AMD THREADRIPPER VS INTEL SKYLAKE X

[B1gbeard]

Ah, the BIOS argument... A new BIOS is not going to magically push it performance up by 30% to get it inline with where is should be.

I'll tell you what its doing, it uses so much power its hitting the celling and throttling, it can't run those 4.5Ghz low threaded clocks because that puts it way over its 140 Watt TDP and thermal limits, its a 6950X on steroids, yet those steroids are causing it to crash and burn, literally.
Actually you might be right, with a BIOS update to give it less aggressive volts its performance would improve as it would put it back inline with the 6950X power levels, but thats not what Intel want, they are stuck on the same path AMD was with Bulldozer, the only way to get the performance up is to up the power and Mhz, the only problem is thermals and motherboard TDP protection limits prev ent it from clocking that high.
------------

Nigel is the unfortunate guy, at an un-namable competitor high-street store, Nigel dew the short straw this morning, Nigel was patted on the back by his manager and told "its ok, they will believe anything you tell them"
Nigel nervously takes his position on the shop floor, the one infront of the two screens, Nigel puts his best fake smile on and proceeds to tell a small gathering of customers to whiteness with their own eye's about how the £50 HDMI lead on the left connected to the hidden Blueray Player has so much better image quality than the £10 HDMI lead on the right connected to the hidden DVD Player.

Nigel, "your friendly Know How Customer Service Agent"

Thats who he is, we have all seen Nigel, he's an obvious Nigel.

I didn't mention the BIOS as an argument buddy bloody hell stop being so defensive. Just stating that we can't see who's using what for both sides.

I'm more disappointed that he is not called Nigel tbh. That shall be his name from now until the end of time.

Hopefully Threadripper will make our decision easy.

 

[zornyan]

Other reviews show the opposite, while a few have a suspicious lack of game reviews.

older bios, memory performance was behind until wednesday when a new bios revision came out, hence why oc3d now has the 7900x faster than the 7700k in games and overall much better performance than other reviewers.

 

[zornyan]

Not sure why it's that surprising. It's well known that most new games use more than 4 cores effectively, it's just that the balance of "more, slower cores" versus "fewer, faster cores" was tipping either way depending on the game when considering average frame rates with the R7 1800X and the i7-7700K. Now we basically have something closer to the per-core speed of the i7-7700K with the high core count of the R7 1800X, it's obviously going to be faster, on average (some games will still favour i7-7700K @ 5 GHz for sure). Of course you have to pay a hell of a lot for that performance though.

Those who were spouting "single core performance is still king in gaming" after Ryzen was released will now have to rationalise that somehow with these results. I look forward to seeing how they do that and I also look forward to seeing what Zen+ brings - by the time it's released newer games will certainly be more optimised for Ryzen than a lot of the current crop are. If it can also clock higher to close the gap with Intel on that front, it will tip the scales so much that the quad core i7s will become obsolete (at their current prices anyway).

This is all talking about 1080p 144 Hz gaming by the way. At 1440p and 4K the gap is already so small as to make no difference with all but the very highest end GPUs, and modern GPUs can run 1080p 60 Hz with pretty much any CPU for most games anyway.

more showing that It was obviously a bios bug with first review he done, his 7820x at 4.8ghz should beat out the 4.6ghz 7900x since those game won't scale past that amount of threads, yet obviously this bios revision fixed memory issues and now it's a fsir bit faster (enough to beat a 200mhz clock defecit)

 

[humbug]

I didn't mention the BIOS as an argument buddy bloody hell stop being so defensive. Just stating that we can't see who's using what for both sides.

I'm more disappointed that he is not called Nigel tbh. That shall be his name from now until the end of time.

So why care about what BIOS revisions reviewers are using?

Nigel is good name for him, all he needs is the 'Know-How' branded name tag.

 

[B1gbeard]

So why care about what BIOS revisions reviewers are using?

Nigel is good name for him, all he needs is the 'Know-How' branded name tag.

Mostly from an AMD perspective really. If they are not using the latest one that improves memory speeds they are pretty pointless to me anyway as it seems to have a big impact on my usage for the CPU.

 

[humbug]

Mostly from an AMD perspective really. If they are not using the latest one that improves memory speeds they are pretty pointless to me anyway as it seems to have a big impact on my usage for the CPU.

Its understandable given the huge performance gain AMD get from running fast memory vs slow, but there is a very good reason why, that reason simply doesn't exist with Intel, the difference on Intel between even a massive memory speed gap like 2133Mhz and 3200Mhz is negligible, its there, but only just.

I maintain what i said before.....

