Enermax first out the gate with AIO for Threadripper.

[dizmo]

23 minutes ago, Brooksie359 said:

your joking right? they are the same chip so there is no reason for the r5 1600 to beat the r7 ever when they are clocked at the same speed. the only reason why any reviews showed the r5 being faster when at the same clocks is because they didn't retest the r7 cpus with all the updates to games and bios. is it worth the price difference for the small increase in gaming performance? no, not really but to say the r7 isn't faster is complete nonsense.

Take a look at the Digital Foundry video from yesterday. The proof is there. 

[Brooksie359]

3 minutes ago, mariushm said:

That makes it possible to set the cores for the application only.  The application (game) interacts with video card driver, direct x , other things (like steam overlay for example), all those can have threads randomly created on the other die of the CPU.

You can't guarantee 100% of the time that everything which is used to run the game will run on just one die.

 

Your game is not something like a command like video encoder (x264), which is self contained and does everything from one executable.

The game relies on external things (DirectX , OpenGL, Cuda, Physix) and internal things like sound engines like Miles3D , physics engines like Havok, AI and other helper libraries ex pathfinding etc  etc)

 

yes but if you set the affinity you can basically eliminate the majority of it. also if you really wanted to you could use vms to force it to run on one die. 

[Brooksie359]

3 minutes ago, dizmo said:

Take a look at the Digital Foundry video from yesterday. The proof is there. 

yeah the r7 is faster in that video so idk what you are talking about. also if you really want to argue that the r5 1600 is faster you can even change the r7 to an r5 1600 configuration in the bios if you so choose so there is no reason at all to say the r5 preforms better in games.

[Brooksie359]

6 minutes ago, dizmo said:

Take a look at the Digital Foundry video from yesterday. The proof is there. 

also go to 14:35 in the video and he says this "the ultimate gaming performance you can get right now from ryzen, well that's the ryzen 7 series with a 4.0ghz overclock" so don't go trying to prove your point with a video that literally say in that the 7 series is faster in gaming. is it worth the extra cost if you just want to game? probably not but again the r7 is still objectively faster.

[dizmo]

 

7 minutes ago, Brooksie359 said:

also go to 14:35 in the video and he says this "the ultimate gaming performance you can get right now from ryzen, well that's the ryzen 7 series with a 4.0ghz overclock" so don't go trying to prove your point with a video that literally say in that the 7 series is faster in gaming. is it worth the extra cost if you just want to game? probably not but again the r7 is still objectively faster.

I wouldn't really call a few fps "faster" that's basically margin of error and not even noticeable. If it was a gap like we see in the i5/i7 comparison, then yeah, it's undoubtedly faster. Not only that the 1600x passes it several times.

Why change it? That's irrelevant data.

Your timestamp must be wrong, because there's no speech at that part of the video. In fact, if you watch for another 10 seconds you'll see the massive dips in frame timing that R7 is experiencing, and, like I said, a 5 fps delta.

[Brooksie359]

Just now, dizmo said:

 

I wouldn't really call a few fps "faster" that's basically margin of error and not even noticeable. If it was a gap like we see in the i5/i7 comparison, then yeah, it's undoubtedly faster. Not only that the 1600x passes it several times.

Why change it? That's irrelevant data.

Your timestamp must be wrong, because there's no speech at that part of the video. In fact, if you watch for another 10 seconds you'll see the massive dips in frame timing that R7 is experiencing, and, like I said, a 5 fps delta.

if you start at 14:35 you will see him say it within the next 10 seconds. also if you look at the crisis gameplay you will see the fps difference is much higher than a couple fps. bottom line is the r7 is faster no matter how much you want to deny it. is it a huge difference? no, like I said before it really isn't something to choose the r7 over the r5 for but it is still there.

[Doobeedoo]

Awesome. Finally to see AiO for TR and it's huge cooling plate. Thought maybe someone other would come with it first. But surely others to follow. 

[tom_w141]

1 hour ago, dizmo said:

Kind of, not really. When a company makes as much money as Enermax,

Wanna bet that it doesn't cost that much money?

