Poll: ** The AMD VEGA Thread **

[Jono8]

That's just another issue so far. The engineering sample was running Debugging Fiji drivers, on debugging usb monitoring PCB, and what seems to be 1200Mhz. Yet performed better than Frontier Edition.

So many unanswered questions and issues that need to be addressed.

I asked earlier in the thread, but where is this 1200mhz thing coming from? Is it just that 3dmark screenshot?

 

[WeSbO]

it will be closer to 484, since he did mention it was under 500mm^2 previously, so anywhere between 484 and 500. when i looked at PCPers teardown, he was out by 1mm when measuring the cores length, which to me put it around 493mm^2.

If it was 484 that would make PCper overall 5mm out on his measurements, so at best 2.5 on each side, maybe more on one side and less on another (because you can see physically that it isn't square), in all cases 5mm (or 2 for that matter) is a lot even when eyeballing with a caliper. I would bet any kind of money that the FE die is not 484mm2 and is closer to PCpers measurements,

EDIT : and again why wouldn't AMD just come out and give the dimensions, not likes it's ultra top secret info, Nvidia already know the size within 0.001 mm probably by now...

EDIT 2: just realized I F'ed up my calculations, yeah, wouldn't have to be that far out to get 484mm2.

 

[N19h7m4r3]

I asked earlier in the thread, but where is this 1200mhz thing coming from? Is it just that 3dmark screenshot?

So far yes, for the same device ID.
Same ID was in Ashes tests scoring close 1080 as well.

As right now all Frontier Edition 3DMark tests report 1600Mhz, even if it's recorded running at 1300-1450Mhz. So for some reason 3DMark always reports the max clocks, even if the card itself never gets near it.

Looking at the 3D Marks scores from the Engineering Sample at 1200Mhz, it was between 1070 and 1080 at reported 1200Mhz.
Frontier Edition at reported 1600Mhz is at GTX 1080 scores in Firestrike.

So stuff it still off on specs to performance on Frontier Edition. Hell, it's as fast as my Hybrid 980Ti :/


Edit. Seems Frontier Edition is doing worse in Sniper Elite 4 than the last demo's AMD showed with an FPS counter as well. They were showing around high 60's - mid 70's. FE can't even break 60.

 

[Tommy_Here]

Are you ignoring the multiple freesync monitor users like myself for whom an Nvidia card is simply out of the equation?

I don't care that it's a year or so after whatever Nvidia cards have released. That just means my upgrade cycle for gpus is different to someone with an Nvidia card.

If Vega performs at 1080 levels for a decent price then people like myself with freesync screens will more than likely buy it. I will not buy a Nvidia card until they support freesync

Freesync G-sync whatever.... that's why I won't buy a monitor that makes me stick to either AMD or Nvidia.

My point still stands releasing a card 15 months after a 1080 and only matching it's performance is pathetic imo.

 

[Jono8]

So far yes, for the same device ID.
Same ID was in Ashes tests scoring close 1080 as well.

As right now all Frontier Edition 3DMark tests report 1600Mhz, even if it's recorded running at 1300-1450Mhz. So for some reason 3DMark always reports the max clocks, even if the card itself never gets near it.

Looking at the 3D Marks scores from the Engineering Sample at 1200Mhz, it was between 1070 and 1080 at reported 1200Mhz.
Frontier Edition at reported 1600Mhz is at GTX 1080 scores in Firestrike.

So stuff it still off on specs to performance on Frontier Edition. Hell, it's as fast as my Hybrid 980Ti :/


Edit. Seems Frontier Edition is doing worse in Sniper Elite 4 than the last demo's AMD showed with an FPS counter as well. They were showing around high 60's - mid 70's. FE can't even break 60.

Hm, i would be wary of using that reported 1200mhz to base anything off. I mean it would be good if that was true, but 3dmark often seems a bit random in what clocks it reports.

 

[N19h7m4r3]

Hm, i would be wary of using that reported 1200mhz to base anything off. I mean it would be good if that was true, but 3dmark often seems a bit random in what clocks it reports.

As I added in after FE can't even match their most recent 4K game demo where FPS was shown. It's 10-15FPS slower in Sniper Elite 4.

Something off there as well.

Raja will have a lot of explaining to do at Siggraph, and Lisa and Co will likely get hammered during the Investor call for Q2. Looking forward to that.

 

[Shaz12]

If the RX Vega is that mich faster than the frontier then why does the FE even exist? It does nothing right.

 

[Shaz12]

So far yes, for the same device ID.
Same ID was in Ashes tests scoring close 1080 as well.

As right now all Frontier Edition 3DMark tests report 1600Mhz, even if it's recorded running at 1300-1450Mhz. So for some reason 3DMark always reports the max clocks, even if the card itself never gets near it.

Looking at the 3D Marks scores from the Engineering Sample at 1200Mhz, it was between 1070 and 1080 at reported 1200Mhz.
Frontier Edition at reported 1600Mhz is at GTX 1080 scores in Firestrike.

