wccftechAMD Pirate Islands : R9 300 Series Alleged Specifications Detailed

[Silverforce11]

It's max o/c vs max o/c. You're just looking for possible discrepancies and they aren't very compelling. Doesn't matter what the boost is. Max o/c is max o/c. How the cards perform respectively at their max o/c's is what you get.

Yes YMMV but this applies to both cards.

Sadly I haven't seen a reputable site test both latest drivers max OC vs max OC single cards, multi cards and 4K. Do you have some new data that can enlighten us all?

Because what I brought up is sound logic, current data with these latest drivers, we have very close clock for clock performance between the 780ti and R290X.

How much headroom advantage do you want to give to the 780ti on air, max 100mhz? 200mhz?.. I'm a reasonable guy, so I think 780ti potential of 1.35ghz boost is generous, with good custom cards and a tweak bios, most of them should hit 1.3ghz. Sounds fair to you?

 

[TreVader]

It's max o/c vs max o/c. You're just looking for possible discrepancies and they aren't very compelling. Doesn't matter what the boost is. Max o/c is max o/c. How the cards perform respectively at their max o/c's is what you get.

Yes YMMV but this applies to both cards.

What is "max oc"? On air? Using only reference? Stock voltage? Water? LN2?
Because i'd be more than happy to compare a voltage unlocked lighting or Cryovenom to a reference 780ti or 780. It's too bad nvidia made such a poor OCing card in comparison to the cryovenom.




Every "max oc" is different so that's a totally arbitrary measurement. You can only compare stock (stock aftermarket, etc) and since 95% of users never OC it's actually almost totally irrelevant to anybody but people on forums like this.

 

[Gloomy]

that 780Ti Ghz goes over 1200Mhz

 

[Silverforce11]

that 780Ti Ghz goes over 1200Mhz

I'm sure Russian knows this, many custom 780ti boost way above 1ghz. How much headroom do they have left?

Russian, since you linked that prior bench, you should read this carefully: The 780ti Ghz ed is boosting to 1.22ghz and sustaining 1.2ghz clocks by default. When manually OC and pushed with extra voltage, it managed ~58mhz extra, with minimal performance gains:
http://www.hardwarecanucks.com/foru...gigabyte-gtx-780-ti-ghz-edition-review-8.html

Here's another of those custom cards, high boost out of box. ~1.2ghz OC:
http://www.hardwarecanucks.com/foru...ws/64269-evga-gtx-780-ti-sc-acx-review-8.html

And here's a big list with a lot of 780ti models and their OCs:
http://www.techpowerup.com/reviews/EVGA/GTX_780_Ti_Classified/27.html
Most of them hit a boost of 1.15ghz to 1.25ghz OC.

Looking at the HC results, with older drivers, it seems there is only a tiny advantage to the 780ti clock for clock. Comparisons with recent drivers have it neck and neck at the same clocks.

Edit: I am being very generous when I say 780ti could get 1.3ghz easily, there's no data from review sites to back that up with any high frequency. The average so far from reading a lot of reviews, seem to be around 1.2ghz boost. With some reaching 1.25ghz while some are worse.

 

[Keysplayr]

Semantics are fun.

 

[Keysplayr]

What is "max oc"? On air? Using only reference? Stock voltage? Water? LN2?
Because i'd be more than happy to compare a voltage unlocked lighting or Cryovenom to a reference 780ti or 780. It's too bad nvidia made such a poor OCing card in comparison to the cryovenom.




Every "max oc" is different so that's a totally arbitrary measurement. You can only compare stock (stock aftermarket, etc) and since 95% of users never OC it's actually almost totally irrelevant to anybody but people on forums like this.

Ok, so the next time you use an o/c'd card to make a point, we can direct you to this post here. Much obliged.

 

[Gloomy]

That small headroom on the Ti is because of Nvidia's limit though, to be fair. It does have a lot of headroom left over. Probably.

 

[Silverforce11]

That small headroom on the Ti is because of Nvidia's limit though, to be fair. It does have a lot of headroom left over. Probably.

Yes to be fair, if you use modded bios you could get 1.3ghz regularly.

But maybe that's just one step too far for most consumers who buy these cards. I don't know many who regularly flash modded bioses on very expensive cards.

 

[Keysplayr]

Yes to be fair, if you use modded bios you could get 1.3ghz regularly.

But maybe that's just one step too far for most consumers who buy these cards. I don't know many who regularly flash modded bioses on very expensive cards.

