[PCWORLD] AMD provides a sneak peek at its Radeon HD 7990

[bystander36]

The problem with AMD is that it's just a smaller company, and probably hurting in terms of finances. What that means is a smaller engineering group with less resources. That doesn't mean they can't pull through still, but if you've got your entire driver team tied up trying to maximize single-gpu performances for games, then you won't have anyone to work on other projects. Even with Nvidia's (presumably much larger) group, they haven't truly fixed SLI so it really may be out of AMD's control for the moment.

What they need to do is hire the guy that made RadeonPro, or someone similar (if they haven't already). But it doesn't help that they're laying off or losing workers, as opposed to gaining them.

I'm sorry, but RadeonPro is not AMD's solution. Everyone who uses RadeonPro uses it as a FPS limiter. Setting the max FPS to be lower than their average FPS for the game and if the game varies a lot in FPS, you have to keep lowering that FPS limit for the worst case scenario.

Limiting FPS, while it helps a lot, does not make your cards look good in benchmarks.

 

[jackstar7]

I'm sorry, but RadeonPro is not AMD's solution. Everyone who uses RadeonPro uses it as a FPS limiter. Setting the max FPS to be lower than their average FPS for the game and if the game varies a lot in FPS, you have to keep lowering that FPS limit for the worst case scenario.

Limiting FPS, while it helps a lot, does not make your cards look good in benchmarks.

I'll take play experience over benchmarks any day.

 

[bystander36]

I'll take play experience over benchmarks any day.

My point is that SLI gives you good playing experience without limiting FPS. AMD is going to want to do the same. RadeonPro isn't a cure all, it is a work around. If AMD wants to give their customers a fix, they will be wanting to do better than forcing heavy FPS limiting on people.

After all, people choose crossfire for higher FPS or settings. FPS limiting kind of goes against that.

 

[jackstar7]

My point is that SLI gives you good playing experience without limiting FPS. AMD is going to want to do the same. RadeonPro isn't a cure all, it is a work around. If AMD wants to give their customers a fix, they will be wanting to do better than forcing heavy FPS limiting on people.

After all, people choose crossfire for higher FPS or settings. FPS limiting kind of goes against that.

I may have this wrong, but I was under the impression that nV uses frame limiters.

 

[ICDP]

My point is that SLI gives you good playing experience without limiting FPS. AMD is going to want to do the same. RadeonPro isn't a cure all, it is a work around. If AMD wants to give their customers a fix, they will be wanting to do better than forcing heavy FPS limiting on people.

After all, people choose crossfire for higher FPS or settings. FPS limiting kind of goes against that.

I have had my fair share of games that required FPS limits set on SLI hardware in the past. In general though the SLI is far more usable out of the box but lets not propogate a myth that it is perfect all the time.

I use CF 7950s and I get both higher eye candy and far higher average FPS. So while an FPS limiter does stop the cards running balls to the wall, they are still much faster than a single GPU.

For example a single HD 7950 @ 1100/1450 gave me ~40 FPS on average in TR3 @ 2560x1600 res, shadows high. CF 7950s at 1100/1450 with the Ultra Shadows gives me 60 FPS. 50% higher performance and higher graphics settings.

 

[Silverforce11]

I'm sorry, but RadeonPro is not AMD's solution. Everyone who uses RadeonPro uses it as a FPS limiter. Setting the max FPS to be lower than their average FPS for the game and if the game varies a lot in FPS, you have to keep lowering that FPS limit for the worst case scenario.

Limiting FPS, while it helps a lot, does not make your cards look good in benchmarks.

Wrong. I use Radeon Pro for the dynamic vsync, so i can play with vsync on in games and dont suffer the 60 -> 30 fps jumps when it dips below 60.

With that, can disable tripple buffering as thats my prior solution, and now games run smooth and no screen tearing.

 

[lavaheadache]

You guys might as well be arguing with stone walls and finger puppets. Brand loyalist are going to cherry pick, use part truths and flat out deny anything that goes against what they want to believe. There are very few of us Neutral Enthusiast that buy and use products from either camps and share the pro's and con's of each.

Up until about 3 months ago I use to let the ignorance get to me, since then I've let maturity get the best of me and just try to share my own experience with others without getting involved in these trivial arguments with forum trolls.

Radeon Pro is great and definately helps alleviate the stresses of dealing with AMD's seemingly inferior driver team. I highly recommend it to anyone who feels that the are affected by frame rate inconsistencies. IF you're not then kudos to you and enjoying your awesome hardware.

I've yet to try out Nvidia's current implementation of SLi but I'm hoping to pick up another Titan for short term anaylsis so I can see for myself if I feel like Sli has improved over previous iterations. Sli has come a long way since the adomination that was 7950 GX2 Quad.

