be released tool is still risky.  

[Mr_Troll]

 

17 minutes ago, ONOTech said:

Another review is live. They bought the CPU themself (ie not a review sample), meaning that this review could be closer to real world benchmarks.

https://www.bit-tech.net/hardware/2017/06/16/intel-core-i9-7900x-and-x299-chipset-revie/1

 

so cpu benchmarks  are as expected with 2 more cores + higher clock speed... but i wouldnt call it .... out of this world.hell it geats beaten by 6950x in some tests.if ryzen 7 had a 10 core part it would be that far off....the gaming performance didnt blow me away either....he power consumption is quite high for my liking at 378 W.....  and not a single mention of temps.... must be runing quite hot ... as the overclocker der8auer showed.... overall... its an ok cpu... but not at that price...

[DildorTheDecent]

11 minutes ago, Spectrez said:

Oh i wish i had the money and the balls to delid this kind of product...

Delids aren't even hard since there's n00b tools out these days. 

[PCGuy_5960]

Just now, done12many2 said:

I'm not sure that we're going to want to pursue that method of delidding with these chips.  Der8bauer was quick to point out how busy the PCB was surrounding the die.  The area adjacent the IHS border was just filled with shit waiting to get broken.  He further advise that delidding, even with his soon to be released tool is still risky.  

Any idea as to why they are using 2 substrates?

[done12many2]

1 minute ago, Mr_Troll said:

so cpu benchmarks  are as expected with 2 more cores + higher clock speed... but i wouldnt call it .... out of this world.hell it geats beaten by 6950x in some tests.if ryzen 7 had a 10 core part it would be that far off....the gaming performance didnt blow me away either....he power consumption is quite high for my liking at 378 W.....  and not a single mention of temps.... must be runing quite hot ... as the overclocker der8auer showed.... overall... its an ok cpu... but not at that price...

 

Great username.  

[PCGuy_5960]

Just now, DildorTheDecent said:

Delids aren't even hard since there's n00b tools out these days. 

Nope, not with Skylake-X, der8auer said it himself, delidding Skylake-X is too risky even with a delid tool

[PCGuy_5960]

1 minute ago, done12many2 said:

Great username.  

And accurate