Saying those expenses are invalid because a company makes plenty of money is BS. It takes money to make money. New products are expensive, especially if you are 1st and not directly copying something. They have spent money therefore they are invested.

[Brooksie359]

Just now, tom_w141 said:

Wanna bet that it doesn't cost that much money?

honestly the biggest cost is going to be creating a production line for it. everything else would be quite easy. if they have to buy new machines and create new tooling it will cost alot of money. I'm unsure just how much but it will most likely be one of the biggest costs related to creating a new liquid cooler like this.

[Jito463]

1 hour ago, Zando Bob said:

I already heard they ticked off George Lucas by adding paper in TLJ. 

I read that initially as The Longest Journey, and was scratching my head trying to figure out how Lucas had anything to do with that video game. 

[Zando_]

7 minutes ago, Jito463 said:

I read that initially as The Longest Journey, and was scratching my head trying to figure out how Lucas had anything to do with that video game. 

Haha! Yeah, I was abbreviating The Last Jedi. When I first got onto Star Wars forums and stuff, I couldn't figure out for the life of me what they meant by ROTJ, ESB, and ANH. 

[Marinatall_Ironside]

2 hours ago, dizmo said:

I wonder how Threadripper will do in gaming...

The Ryzen Threadripper CPUs will have a Precision Boost of up to 4.0 GHz (on a couple of CPU cores). I'm not sure of the turbo frequency on all cores, but that'll come in the performance reviews.

[ravenshrike]

4 hours ago, Brooksie359 said:

your joking right? they are the same chip so there is no reason for the r5 1600 to beat the r7 ever when they are clocked at the same speed. the only reason why any reviews showed the r5 being faster when at the same clocks is because they didn't retest the r7 cpus with all the updates to games and bios. is it worth the price difference for the small increase in gaming performance? no, not really but to say the r7 isn't faster is complete nonsense.

Technically it has more cache per core than the r7.

[SurvivorNVL]

Well, now I need Threadripper, and that AIO.

[imreloadin]

Anyone know when the embargo lifts on reviews of TR?

[The Benjamins]

2 hours ago, imreloadin said:

Anyone know when the embargo lifts on reviews of TR?

I expect it to be 9am PST

[EnergyEclipse]

Thank flaming goodness someone finally has made a cooler adequate enough for a GTX 480, all these many years later.

 

Wait, 

does that mean Enermax were using Internet Explorer to keep up with the 'latest' news?

[ravenshrike]

1 hour ago, EnergyEclipse said:

Thank flaming goodness someone finally has made a cooler adequate enough for a GTX 480, all these many years later.

That's... an interesting point. It should be relatively simple to cut down the cold plate to Vega size and attach a 92mm fan that would sit in the empty space on a Vega board routed over the VRMs?

[dizmo]

10 hours ago, tom_w141 said:

Wanna bet that it doesn't cost that much money?

Saying those expenses are invalid because a company makes plenty of money is BS. It takes money to make money. New products are expensive, especially if you are 1st and not directly copying something. They have spent money therefore they are invested.

Think about what they actually had to make. The block. That's it. Pretty much everything else is already there, there's very little R&D going into the radiator, etc. They're not completely building this from the ground up. I'm not saying it's free, but it's not as expensive as you think.

[Jito463]

2 minutes ago, dizmo said:

Think about what they actually had to make. The block. That's it. Pretty much everything else is already there, there's very little R&D going into the radiator, etc. They're not completely building this from the ground up. I'm not saying it's free, but it's not as expensive as you think.

While I agree with you to some extent, there's still engineering required to produce this correctly.  They have to ensure the thickness of the plate is appropriate for it's size, and that the heat will transfer adequately.  They have to ensure the pump and the liquid flow is adequate to move the heat away from the CPU within the given TDP window (likely with a margin appropriate for overclocking).  It's also possible they've created a thicker (or thinner) rad for TR, so they may well have spent R&D on that front.

 

While much of that could be worked out from their existing line, it still requires time and resources to make sure it's done right.  How much cost that entails, only Enermax knows.