So you're excluding a thing that would give the edge. I see. Fair as all hades.

 

[Silverforce11]

No I didn't exclude it, in fact several posts ago and repeatedly I have mentioned that it has an advantage with modded bios. You could squeeze an extra 5-10% headroom from that. I just noted, flashing with modded bioses is not something people do and if they do enjoy that, sure, then its an edge.

Otherwise for the regular joe enthusiast, no its no edge. I enjoy having warranty on my cards also, I'm sure most users enjoy that too.

ps. I'm very fair. I do use both NV and AMD gpus, recently with a 670. Because i am fair, I never touch an AMD FX CPU because its trash. Their GPUs are highly competitive though, currently extremely competitive, something they haven't managed for a long long time.

 

[raghu78]

There is, in % terms. 7970 came clocked at 925mhz and many chips hit 1175-1200mhz, the Lightning and Matrix could do 1250-1300mhz. In % terms, 7970 kills R9 290X.

The HD 7970 came at 925 mhz and then a HD 7970 Ghz SKU came at 1050 mhz. R9 290X came at 1000 Mhz. Both Hawaii and Tahiti on golden samples overclocked to 1300 - 1325 Mhz. So i don't understand how Tahiti kills Hawaii in % overclock. even with the first gen HD 7970 the difference is just 75 mhz.

See, 7970 was about 8-10% behind 680 but when both were overclocked, it could match an overclocked 680. R9 290X cannot match an overclocked 780Ti. We look at base performance and then look at % overclock because enthusiasts do overclock. While R9 290X is close to a stock 780Ti, once the 780Ti is overclocked, the difference grows in favour of NV, not shrinks.

R9 290X is right behind GTX 780 Ti on a clock for clock basis. roughly 5 - 7% slower on avg across many games at 1440p.

Gigabyte GTX 780 Ghz (1258 mhz)
http://www.hardwarecanucks.com/foru...gigabyte-gtx-780-ti-ghz-edition-review-8.html

PCS+ R9 290X (1219 Mhz)
http://www.hardwarecanucks.com/foru...5702-powercolor-r9-290x-pcs-4gb-review-8.html

Agreed the GTX 780 Ti is clocking higher on average and max by about 100 Mhz. But still GTX 780 Ti can only claim single monitor upto 1600p and single GPU leadership. Multi monitor and ultra high res single monitor (4k) with multi GPU go to AMD clearly. For enthusiasts who want the best 4K or multi monitor gaming performance look no further than the R9 290X CF watercooled.

With Maxwell bringing 35% increase in IPC, and it'll be more likely to win the overclocking % battle as has the 285/480/580/780Ti before it.

Again you are jumping to conclusions without having a clue as to what AMD has for their next gen GPUs. You were so vocal last year saying AMD cannot beat GTX 780 and match GTX Titan with a sub 500 sq mm die. Guess what, AMD proved you wrong. they did it at 438 sq mm. It required a fully enabled higher clocked GK110 to take back the crown.

R9 290X at 1.22 Ghz can't even beat a 780Ti Ghz edition that's not overclocked.
http://www.hardwarecanucks.com/foru...5702-powercolor-r9-290x-pcs-4gb-review-8.html

Yeah when the GTX 780 Ti is clocked at 1.2 Ghz out of the box its going to do well. anyway the GTX 780 Ti Ghz in the hwc review didn't gain much when overclocked over its stock speed. see the links above.

But Kepler is an old architecture, tracing its roots back to Fermi. It's actually more amazing that Kepler can even keep up with GCN. Maxwell is in turn NV's "GCN" - it's their new architecture in a while.

http://www.anandtech.com/show/2116

"We always get very excited when we see a new GPU architecture come down the pipe from ATI or NVIDIA. For the past few years, we've really just been seeing reworked versions of old parts. NV40 evolved from NV30, G70 was just a step up from NV40, and the same is true with ATI as well. Fundamentally, not much has changed since the introduction of DX9 class hardware. But today, G80 ushers in a new class of GPU architecture that truly surpasses everything currently on the market."

http://www.anandtech.com/show/2849/3

"NVIDIA keeps referring to Fermi as a brand new architecture, while calling GT200 (and RV870) bigger versions of their predecessors with a few added features."