 

[Grooveriding]

Some guy from EA put a picture on Reddit. They are using the 7990 for BF4 testing.

That is a cool looking card. With that design I'm assuming they didn't want to repeat the noise levels of the 6990.

 

[Final8ty]

I have had my fair share of games that required FPS limits set on SLI hardware in the past. In general though the SLI is far more usable out of the box but lets not propogate a myth that it is perfect all the time.

I use CF 7950s and I get both higher eye candy and far higher average FPS. So while an FPS limiter does stop the cards running balls to the wall, they are still much faster than a single GPU.

For example a single HD 7950 @ 1100/1450 gave me ~40 FPS on average in TR3 @ 2560x1600 res, shadows high. CF 7950s at 1100/1450 with the Ultra Shadows gives me 60 FPS. 50% higher performance and higher graphics settings.

You would get the impression that Nvinspector didn't exist.

 

[BallaTheFeared]

It's bad enough with Nvidia's frame metering, it's lost performance, and it's not like you're getting a discount on the second or third card, it's full price.

100% cost increase for 50% performance is awful, there is no other way to look at it for me.

To each their own, but doubling my cost for slightly better shadows strikes me as a joke.

 

[railven]

After all, people choose crossfire for higher FPS or settings. FPS limiting kind of goes against that.

Odd, as someone who normally uses v-sync, I found adding a second card let me add far more IQ and setttings and still maintain 60 FPS.

I've yet to have any of these insane microstutter issues and so far it's still just WoW that gives me a little stutter.

I must have gotten two golden cards, can run them 1125/1550 @ 1.110v. So far I'm very happy with my CFX adventures.

 

[UaVaj]

if radeonpro is truly amd's fix (reduce performance to reduce microstutter).

using icdp's example from post #88.
as is = 79fps.
with radeonpro frame cap = 50fps.
giving up 37% :'( of performance to get playable/enjoyable frame.

amd is 10% cheaper (2gb) or 27% cheaper (4gb) than nvidia
powercolor 7970 - $380 (after MIR)
sparkle 680 2gb - $424 or pny 680 4gb - $484

with frame capping - the performance to price-ratio has now shifted toward to nvidia.

-----

for anyone considering multi-gpu - the priority goal is typically performance with maximum scaling. why would any consider crossfire just to neuter it? after neutering - even nvidia 4gb edition (incase you run high high resolution) is offering a better performance to price-ratio. the nvidia 2gb edition is no contest.

7970 crossfire (aka 7990) = epic failed.

 

[ICDP]

It's bad enough with Nvidia's frame metering, it's lost performance, and it's not like you're getting a discount on the second or third card, it's full price.

100% cost increase for 50% performance is awful, there is no other way to look at it for me.

To each their own, but doubling my cost for slightly better shadows strikes me as a joke.

Ironic coming from someone who has defended Titan performance given it gives far worse performance gain for even higher % cost.

I like how you separated those 2 improvements like they are mutually exclusive. I didn't get one or the other, I got both at the same time. The ultra shadows bring the average FPS down to ~35 FPS when enabled at that resolution and the game becomes unplayable in some places as it drops into low/mid 20s. So the performance improvement is 71% when taken at apples to apples settings.

Still not worth it?

 

[UaVaj]

You guys might as well be arguing with stone walls and finger puppets. Brand loyalist are going to cherry pick, use part truths and flat out deny anything that goes against what they want to believe. There are very few of us Neutral Enthusiast that buy and use products from either camps and share the pro's and con's of each.

I've yet to try out Nvidia's current implementation of SLi but I'm hoping to pick up another Titan for short term anaylsis so I can see for myself if I feel like Sli has improved over previous iterations.

neutral enthusiast here too. purchase a pair of 7970 for crossfire even after all the warnings. test it out and it was no joy. i admit. i did not neuter it (37%) like icdp. that is simple again principal of multi-gpu. 680 sli now and smooth sailing. zero neutering.

pcper actually has the titan sli frame-rating results. for whatever their reasons is - they are simply not publishing it. at least not yet.

anyway when you do get titan sli. do post the results. that is the only thing keeping me from picking up a pair of titan. want some hard evidence as to how sli is working on titan before jumping on the wagon.

 

[railven]

if radeonpro is truly amd's fix (reduce performance to reduce microstutter).

using icdp's example from post #88.
as is = 79fps.
with radeonpro frame cap = 50fps.
giving up 37% :'( of performance to get playable/enjoyable frame.

amd is 10% cheaper (2gb) or 27% cheaper (4gb) than nvidia
powercolor 7970 - $380 (after MIR)
sparkle 680 2gb - $424 or pny 680 4gb - $484

with frame capping - the performance to price-ratio has now shifted toward to nvidia.

