7900X Coming In Hot?!

[ZeusXI]

On 6/21/2017 at 3:06 PM, tom_w141 said:

Well damn no wonder they have a whole separate video planned on thermals and power upcoming! These are just key notes and I look forward to seeing GN's dedicated video on the topic as these guys are super thorough and have excellent test methodology. 95 degrees at a voltage as low as 1.275V?! Damn! what's your thoughts on that guys? Unlucky to anyone who loses the silicon lottery and needs to push 1.3V or more for their OC. It's not the end of the world you can rectify it by delidding and doing the job Intel should have done in the 1st place, I guess the thing to take away from this for me personally is why did they have to cut corners with the enthusiast platform like lack of solder that they didn't cut before?

 

 

My thoughts? Id either get the 7700k, or another ryzen cpu. sticking away from anything over 125W TDP. those things are ovens. Ive seen a video, think it was paulshardware or something, but even with a liquid cooler, it was still hot as all get out. Not worth it as well since cannon lake is it? coming out in the next few months

[Taf the Ghost]

https://www.overclock3d.net/reviews/cpu_mainboard/intel_i9_7900x_skylake-x_review/16

 

 

I don't think the OC3D review of the 7900X got mentioned.  The really interesting part is that the 8c & 10c are showing the same power draw, which suggests that the Heat issue might actually be outright holding the CPU back.

 

Quote

It is certainly a warm processor though, as you would expect from something with this much silicon under the cover, so a good cooling solution is a bare minimum. We'd certainly not want to cool this with anything smaller than a 240mm AIO if youre planninjg on overclocking and 1.2v is probably going to be a ceiling even with a cooler like this if you like a quiet system. Delis are a viable option to reduuce a speculated 20c-30c off of your load temps but this comes with a great deal of risk and totally voiding your warranty. However, if you think ten cores at 4 GHz will be enough performance for you - and unless you're Pixar it probably is - then the 7900X also rewards undervolting. We managed to get it down to 1v VCore with the temperatures dropping to just a hair over 50°C. Excellent.

 

[tom_w141]

23 minutes ago, Taf the Ghost said:

I don't think the OC3D review of the 7900X got mentioned.  The really interesting part is that the 8c & 10c are showing the same power draw, which suggests that the Heat issue might actually be outright holding the CPU back.

Got mentioned on the 1st few pages but ignored due to its inconvenience/reluctance to admit the more popular 8 core will also get hot Monday is only 2 days away so they will soon see, as long as they are cool (not literally ) with it then that's fine  The entire point of the thread was to warn people not comfortable running near to 100 degrees that it might not be for them  Of course you are fine if you don't overclock but if you don't overclock the gains over the Ryzen costing between half - a third less (depending on whether you compare CPU or Platform) are quit insignificant. 

 

Steve @ Techspot/HWUnboxed:

 

Quote

Let's talk 8-core CPUs for a minute. The Ryzen 7 1800X costs $460, but smart shoppers will opt for the much cheaper $310 1700, which can be overclocked to achieve the same performance. The Core i7-7820X comes in at $600, making it 30% more expensive than the 1800X, but almost twice the price of the 1700. However looking at the real-world applications, the 7820X was on average just 10% faster.

 

tldr of this whole thread basically:  Good silicon, knee capped by poor thermal efficiency between die and IHS. Resulting in crazy performance that can't be harnessed without a custom loop and a delid.

 

 

[tom_w141]

2 hours ago, done12many2 said:

 

It's good to see that nothing gets by you.  

Just realised if you dedicate your X299 build to me then you are calling me hot  thanks x

[PCGuy_5960]

15 minutes ago, tom_w141 said:

 However looking at the real-world applications, the 7820X was on average just 10% faster.

Unless those applications use (you guessed it) AVX!

[tom_w141]

Just now, PCGuy_5960 said:

Unless those applications use (you guessed it) AVX!