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/2

"For NVIDIA, Kepler is the embodiment of that concept. Kepler brings with it some very important architectural changes compared to Fermi, but at the same time it’s still undeniably Fermi. From a high level overview Kepler is identical to Fermi: it’s still organized into CUDA cores, SMs, and GPCs, and how warps are executed has not significantly changed. Nor for that matter has the rendering side significantly changed, with rendering still being handled in a distributed fashion through raster engines, polymorph engines, and of course the ROPs. The fact that NVIDIA has chosen to draw up Kepler like Fermi is no accident or coincidence; at the end of the day Kepler is the next generation of Fermi, tweaked and distilled to improve on Fermi’s strengths while correcting its weaknesses."


http://www.anandtech.com/show/7764/the-nvidia-geforce-gtx-750-ti-and-gtx-750-review-maxwell/2

"In short, Maxwell only offers a handful of new features compared to Kepler. Kepler itself was a natural evolution of Fermi, further building on NVIDIA’s SM design and Direct3D 11 functionality. Maxwell in turn is a smaller evolution yet."

So there you have it. Nvidia's legendary G80 Tesla which brought unified shaders and DX 10 support and and GF100 which brought DX11 support and a highly scalable gpu architecture are the two major architectures after NV30 and DX9 way back in 2003. Incidentally Nvidia's major architectures have coincided with a new Microsoft Directx API.

So no Maxwell is not a major new architecture. That would be Pascal or Volta which recently got pushed back on Nvidia's roadmaps.

On the contrary GCN is AMD's Fermi. It is AMD's clean sheet start to build a from the ground up DX11 scalable graphics and compute architecture. Its driving the next gen consoles and has a long life ahead. AMD can definitely innovate and improve performance and efficiency as Nvidia did with Kepler and Maxwell.

http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review

"GCN would be AMD’s Fermi moment, where AMD got serious about GPU computing and finally built an architecture that would serve as both a graphics workhorse and a computing workhorse"

The performance/watt of 750Ti is spectacular and that's only on 28nm.....once Maxwell moves to 20nm, it'll be even better. I hope for the PC gaming landscape that AMD can compete for flagship performance next round but I am not too optimistic. Last time NV introduced a new GPU architecture with a node shrink, it blew AMD away - 8800GTX. Maxwell = brand new architecture (newer than GCN) and a node shrink. Maxwell is already very efficient at script mining with 750Ti. It should fix Kepler's compute weakness per mm2. GPUs are now primarily power constrained. Since Maxwell architecture was designed for mobile first, once you scale such an efficient design, it will be very powerful on the desktop inside a 250W TDP. I doubt GCN 2.0 was designed as efficiently.

Maxwell is efficient. Thats proven. Now we need to see AMD's response. Be patient. I am sure they will have something competitive just as R9 290X was to GTX 780 Ti. R9 290X and GTX 780 Ti are competitors. When GTX 780 Ti SLI cannot decisively beat R9 290X CF it cannot claim all round GPU leadership. It does have the fastest single GPU crown for single monitor upto 1600p.

NV never had a flagship with 294mm2 die size since 8800GTX. That in itself was proof that GTX680 was never a real flagship. It simply took the place of NV's flagship due to market landscape at the time. Kepler is hands down the biggest leap in inter-generational performance from NV since 7900GTX to 8800GTX. 780Ti is at least 2x faster than 580. It wasn't that 7970 was poor, but that Kepler ended up exceeding NV's expectations.

fair enough. But to give you a clue as to how good G80 aka 8800 GTX was it provided all the perf gain at the same 90 nm process node as 7900 GTX. Simply outstanding. I doubt Nvidia can repeat a feat like that. Unified shaders is probably the number one most important innovation in the GPU industry in the last 2 decades. Fermi to Kepler involved a huge process node transition from 40G (polysilicon) to 28HP (high k metal gates).

On paper, if mid-range Maxwell and R9 380X beat 780Ti and they are priced at $499-549, they'll sell like hot cakes since 780Ti is now at $650-750; and that's why AMD and NV can do a 1-2 wave launch of 20nm like they did 7970-> R9 290X and 680 -> 780Ti. Sucks for us gamers.

I don't expect Nvidia or AMD to sell a 20 nm GPU which beats GTX 780 Ti by a significant margin for less than USD 700. Both are going to fleece the consumers. 20nm is costly. dual patterning adds wafer processing steps and cost. cost per transistor is not seeing a major reduction as in previous generations. So expect higher prices.

http://www.anandtech.com/show/7739/arm-cortex-a17

 

[RussianSensation]

I'm not going to bother comparing older driver reviews, it seems they are no longer correct since the gains are significant for both NV and AMD. Even if you get a 10% OC advantage, the difference overall is not significant anymore single card, especially with Mantle and BF4 MP results thrown into the data.