It uses so much power its hitting the celling and throttling, it can't run those 4.5Ghz low threaded clocks because that puts it way over its 140 Watt TDP and thermal limits, its a 6950X on steroids, yet those steroids are causing it to crash and burn, literally.
Actually you might be right, with a BIOS update to give it less aggressive volts its performance would improve as it would put it back inline with the 6950X power levels, but thats not what Intel want, they are stuck on the same path AMD was with Bulldozer, the only way to get the performance up is to up the power and Mhz, the only problem is thermals and motherboard TDP protection limits prevent it from clocking that high.

So yes in a different way i think you have a point, but its not memory, its simply that Intel in a bid to get more performance out of it have set it up so aggressively its power and thermal envelopes can't take it and it throttles.

The Motherboard is protecting its self from burning its VRMs and the CPU is protecting its self from its self.
Again if Intel tone it down a bit the performance isn't going to be as high as they like but it will be higher than it currently is.

 

[B1gbeard]

Its understandable given the huge performance gain AMD get from running fast memory vs slow, but there is a very good reason why, that reason simply doesn't exist with Intel, the difference on Intel between even a massive memory speed gap like 2133Mhz and 3200Mhz is negligible, its there, but only just.

I maintain what i said before.....



So yes in a different way i think you have a point, but its not memory, its simply that Intel in a bid to get more performance out of it have set it up so aggressively its power and thermal envelopes can't take it and it throttles.

The Motherboard is protecting its self from burning its VRMs and the CPU is protecting its self from its self.
Again if Intel tone it down a bit the performance isn't going to be as high as they like but it will be higher than it currently is.

It's tough for someone to make an informed decision. I want to see Ryzen at it best against a 7820 or 7900 at it's best but we don't seem to get that or at least many of the reviews are not stating they are using the beta Ryzen BIOS. I wonder if Threadripper will have similar teething problems.

 

[humbug]

Yeah i agree completely.

And IMO Threadripper will have a problematic launch. I can't remember the last time AMD had a perfect launch.

 

[smogsy]

Its understandable given the huge performance gain AMD get from running fast memory vs slow, but there is a very good reason why, that reason simply doesn't exist with Intel, the difference on Intel between even a massive memory speed gap like 2133Mhz and 3200Mhz is negligible, its there, but only just.

I maintain what i said before.....



So yes in a different way i think you have a point, but its not memory, its simply that Intel in a bid to get more performance out of it have set it up so aggressively its power and thermal envelopes can't take it and it throttles.

The Motherboard is protecting its self from burning its VRMs and the CPU is protecting its self from its self.
Again if Intel tone it down a bit the performance isn't going to be as high as they like but it will be higher than it currently is.

you watched the Pcper live stream fully, they asked intel about the 70W extra power & they stated they are not seeing that in their testing Seems like BIOS Bug on ASUS boards at this point
not defending Intel Here, BUt looks like Ryzen/SkyLake-X both have/had early releases bite them in the ....

heck Nigel Guy Cpu-Z was showing 1.9V at some points... (no wonder heat is a issue)

 

[gavinh87]

you watched the Pcper live stream fully, they asked intel about the 70W extra power & they stated they are not seeing that in their testing Seems like BIOS Bug on ASUS boards at this point
not defending Intel Here, BUt looks like Ryzen/SkyLake-X both have/had early releases bite them in the ....

heck Nigel Guy Cpu-Z was showing 1.9V at some points... (no wonder heat is a issue)

At 1.9v that cpu would be dead.

 

[HeX]

Yeah i agree completely.

And IMO Threadripper will have a problematic launch. I can't remember the last time AMD had a perfect launch.

I think AMD have pulled a blinder with the TR launch and it won't be as problematic as you think.

They've cleverly used the mainstream Ryzen launch as a 'beta' for EPYC and Threadripper, they've had months of Ryzen testing, rolled out the new AGESA updates, and given the mobo makers time to get up to speed with the new BIOS code.

All that knowledge, bug fixing, and tuning is directly applicable to EPYC and TR. Which should hopefully mean a much smoother launch.

 

[subbytna]

It's tough for someone to make an informed decision. I want to see Ryzen at it best against a 7820 or 7900 at it's best but we don't seem to get that or at least many of the reviews are not stating they are using the beta Ryzen BIOS. I wonder if Threadripper will have similar teething problems.

Ryzen is a totally new built from the ground up platform. Intel have just rehashed the existing platform for a couple of generations now. You're always going to have optimisation issues with new platforms. I'm going to put my balls on the line here and say that AMD will make a huge comeback with CPU's (versus INTEL) in the next 12-18 months but with GPUs they will always be playing second fiddle to Nvidia.