-----

for anyone considering multi-gpu - the priority goal is typically performance with maximum scaling. why would any consider crossfire just to neuter it? after neutering - even nvidia 4gb edition (incase you run high high resolution) is offering a better performance to price-ratio. the nvidia 2gb edition is no contest.

7970 crossfire (aka 7990) = epic failed.

Neuter it? Some of us don't like tearing. I can SLI GTX 680's I'd still use v-sync, so I guess by your explanation here I Just neutered my cards.

Ignoring that locking my FPS to 60 allows excess GPU power. This is like that awful comparative Keys made. If you're just going to use v-sync why even buy a 680, just get a 660 Ti.

My CFX runs at >90% load on each card with 60 FPS v-sync on a 1440p monitor without tearing. I guess I can be you and just remove the SSAA, get 140 FPS with >90% load on each and tearing left and right.

 

[UaVaj]

Ironic coming from someone who has defended Titan performance given it gives far worse performance gain for even higher % cost.

I like how you separated those 2 improvements like they are mutually exclusive. I didn't get one or the other, I got both at the same time. The ultra shadows bring the average FPS down to ~35 FPS when enabled at that resolution and the game becomes unplayable in some places as it drops into low/mid 20s. So the performance improvement is 71% when taken at apples to apples settings.

Still not worth it?

very good call. at the end of the day. it is all about performance vs price. unless you have a money tree in your back yard - then that does not apply to you.

titan performance/price ratio is probably as bad as 7970cf neutered if not worse.

 

[ICDP]

if radeonpro is truly amd's fix (reduce performance to reduce microstutter).

using icdp's example from post #88.
as is = 79fps.
with radeonpro frame cap = 50fps.
giving up 37% :'( of performance to get playable/enjoyable frame.

Look again, it was 69 FPS... not 79. So I gave up 27% performance to get stutter free and tear free gaming.

amd is 10% cheaper (2gb) or 27% cheaper (4gb) than nvidia
powercolor 7970 - $380 (after MIR)
sparkle 680 2gb - $424 or pny 680 4gb - $484

with frame capping - the performance to price-ratio has now shifted toward to nvidia.

Two very valid points need to be made that render your little exercise pointless.

1. My testing was done on 7950s, not 7970s. You can't take my 7950 perf numbers and apply them to a comparison between a 7970 vs 680.
2. As pointed out above the performance drop was 27% with the FPS cap, which throws your calculations right out the window.

-----

for anyone considering multi-gpu - the priority goal is typically performance with maximum scaling. why would any consider crossfire just to neuter it? after neutering - even nvidia 4gb edition (incase you run high high resolution) is offering a better performance to price-ratio. the nvidia 2gb edition is no contest.

You have taken what you feel to be the reason for purchasing multi-GPU and applied it to absolutely everyone. Just because you are happy to see massive screen tearing does not mean I or everyone else is. I and others in this thread are happy that with some tweaking they get excellent stutter and tear free gaming using their CF setup. I had 2x GTX680s in SLI and the experience was absolutely excellent, much easier to setup than CF 7950s. The problem was I always set 60 FPS with vsync to eliminate screen tearing. So I was neutering my GTX680 SLI way more considering they cost almost double the price of my 7950 CF setup. They lasted a few short weeks in my system before I sold one of them to cut my losses. Great setup but for me they didn't justify the cost given that I was limited to 60 FPS anyway.

7970 crossfire (aka 7990) = epic failed.

I understand you had an absolutely terrible experience with your 7970 CF setup. I would be angry as well considering the price you paid for what is undoubtedly a feature that needs a lot of work. The problem is that you can't conclude that everyone who has CF will share the same opinion.

 

[bystander36]

Wrong. I use Radeon Pro for the dynamic vsync, so i can play with vsync on in games and dont suffer the 60 -> 30 fps jumps when it dips below 60.

With that, can disable tripple buffering as thats my prior solution, and now games run smooth and no screen tearing.

All v-syncs do a 60/30/60/30 type solution if v-sync is actually enabled. I'm not certain if dynamic v-sync is an actually v-sync, so it may avoid it. Adaptive v-sync turns v-sync off below your refresh rate, so it also avoids it.

They may want to add something like dynamic v-sync as an option, but this is not the solution AMD wants for frame metering. This would be an additional type of v-sync that may be useful. A lot of people do not want a FPS limiter or v-sync. Currently that is your only option.

 

[ICDP]

very good call. at the end of the day. it is all about performance vs price. unless you have a money tree in your back yard - then that does not apply to you.

titan performance/price ratio is probably as bad as 7970cf neutered if not worse.

Titan price/perf ratio is absolutely horrendous here in the UK. One Titan costs almost twice the price of two 7950s or GTX660Tis.

 

[bystander36]

I may have this wrong, but I was under the impression that nV uses frame limiters.