So you are saying March 2014 reviews are suddenly irrelevant? I see here a great PowerColor PCS+ 290x and it gets its ass handed by a Gigabyte GTX780Ti Ghz. It's not even close.

http://www.hardwarecanucks.com/foru...5702-powercolor-r9-290x-pcs-4gb-review-3.html

Then SLI vs CF and 4K results, 780ti is blown away, its actually quite a large defeat. Quite ridiculous in BF4 MP and a solid lead even in Crysis 3 and FC3. [/url]

OK but that's 1 benchmark. And honestly 4K is mostly marketing right now. Most people would rather game at max settings at 1440/1600P than at Low/Medium settings on 4K. You think games will be at BF4 level of graphics for the next 2-3 years? Once games like Witcher 3, the Division hit, they'll wipe out any advantage Maxwell. We've seen the same thing over the last 10 years. First GPUs outpace PC games and then we get a new wave of next generation titles that mop the floor with our GPUs. Right now it seems that 4K is within reach but it's only because 99% of PC games have outdated game engines and last gen graphics. Wait until next generation games start arriving and 4K will again turn into a pipe-dream without going with 2-3 flagship Maxwells.

If you fret about max OC vs max OC, I don't see many reputable site doing a R290X @ 1.2ghz vs 780ti @ 1.3ghz bench off, certainly not with recent drivers.

HWC has the 1.05Ghz 290X PCS+ vs. 1.15Ghz 780Ti and it loses badly. You add 150 mhz to 290X and 150mhz to 780Ti, the picture won't change much. Also, for most gamers, 1080P and 1440/1600P still rules and there 780Ti OC beats 290X OC even more.

Again, that's a big GK110 die versus the smaller Hawaii die. Lets assume Maxwell has 35% IPC gains and GCN 2.0 has ZERO gains, AMD could still catch up with a large die if they wanted to go for it.

How? Think about it a GTX780Ti OC beats R9 290X OC without any problems and that's on older Kepler tech. Let's assume NV and AMD will both benefit equally from a 20nm node shrink (30-40% boost), NV brings a 35% increase in IPC. AMD needs to use some more exotic technologies or go far beyond 438mm2 to overcome that 35% deficit.

I'd like to be proven wrong by AMD but for the next round I am even more inclined to believe NV will win: (1) NV has won every round so far since 8800GTX when it came to single flagship GPU performance. Why would it suddenly change? (2) This round NV is bringing a brand new architecture while AMD is working off a 2-year-old GCN. Advantage: NV.

ps. I disagree that GPUs are power constrained, that only occur for mobiles and low end stuff. High end stuff, especially since you love to talk about OC, I've seen users on these forums with modded Bios and a 780 pull 500W.

You are looking at it from an overclocker's point of view. I am talking about from an engineering point of view. NV will try to make the fastest possible GPU within a 250W TDP. To do that, they have revamped their entire GPU strategy from the ground-up to make the smallest building block of Maxwell as efficient as possible by targeting mobile. With Maxwell, a desktop GeForce is just a mobile architecture expanded to the desktop. The efficiency should permeate from top to bottom. Considering AMD couldn't win with the newer GCN, now with Maxwell, I am even more convinced NV will sweep the next round for top flagship GPU performance. NV also has GameWorks and their driver team continues to make improvements in AMD GE titles. OTOH, in games where NV dominates (AC3, FC3, C3, Project CARS), AMD has hardly made an impact. I'll give AMD props for BF4's Mantle performance but it's only 1 game.

 

[Silverforce11]

RS, if you think that HC review with the 780ti Ghz ed with its 1.2ghz boost clock winning by a small margin in Crysis 3 and even less in TR at 1440p is ass handing, you are delusional.

It's absurd you think that actually, given in their own tests, that card had next to no OC room left and you are making it sound as if its competing at stock. Yeah, it's "stock" is already a massive OC by any other standards. Go look around on most reviews, 780ti boost between 1.15 to 1.25ghz OC.

4K is no gimmick, CF R290/X has the power to run modern games at 4K at near 60 fps even with MSAA. And its not just 1 bench/game as you claimed:

I do agree previous gen NV won on raw performance with their big die approach, there's no point debating that.