[Tam3n]

Well, one problem the AIOs generally have vs custom loop, is too weak flow rate. If they have remedied that, it could be cood. Coupled with a decent triple fan radiator it should provide enough cooling for the TR.

[Skanky Sylveon]

15 hours ago, mariushm said:

There is a bit of latency when two threads across dies have to talk to each other or transfer data. Threads within a single die would work faster.

 

There is also latency when two CCXs need to communicate with each other, this is a limitation with the infinity fabric.  You can minimize the latency by getting fast RAM, as the infinity fabric's speed is dependent on RAM speeds.

15 hours ago, mariushm said:

 

If you have a game using a few threads that need to exchange small bits of data but lots of times, that game may run slightly better on a Ryzen because all threads will be equal and would work on same die.

With Threadripper, every once in a while, the operating system will put some threads on one die and some threads on the other die. The OS may even periodically move threads from die to die in order to allocate core resources better.

Fun fact, this may be one of the reasons why Ryzen, in general performs worse in games than Intel's CPUs.  Like I said before, this limitation isn't only on Threadripper.

15 hours ago, mariushm said:

 

Even if a game is extra careful about placing its threads and keeping them on one die, or maybe if AMD comes up with custom cpu driver that would place threads of one application preferentially on one die, with games you'd still have DX 12 threads, video card driver threads, then game threads .. which would be considered separate applications (i think).

The 1700 has four cores per CCX, while the 1600x has only 3 cores per CCX, same with Threadripper, with the 1900x having only 2 cores per CCX.  Depending on the chip, it would be impossible to have certain games run on only one CCX.

15 hours ago, mariushm said:

So one Ryzen and one Threadripper both at same frequencies, ignoring the benefits of quad channel ddr4 of Threadripper... there's quite a possibility that Ryzen system will be slightly faster (maybe less than 1% difference in frame rates and all that)

All of the issues that I stated above could be more severe in Threadripper, although I don't think that the difference would be significant.  The fact that Threadripper has quad channel RAM may actually make it perform slightly better in games, since like I said before, the infinity fabric's speed is dependent on RAM speeds.  A faster infinity fabric means less latency.

15 hours ago, mariushm said:

It will be very small because threads in real world don't really talk a lot between each other and basically, there's a huge gap between  percentage of time spent talking and percentage of  time doing their job, so while there is some latency, the impact is very small in most cases.

Most cases, yes.  Games usually have more frequent core to core communication than most other applications though.

[Bananasplit_00]

6 hours ago, ravenshrike said:

That's... an interesting point. It should be relatively simple to cut down the cold plate to Vega size and attach a 92mm fan that would sit in the empty space on a Vega board routed over the VRMs?

The Vega VRM puts out like 40W, I'm not sure a 92mm fan directed at it will cut it lol

[mariushm]

25 minutes ago, Bananasplit_00 said:

The Vega VRM puts out like 40W, I'm not sure a 92mm fan directed at it will cut it lol

The Vega FE VRM had 16 phases (8 plus doublers), 40w spread over 16 phases is more or less around 2.5w per phase ... the mosfets are rated for something like 5-10c per w above ambient with the tiny amount of heatsink and dissipation in pcb..  so the vrm will sit at around 50-70c while all mosfets are rated for up to 125c

 

Maybe Vega56 will only be 6phase and the Vega64 will be plain 8 phase to make the cards cheaper, but they'll still be cool enough

[EnergyEclipse]

11 hours ago, ravenshrike said:

That's... an interesting point. It should be relatively simple to cut down the cold plate to Vega size and attach a 92mm fan that would sit in the empty space on a Vega board routed over the VRMs?

Actually yes, I didn't realise that. Maybe once VEGA comes around we will see a new Kraken AIO Mount for it, were for the types of TDPs I expect the VEGA then that kind of a cooler would be amazing  I'd say however you're probably going to want both heat sink and fan for the VRMs, as I think were going back to an era of "Did I buy a GPU or a barbeque?"