If it spikes to high 90s in non AVX workloads like Aida64 CPU only, then I can only imagine the amount of throttling when AVX is actually utilised  

 

Also you are quoting Steve there not me  

[PCGuy_5960]

3 minutes ago, tom_w141 said:

If it spikes to high 90s in non AVX workloads like Aida64 CPU only, then I can only imagine the amount of throttling when AVX is actually utilised  

Go big or go home, get a custom loop and overclock or don't overclock  The 7900X performs like a 4.1-4.2GHz 6950X at stock, so its pretty fast even if you don't overclock

3 minutes ago, tom_w141 said:

Also you are quoting Steve there not me  

I know

[Taf the Ghost]

52 minutes ago, tom_w141 said:

Got mentioned on the 1st few pages but ignored due to its inconvenience/reluctance to admit the more popular 8 core will also get hot Monday is only 2 days away so they will soon see, as long as they are cool (not literally ) with it then that's fine  The entire point of the thread was to warn people not comfortable running near to 100 degrees that it might not be for them  Of course you are fine if you don't overclock but if you don't overclock the gains over the Ryzen costing between half - a third less (depending on whether you compare CPU or Platform) are quit insignificant. 

 

Steve @ Techspot/HWUnboxed:

 

 

tldr of this whole thread basically:  Good silicon, knee capped by poor thermal efficiency between die and IHS. Resulting in crazy performance that can't be harnessed without a custom loop and a delid.

 

 

OC3D did a 7820X (8c) on launch, the 7900X (10c) review came out Thursday at some point. So it's a later review, with roughly the same thermal issues, and it is another 7900X that clocks at 4.6 Ghz.

[Kinda Bottlenecked]

2 hours ago, done12many2 said:

 

27 pages later we know that running a 10c/20t CPU at high clock speeds with an off the shelf cooler is going to result in high temps.  To that I say, no shit.

 

yea u need a "custom loop with a thicc rad"

[Valentyn]

59 minutes ago, Pohernori said:

 

yea u need a "custom loop with a thicc rad"

Extra T H I C C!
 

 

[done12many2]

I'm just happy the return on that power consumed is there!

 

We all tend to look at things from different angles.  Keeping things in perspective won't net you the dramatic effect that some are going for, but it is fair thing to do.

 

 

 

 

 

 

[XenosTech]

8 minutes ago, done12many2 said:

I'm just happy the return on that power consumed is there!

 

We all tend to look at things from different angles.

Yeah but the 7820x isn't much better than 1800 or 1700 but 2 extra cores and better ipc on the 7900x will obviously give it an advantage

[done12many2]

1 minute ago, XenosTech said:

Yeah but the 7820x isn't much better than 1800 or 1700 but 2 extra cores and better ipc on the 7900x will obviously give it an advantage

 

Agreed.  There's something going on with the 7820X with regards to power draw at this moment in time, but I'm sure that it will all get worked out.  

 

The 10 core 7900X is actually giving us more performance then expected per extra watt consumed.  

[Deli]

18 minutes ago, done12many2 said:

 

Agreed.  There's something going on with the 7820X with regards to power draw at this moment in time, but I'm sure that it will all get worked out.  

 

The 10 core 7900X is actually giving us more performance then expected per extra watt consumed.  

I think wait for the 10 core TR then compares to 7900X will be more apple to apple comparison. Although I expect 7900X would still beat the 10 core TR hand down in term of performance.

 

Honestly, I don't care too much about power consumption. As long as the performance is there, and not hard to keep it cool. The second part is a minor headache at this moment.

[XenosTech]

15 minutes ago, done12many2 said:

 

Agreed.  There's something going on with the 7820X with regards to power draw at this moment in time, but I'm sure that it will all get worked out.  

 

The 10 core 7900X is actually giving us more performance then expected per extra watt consumed.  

Compared to the 6950x it's power draw to performance seems odd... 87 more watts for like 204 extra points in cinebench

[Jumper118]

41 minutes ago, done12many2 said:

I'm just happy the return on that power consumed is there!

 

We all tend to look at things from different angles.  Keeping things in perspective won't net you the dramatic effect that some are going for, but it is fair thing to do.

 

that is the whole system not just the cpu power draw. the x299 board will have way more features and prts so it will also use more power than x370. another thing to consider is that the ryzen performance scores are very out of date and performance of those cpus have been greatly increased whereas intel cpu's dont improve over time as they are already well optimised. also also the 1700 oc used less power and got more or less the same performance as the 1800x while being cheaper and also the 7900K will use less power once delided, and use more power and oc higher and can run higher voltage making it faster and more efficient, but draw even more more powers. 

 

 

 

The point is that those grarfs are not really very helpful as some of the data is out of date and some is not fully optimised. 