The fact is currently with the latest drivers, AMD's smaller die is matching NV's big die, and at 4K & multi-GPU, it kills it. Importantly it's ONLY GCN 1.1 with a focus on 1/2 DP compute. Yet you see doom and gloom for the future vs Maxwell.. I'm not seeing it. AMD has never managed to come so close with their smaller die to NV's big die for a long time until now.

ps. With any major driver update, older data is irrelevant for current comparisons, obviously. Else we all think the 680 is still faster than the 7970/Ghz. We only have to look at Rome 2 TW to see how irrelevant it would be if it was tested with an older NV driver. Utterly broken. NV fixed it and managed "up to 79%" (or something fantastic!) perf improvement.

 

[TreVader]

Ok, so the next time you use an o/c'd card to make a point, we can direct you to this post here. Much obliged.

That's fine. AMD already has aftermarket cards that run at 1150 stock (cryovenom). It's perfectly reasonable to compare two aftermarket stock-OC cards.



AMD already has stock-oc'd cards that dominate nvidia if you don't OC them. The Cryovenom 290 destroys the Ghz edition 780, I think it even wins in Heaven.

 

[raghu78]

How? Think about it a GTX780Ti OC beats R9 290X OC without any problems and that's on older Kepler tech. Let's assume NV and AMD will both benefit equally from a 20nm node shrink (30-40% boost), NV brings a 35% increase in IPC. AMD needs to use some more exotic technologies or go far beyond 438mm2 to overcome that 35% deficit.

I'd like to be proven wrong by AMD but for the next round I am even more inclined to believe NV will win: (1) NV has won every round so far since 8800GTX when it came to single flagship GPU performance. Why would it suddenly change? (2) This round NV is bringing a brand new architecture while AMD is working off a 2-year-old GCN. Advantage: NV.

Sorry read my post above. Tesla G80 and Fermi GF100 are the 2 major architectures for Nvidia in the last 8 years. Nvidia worked off a 2 year old Fermi with Kepler. It worked out well. They worked off a 2 year old Kepler with Maxwell. Worked out very well. Why do you think AMD cannot innovate with GCN. Give solid reasoning.

Considering AMD couldn't win with the newer GCN, now with Maxwell, I am even more convinced NV will sweep the next round for top flagship GPU performance. NV also has GameWorks and their driver team continues to make improvements in AMD GE titles. OTOH, in games where NV dominates (AC3, FC3, C3, Project CARS), AMD has hardly made an impact. I'll give AMD props for BF4's Mantle performance but it's only 1 game.

How do you say AMD couldn't win with newer GCN. The way I see it at the high end both Nvidia and AMD are tied. Single monitor 1600p - GTX 780 Ti . 4K and multi monitor which needs multi GPU for playing at high quality settings goes to R9 290X CF. Don't you think XDMA has a role to play in that. XDMA is also a part of GCN tech. the forward looking 64 ROP design also helps. Why didn't you give credit to AMD for both of that.

btw Farcry 3 and Crysis 3 run very well on AMD R9 290X CF. How do you dominate a game when you lose at the highest resolutions and highest settings.

Farcry 3

http://hardocp.com/article/2014/04/08/amd_radeon_r9_295x2_video_card_review/5#.U0lftaKfZ8E

http://www.hardwarecanucks.com/foru...md-radeon-r9-295x2-performance-review-11.html

Crysis 3

http://hardocp.com/article/2014/04/08/amd_radeon_r9_295x2_video_card_review/4#.U0lkx6KfZ8E

http://www.pcper.com/reviews/Graphics-Cards/AMD-Radeon-R9-295X2-8GB-Graphics-Card-Review/Crysis-3

http://www.hardwarecanucks.com/foru...md-radeon-r9-295x2-performance-review-10.html

http://www.techpowerup.com/reviews/AMD/R9_295_X2/13.html

 

[AnandThenMan]

Again you are jumping to conclusions without having a clue as to what AMD has for their next gen GPUs.

100% agreed with this. RS is usually very level headed but this time not sure what is going on. My hunch, GCN 2.0 will be a very strong chip. I have a feeling AMD was stretched quite thin dealing with getting Mantle off the ground and console GPUs. Not the case anymore.

 

[Leadbox]

100% agreed with this. RS is usually very level headed but this time not sure what is going on. My hunch, GCN 2.0 will be a very strong chip. I have a feeling AMD was stretched quite thin dealing with getting Mantle off the ground and console GPUs. Not the case anymore.