When developers start to get away from the fact that they predominately optimise software for quadcores now, and focus on 6 or more cores, then we'll see AMD make a comeback. Most people, outside the enthusiast community, still run either dual or quads so developers optimise for the majority, not the minority. If AMD can make 6 cores as the new minimum, then we'll see huge movement in developers using those extra cores/threads and the gap between INTEL and AMD will close significantly.

 

[humbug]

you watched the Pcper live stream fully, they asked intel about the 70W extra power & they stated they are not seeing that in their testing Seems like BIOS Bug on ASUS boards at this point
not defending Intel Here, BUt looks like Ryzen/SkyLake-X both have/had early releases bite them in the ....

heck Nigel Guy Cpu-Z was showing 1.9V at some points... (no wonder heat is a issue)

Of course Intel are not seeing that in their testing, yet as Ryan said everyone 'outside of Intel's testing' is.

What do you expect them to say? "oh yeah, not very power efficient are they? eh? dropped a stinker on our customers there didn't we"

No, Intel think they are perfect, its something or someone else...

 

[SiDeards73]

"older bios, memory performance was behind until wednesday when a new bios revision came out, hence why oc3d now has the 7900x faster than the 7700k in games and overall much better performance than other reviewers."

Lol'ing hard at this, from the same person who argued AMD Ryzen cpu's were rubbish when they first come out, double standards much? i like how you picked your words extremely carefully, shame you cant give AMD the same credit....

Shill is a shill is a shill tbh...

 

[smogsy]

Of course Intel are not seeing that in their testing, yet as Ryan said everyone 'outside of Intel's testing' is.

What do you expect them to say? "oh yeah, not very power efficient are they? eh? dropped a stinker on our customers there didn't we"

No, Intel think they are perfect, its something or someone else...

maybe or Maybe theirs just a huge bug.
OC3d Stated that his cPu/Mobo combo just defaullted to turbo clocks 24/7 so well see

 

[zornyan]

"older bios, memory performance was behind until wednesday when a new bios revision came out, hence why oc3d now has the 7900x faster than the 7700k in games and overall much better performance than other reviewers."

Lol'ing hard at this, from the same person who argued AMD Ryzen cpu's were rubbish when they first come out, double standards much? i like how you picked your words extremely carefully, shame you cant give AMD the same credit....

Shill is a shill is a shill tbh...

difference is this has been fixed pre release, ryzen is still being fixed months after release

that and ryzen is still sub par in comparison

 

[KompuKare]

I was wondering myself where were the 720p benchies.

Well, I think far more interesting is why so few sites did gaming benches in the first place?

If it is as suspect in that Intel told them not to, or promised that with new BIOS versions or microcode there would be big improvements, why did nobody mention this?

Obviously, for reviewers clicks = money so with only very few reviewers actually being privileged with getting samples you'd wonder if those not toeing the line are going to get samples next time?

Of course, if new BIOS/microcode revisions come out with improve games I would expect reviewers to revisit their benches (well the not lazy ones), but during the whole Ryzen launch was not an often quoted reason to go with Intel that their platform is more polished and that justifies the extra cost? (This was aside from the fact that X99 had major teething issues too.)

But the real problem (at least for now) seems to be that Skylake-X's new cache structure doesn't suit some loads. Whether that can be fixed or not is the question. In way, the speculation about the new cache is that is designed for AVX512 and HPC loads at the expense of more normal loads. Still no proper server load reviews so hard to tell how the new cache performs there, but so far we have seen that for games it is not good. Now, since they are using two dies for Skylake-X and really AVX512 and HPC are more suited to the 18+ core chips, would it not have made more sense to confine the new cache to those bigger dies (Xeon LGA 3647) and design a mid-way cache structure between desktop and that for the medium and small Skylake-X dies? Of course, cost and time to market may have precluded that.

 

[KompuKare]

All that knowledge, bug fixing, and tuning is directly applicable to EPYC and TR.

This is another advantage of their 'one die to rule them all' approach. Obviously, quad or octo channel will have differences but most of the work has been done.
Skylake is now nearly out for 2 years (Aug 2015) which should have given Intel plenty of time to polish Skylake-X, but not only is HEDT (and the corresponding Xeons) a different die, this time they massively changed the way their cache works making it far harder to heed the lessons they got with consumer Skylake.

 

[Panos]

difference is this has been fixed pre release, ryzen is still being fixed months after release

that and ryzen is still sub par in comparison

You are the one who has clearly lowered this thread to trolling trash.
Give him the troll medal someone.

 

[humbug]

Zornyan will have none to talk to if he still thinks his ignore list will make everyone stop responding to him