[done12many2]

7 minutes ago, Jumper118 said:

that is the whole system not just the cpu power draw. the x299 board will have way more features and prts so it will also use more power than x370. another thing to consider is that the ryzen performance scores are very out of date and performance of those cpus have been greatly increased whereas intel cpu's dont improve over time as they are already well optimised. also also the 1700 oc used less power and got more or less the same performance as the 1800x while being cheaper and also the 7900K will use less power once delided, and use more power and oc higher and can run higher voltage making it faster and more efficient, but draw even more more powers. 

 

 

 

The point is that those grarfs are not really very helpful as some of the data is out of date and some is not fully optimised. 

 

Yeah, I've seen Ryzen chips hit 1800cb and I doubt that 2500cb is the highest we'll see from the 7900X. 

 

I was just trying to put all the talk of excessive power consumption into perspective. 

 

In the end, for every what drawn by the 7900X / x299 platform, you get a decent amount of extra performance in return. 

[Taf the Ghost]

The "System Power Draw" would be a Delta not an absolute.  The i3-7350k to the i7-7700k at stock is probably the best direct guess for "watts per core".  Delta there is 49w for doubling the cores. Puts Mobo + GPU at around 60w in this test. (Though it's going to be slightly different for every Motherboard/Chipset.)

 

So it's probably around a 45-50% Power Draw increase for around a 17% increase in the Cinebench R15 Multi-core result.

[Taf the Ghost]

8 minutes ago, done12many2 said:

 

Yeah, I've seen Ryzen chips hit 1800cb and I doubt that 2500cb is the highest we'll see from the 7900X. 

 

I was just trying to put all the talk of excessive power consumption into perspective. 

 

In the end, for every what drawn by the 7900X / x299 platform, you get a decent amount of extra performance in return. 

It's actually the first 2500 cb score I've seen on the chip, but OC3D also said the BIOS update they got let them get higher RAM timings.  They couldn't get 3600 previously, but they could for the OC tests.

 

I don't feel one should ever use OC results for comparing CPUs or GPUs. Yes, a user might get a better CPU and get more out of it, but it's a bad way to test.

 

For the Encode scores, it's about a 35% uplift from 1800X -> 7900X.  18% from 1800X -> 7820X.  A bit more down-scaling going from 8c to 10c.  Something that's started to crop up in a lot of these testing. 

[Jumper118]

3 minutes ago, Taf the Ghost said:

 

I don't feel one should ever use OC results for comparing CPUs or GPUs. Yes, a user might get a better CPU and get more out of it, but it's a bad way to test.

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. . 

[NuclearKing]

3 minutes ago, Taf the Ghost said:

I don't feel one should ever use OC results for comparing CPUs or GPUs. Yes, a user might get a better CPU and get more out of it, but it's a bad way to test.

I think a chip that overclocks nicely is added value to the product, and should be considered when reviewed. Of course stock vs stock performance should be compared, but most of us slightly more advanced PC users can and will certainly overclock, so we have to take that into consideration.

 

Now of course you don't just go out and find the best overclocking 7900X and call it a fair review and add that to the value, but the average OC for a chip architecture is definitely viable review material. (like 4.5-4.6 for Skylake, or 4.9+ for Kaby Lake)

[VagabondWraith]

I think an interesting test would be to downclock a 7820X (8c/16t) to 4GHz and compare to the 1800X and see where performance stands between the two at the same clocks. I'll do that once my rig is up and running if no one else has yet. Would be a good comparison, I think.

[NuclearKing]

3 minutes ago, VagabondWraith said:

I think an interesting test would be to downclock a 7820X (8c/16t) to 4GHz and compare to the 1800X and see where performance stands between the two at the same clocks. I'll do that once my rig is up and running if no one else has yet. Would be a good comparison, I think.

Spoiler

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.

[PCGuy_5960]

5 minutes ago, VagabondWraith said:

I think an interesting test would be to downclock a 7820X (8c/16t) to 4GHz and compare to the 1800X and see where performance stands between the two at the same clocks. 

I think the 7820X will be 5-10% faster (because it has 5-10% better IPC)

[Binks_no_Sake]

5 minutes ago, VagabondWraith said:

I think an interesting test would be to downclock a 7820X (8c/16t) to 4GHz and compare to the 1800X ...

Do that please ^^ it's a great idea