• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Radeon 7870/7850 Reviews Are up!

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
No?

So the 7870 is faster per clock, at 1000Mhz while being slower in benches?

I'm confused as to what is actually misleading?


772Mhz 580 is faster than 1000MHz 7870, adding 200MHz to the 7870 would not require the 580 to run at 972 to maintain the same performance advantage.

You can see the lower performance per Mhz at stock. I repeat. What's the point of doing such misleading comparison?

Both cards at max OC have almost the same performance but you can push a 2500K above 3.3 Ghz. Again, your comparison is broken.
 
Pretty sure the 580 has better per core, per clock performance.

Think of it like bulldozer vs SB... I guess. A i5-2500k @ 3.3 GHz vs a Bulldozer cpu at 4.2GHz is slightly faster, it doesn't need to continue to match clock speeds at this level to continue being faster because it has higher IPC, meaning it gains more for increased MHz than bulldozer does. 7870 vs 580 works exactly the same way.

so if GTX 580 is so efficient then explain me why its need alot more power than HD 7870 ??? and after all oced HD7870 still consume less power than GTX 580 so your logic is failing hard, and btw GTX 580 shader clocked much more higher than HD 7870 so your logic is fail again.
 

No, it's all about % OC.

If gtx580s get 30% OC on air with as often, then you may say gtx580 can keep up with the OCs, it would still be slower OC vs OC unless say, it had 40% OCs. Then factor in the 1.5gb vram vs 2gb vram difference, it may not affect everyone, but when it does, performance tanks. Then factor in the OC power draw, 20W more for 7870s and more than 100W extra for the gtx580, already drawing >100W more to start with.

Next gen mid-range IS NOT MEANT as an upgrade path for users who own previous gen high-end. Never has been. It's meant for ppl who are still stuck with mid-range, 58xx, gtx460/560. When all factors considered, these cards are a great upgrade. $50 too expensive sure.

But have you ppl factored in the ~180W OC power draw vs >300W OC power draw of gtx560s? That difference in a lot of countries amount to more than the "$50 too expensive" over the 2-3yrs lifetime of the cards, and certainly in noticeable heat output.

It will come down to how gk104 stack up perf/$ and perf/w, as these cards taken as an entire package with their amazing OC and low power use blow away any reason to buy old gen products, period.
 
so if GTX 580 is so efficient then explain me why its need alot more power than HD 7870 ??? and after all oced HD7870 still consume less power than GTX 580 so your logic is failing hard, and btw GTX 580 shader clocked much more higher than HD 7870 so your logic is fail again.

Because power/performance efficiency has nothing to do with what was being discussed?

So my logic is still good there.

No, it's all about % OC.

Percent is not what I argued, I argued clock speeds. 20% of 772 is 926 which is much more reasonable for a 580 on air. Which would be 1200Mhz for the 7870 to do the same, we'll have to see if overclocking really is a boon this time around for the 7870, considering it's already slower I don't see it getting back that performance and making up even more ground to surpass the 580 OC.
 
Because power/performance efficiency has nothing to do with what was being discussed?

So my logic is still good there.



Percent is not what I argued, I argued clock speeds. 20% of 772 is 926 which is much more reasonable for a 580 on air. Which would be 1200Mhz for the 7870 to do the same, we'll have to see if overclocking really is a boon this time around for the 7870, considering it's already slower I don't see it getting back that performance and making up even more ground to surpass the 580 OC.


but shader clock in GTX 580 is much higher than HD 7870 so..................
 
I have to agree with the Price/Performance.

TDP is not uninteresting it's just secondary.


I would MUCH rather have a 200w 7870 that trashes GTX 580.

Same with any other southern islands.


I hate that fact marketing/production doesn't stick to standards and compete on it.


Intel did the same with IB - lowered tdp to sacrifice performance increases on stock to purvey the way for Haswell to show bigger increase to SB/IB.


Stick to a preset, and push the power.
We'll find a way to use it 😉.


That being said, if NVidia goes to price war, we all win.
If they don't, this is disappointing imho to previous generation owners.

Bring on the damn competition, even if Kepler doesn't live totally up to hype it will help the market situation.
 
Percent is not what I argued, I argued clock speeds. 20% of 772 is 926 which is much more reasonable for a 580 on air. Which would be 1200Mhz for the 7870 to do the same, we'll have to see if overclocking really is a boon this time around for the 7870, considering it's already slower I don't see it getting back that performance and making up even more ground to surpass the 580 OC.

I don't even know why you argue based on clock speed as architecture wise both are completely different, obviously the biggest difference is in the hotclocks on NV gpus.

926mhz on gtx580 is stretching it, with a lot of extra voltage and at this OC, its prolly sucking down ~400W.

A 7870 running at 1.196 ghz (near 1.2ghz) at stock volts adds 20W to its ~120W consumption.

Which is why when it comes down to it, final OC vs OC, the 7870 ends up faster and draws a tiny amount of power in comparison.

The same goes for the 7850, its stock performance is meh, but a 33% OC on stock volts, seriously, this needs to be factored to the final perf/$ and perf/w to which many people based their buying dollars on.
 
You know what.

It seems like AMD thinks it' has NVidia all figured out.

And....

NVidia thinks it has AMD all figured out.



One of em, is going to get severely hurt in the next month or so.

Wonder who :O
 
I have to agree with the Price/Performance.

TDP is not uninteresting it's just secondary.

You sound like a spoilt American.

If you think its a bit expensive based on perf/$...

Have you factored in the ~120W difference over 2 yrs? Even at a cheap rate of 12 US cents a kwh, a good avg 6 hrs a day of use, over 1 year that equates to ~$31 USD. Over 2 yrs, thats $62.

Now, if you live in many places around the world where electricity costs twice or three times that much. You can do the maths.
 
How is it faster? I'm just not sure where you're coming up with that.

Power wasn't part of the discussion, you can stop using it as a crutch.

While generaly the GTX 580 has superior IPC to the 7870 (lets ignore the whole hotclocking thing for this), both on stock voltage and when overvolted, the 7870 has far more headroom. I kinda doubt the higher IPC can bridge the 30%+ clock gap.
 
So when will they be available and how many can you run i crossfire? 2? 3? Fiance's 5770 x fire just died last night and need a replacement.
 
While generaly the GTX 580 has superior IPC to the 7870 (lets ignore the whole hotclocking thing for this), both on stock voltage and when overvolted, the 7870 has far more headroom. I kinda doubt the higher IPC can bridge the 30%+ clock gap.

perfrel_1920.gif

The 7870 would have to overclock 30% if the GTX 580 overclocked 20%. They'd be about equal on performance then.
 
Haserat said:
The 7870 would have to overclock 30% if the GTX 580 overclocked 20%. They'd be about equal on performance then.

You can see a 30% OC on a 7870 with volt mod in the techreport review. No other review site has OCed the card with volt mods.

Aside that I don't know how the hell are you measuring the IPC in a complete different architecture with different transistor, units, bandwidth and memory count.
 
Last edited:
7870 Vs. 6970 is an interesting comparison, thanks to the very close number of transistors. Clocks are comparable too, with 7870 having higher core clocks, and 6970 higher memory. From the looks of things gaming perf/transistor has at worst stayed flat with VLIW4 and at best improved modestly.
 
Back
Top