A FPS limiter stops you from ever going beyond a certain threshold. Nvidia's metering system is smarter than that, though not perfect either. It occasionally will lengthen frames that get out of sync. As you can see, SLI scales almost exactly the same as AMD on average, so it doesn't affect performance a lot.

The point I am making is that AMD does have a legitimate problem in crossfire when used without v-sync. FPS limiting is not the solution they will want if they want to look competitive in reviews. Metering is going to be added, as AMD has mentioned they plan to release a fix this summer.

If you use v-sync all the time, then don't worry about it, it is not a problem you deal with. If you do not use v-sync, this is a problem, even if FPS limiting can fix the stuttering, that comes at a cost.

 

[bystander36]

Now seriously people. Those who are going on about how RadeonPro is all you need. Do you really not want AMD to add a frame metering option so you don't have to limit FPS? I'll be happy when it is added.

 

[ICDP]

All v-syncs do a 60/30/60/30 type solution if v-sync is actually enabled. I'm not certain if dynamic v-sync is an actually v-sync, so it may avoid it. Adaptive v-sync turns v-sync off below your refresh rate, so it also avoids it.

They may want to add something like dynamic v-sync as an option, but this is not the solution AMD wants for frame metering. This would be an additional type of v-sync that may be useful. A lot of people do not want a FPS limiter or v-sync. Currently that is your only option.

Dynamic vsync does the same thing, it is a decent solution for me as it does eliminate most, but not all tearing.

Speaking for my own CF experience, I simply get cheesed off when most of the people who slag it off haven't even experienced it. Or if they did they were not prepared to deliberately use FPS caps etc and then proclaim CF totally worthless. I can understand that these tweaks are not for everyone and that AMD have a massive amount of work ahead to improve the CF out of box experience. If someone wants absolute performance with no compromises then Crossfire is not a viable solution right now. But that is not the same as always broken as many imply, because a lot of us are happy to use FPS limiters or vsync. This of course does not mean AMD should be let off the hook, so to speak.

 

[wand3r3r]

Now seriously people. Those who are going on about how RadeonPro is all you need. Do you really not want AMD to add a frame metering option so you don't have to limit FPS? I'll be happy when it is added.

Speaking for my own CF experience, I simply get cheesed off when most of the people who slag it off haven't even experienced it. Or if they did they were not prepared to deliberately use FPS caps etc. I can understand that these tweaks are not for everyone and that AMD have a massive amount of work ahead to improve the CF out of box experience. If someone wants absolute performance with no compromises then Crossfire is not a viable solution right now. But that is not the same as always broken as many infer.

I believe that is the point (about the defensive nature). Most people loudly whining haven't even used crossfire.

Anyways it'll be interesting to see what AMD comes up with. All the attention on crossfire and frametimes has finally forced them to take it seriously. I'm not holding my breath but I suspect they'll be able to pull a workable solution out.

 

[BallaTheFeared]

Ironic coming from someone who has defended Titan performance given it gives far worse performance gain for even higher % cost.

I like how you separated those 2 improvements like they are mutually exclusive. I didn't get one or the other, I got both at the same time. The ultra shadows bring the average FPS down to ~35 FPS when enabled at that resolution and the game becomes unplayable in some places as it drops into low/mid 20s. So the performance improvement is 71% when taken at apples to apples settings.

Still not worth it?

I defending Titan? That's news to me, but to be honest I'd rather have a 1K titan that some gimped poor mans version called CF. There is no positive comparison you can make for CF vs single, not one, at least none that make any sense.

Normally people buy one, then add another later when they have the additional funds. What you did, do, or have done had nothing to do with what I was posting. I wasn't attempting to attack you, I was merely commenting on the situation as most would experience it.

Not in the least.

 

[bystander36]

I believe that is the point (about the defensive nature). Most people loudly whining haven't even used crossfire.

Anyways it'll be interesting to see what AMD comes up with. All the attention on crossfire and frametimes has finally forced them to take it seriously. I'm not holding my breath but I suspect they'll be able to pull a workable solution out.

I have done 5870's and 6950's in crossfire, and I was forced to use v-sync. I was mostly happy, but to be honest, I did have to fuss with drivers a lot more then when I switched the 6950's for 470's (I got a 120hz monitor, and fell in love with 3D, forcing the switch). I hated the 470's due to noise, but the other than the noise, it did just work out of the box with less fuss. I got rid of the 470's when the 680's were first out, as they were reasonably priced for the time and far quieter. I'm quite happy with the 680's, the first time I've been content with my graphics setup since my 9800 gt. Though I was really happy with the 9800 pro, I did not choose to stick with ATI the next card due to driver annoyances.

From my experience, ATI/AMD has always offered good performance for the cost, but it always came at a cost. There is always something that just doesn't work right. I always give them another chance, but eventually get annoyed.