So stuff it still off on specs to performance on Frontier Edition. Hell, it's as fast as my Hybrid 980Ti :/


Edit. Seems Frontier Edition is doing worse in Sniper Elite 4 than the last demo's AMD showed with an FPS counter as well. They were showing around high 60's - mid 70's. FE can't even break 60.

will you look at that list of NVIDIA cards at the top? This chart makes it look like AMD is generations behind NVIDIA in the GPU market.

 

[N19h7m4r3]

If the RX Vega is that mich faster than the frontier then why does the FE even exist? It does nothing right.

will you look at that TitanXP go? This chart makes it look like AMD is generations behind NVIDIA in the GPU market.

The Frontier Edition outside of Tomb Raider is essentially scoring 980Ti numbers, that is just above Fury X.


It's losing to a reference 1080 in Ashes of the Singularity as well!!!

Old engineering sample matched 1080 in Ashes.
https://www.overclock3d.net/news/gpu_displays/alleged_amd_vega_benchmarks_appear_online/1

Double the TFLOPs, and 50% smaller process, 4x the VRAM that's faster than Fiji, and is 50% slower than TitanXp, and losing to its own previous engineering sample demos.

 

[Nablaoperator]

Raja just confirmed on Twitter that Vega is 484mm2.

So it's really comparable to GP102 in size.

Still open to interpretation ... can be either 484 or 529 mm2

C'mon dudes... no one in the community who'd like to activate few industry connects.. to learn more about the transistor count

EDIT: my bad with the manual compute bit.. need to practice patience by waiting for rx vega

 

[Mauller]

If the RX Vega is that mich faster than the frontier then why does the FE even exist? It does nothing right.

It has a focus on pro software performance, but that performance will likely increase if it is true that hardly any of the new vega features are enabled yet.

We will likely see what the architecture is more capable of with RX Vega since there is no emphasis with rendering accuracy, so features that are not considered driver stable for pro workloads can be enabled.

 

[shankly1985]

Are you ignoring the multiple freesync monitor users like myself for whom an Nvidia card is simply out of the equation?

I don't care that it's a year or so after whatever Nvidia cards have released. That just means my upgrade cycle for gpus is different to someone with an Nvidia card.

If Vega performs at 1080 levels for a decent price then people like myself with freesync screens will more than likely buy it. I will not buy a Nvidia card until they support freesync

Agreed!
People on here think GPU warz is the norm lol said it time and time again I couldn't give a toss how Vega compares with Nvidia just so long what I buying is an upgrade for myself.

 

[Nexus18]

Agreed!
People on here think GPU warz is the norm lol said it time and time again I couldn't give a toss how Vega compares with Nvidia just so long what I buying is an upgrade for myself.

But but but..... you have to impress people on an internet forum shankly

 

[chrcoluk]

The profit margins on high end products are absolutely huge. Due you really think the Titan XP costs $500 more than the 1080 ti to make?

The total volume is much smaller and so so that will limit revenue, but its easy money. that is why they do it. If there was no healthy profit margins it would be a pointless endeavor. Low end parts will have in tota; a higher volume and net profits will add up, so hat is the meat and portatoes. But the high end pays back a nice chunk of the R&D for relatively little work, as long as you have a good architecture. 1080/Ti/TXP are basically just scaled up 1050/1060 GPUS so the core design costs are reasonable. The work in aking the 1080ti perform as wlel as it does ill be valuable for next generations 1060 product (e.g. Volta 2060). All the while they might get $150 a card sold.

volume beats profit margin.

Also titans very likely have lower yields than say a gtx 1060 so the profit margin may not be as big as you think.

I agree a titan will have more profit per card, but overall profit it will be left in the dust by 1060s.

 

[shankly1985]

But but but..... you have to impress people on an internet forum shankly

Mine is bigger than yours

 

[LoadsaMoney]

so its a professional card, that you can also do some gaming on as welll if you want to, and if you do, you'll get 1070 level performance.

Doesn't sound too bad, as 1070s are decent at gaming.

 

[D.P.]

It has a focus on pro software performance, but that performance will likely increase if it is true that hardly any of the new vega features are enabled yet.

We will likely see what the architecture is more capable of with RX Vega since there is no emphasis with rendering accuracy, so features that are not considered driver stable for pro workloads can be enabled.

If Vega has a focus on pro software why is it slower than a 1080 running lot drivers

 

[Mauller]

so its a professional card, that you can also do some gaming on as welll if you want to, and if you do, you'll get 1070 level performance.

Doesn't sound too bad, as 1070s are decent at gaming.

It will likely gain a decent performance boost once the new vega features codebase is considered stable and added to the driver. if they actually are just using a fiji like fallback mode at the moment for rendering stability.

 

[shankly1985]

so its a professional card, that you can also do some gaming on as welll if you want to, and if you do, you'll get 1070 level performance.

Doesn't sound too bad, as 1070s are decent at gaming.

It's only bad if you buy this GPU for only gaming. No one in the right mind would or should. And why it's only getting compared to gaming on forums is also very silly.

The GPU main focus is creation with light gaming optimised for them tools for play testing.

 

[Mauller]

If Vega has a focus on pro software why is it slower than a 1080 running lot drivers

I already explained that in the same quote. if its true that the features aren't code stable yet to be considered for a pro driver, they won't be included. but quadro cards have long stable drivers of course.