Maybe he thinks that for 20nm AMD are just going to shrink and add more SPs to a Hawaii type die?
Bonnaire and Hawaii are derivatives of the work AMD and Sony done for the PS4 apu yes, no?

 

[raghu78]

100% agreed with this. RS is usually very level headed but this time not sure what is going on. My hunch, GCN 2.0 will be a very strong chip. I have a feeling AMD was stretched quite thin dealing with getting Mantle off the ground and console GPUs. Not the case anymore.

Just a bit of history. Nvidia and AMD have had 2 major architectures in the last 9 years. AMD's R600 unified shader GPU with a VLIW 5 architecture lasted for 6 years from the original Xbox GPU which launched in Nov 2005 to the HD 6970 (End Of Life) in Dec 2011. AMD started off poorly with HD 2900XT but rapidly turned it around with the RV770 aka HD 4870, tweaked it and added DX11 support with RV870 aka HD 5870 and then finally tweaked it once gain with VLIW4 for HD 6970. GCN was the second major achitecture. For Nvidia it was the G80 Tesla and Fermi GF100 architectures.

GCN is the foundation for AMD GPUs and both the major next gen consoles PS4 / XB1. Its going to be here for a long while. I am guessing atleast 6 years. There will be improvements to the architecture to improve efficiency and support the newer APIs like DX12, but the foundation would still be unmistakeably GCN. Nvidia has shown with Fermi,Kepler and Maxwell that a strong foundation can last a long while. GCN is very much that strong foundation. :thumbsup:

 

[raghu78]

Maybe he thinks that for 20nm AMD are just going to shrink and add more SPs to a Hawaii type die?
Bonnaire and Hawaii are derivatives of the work AMD and Sony done for the PS4 apu yes, no?

Hawaii uses the same 8 ACE compute front end adopted by the PS4. A lot of the work done for PS4 has been utilized in Hawaii.

http://www.slideshare.net/DevCentralAMD/gs4106-the-amd-gcn-architecture-a-crash-course-by-layla-mah

(slide 54)

The audio DSP hardware named TruAudio on R9 290X is also present in the PS4.

http://www.anandtech.com/show/7513/ps4-spec-update-audio-dsp-is-based-on-amds-trueaudio

 

[VulgarDisplay]

So you are saying March 2014 reviews are suddenly irrelevant? I see here a great PowerColor PCS+ 290x and it gets its ass handed by a Gigabyte GTX780Ti Ghz. It's not even close.

http://www.hardwarecanucks.com/foru...5702-powercolor-r9-290x-pcs-4gb-review-3.html

What drivers is that review using? I can't find it listed anywhere. If they are using anything 14.x without the pt1 bios their cards are going to throttle when over clocked.

 

[Atreidin]

I don't see why GCN 2.0 can't have nice IPC gains, it isn't like AMD's engineers are incapable of finding optimizations like Nvidia's engineers. I'm confused with the assumption that it won't.

 

[Elfear]

According to HwBot, the average OC for the 780Ti and 290X are as follows:

780Ti (3201 submissions)
Air - 1187/1940
Water - 1346/2188

290X (1649 submissions)
Air - 1159/1910
Water - 1238/1637

The memory clocks obviously need to be taken with a grain of salt (probably a glitch in how their software reads the submission results) but the core clocks look pretty even on air and ~100Mhz higher for the Ti on water.

It doesn't appear from their data that the Ti has a large overclocking advantage. Some of the top-end aftermarket cards will hit higher on air but I imagine the same is true of the top-end 290Xs. I know my lowly reference 290s will hit 1275 and 1240Mhz on water (benching only).

As to the 390X I hope it brings some big improvements in the IPC department. It would be great if PI and Maxwell were neck and neck.

 

[Enigmoid]

According to HwBot, the average OC for the 780Ti and 290X are as follows:

780Ti (3201 submissions)
Air - 1187/1940
Water - 1346/2188

290X (1649 submissions)
Air - 1159/1910
Water - 1238/1637

The memory clocks obviously need to be taken with a grain of salt (probably a glitch in how their software reads the submission results) but the core clocks look pretty even on air and ~100Mhz higher for the Ti on water.

It doesn't appear from their data that the Ti has a large overclocking advantage. Some of the top-end aftermarket cards will hit higher on air but I imagine the same is true of the top-end 290Xs. I know my lowly reference 290s will hit 1275 and 1240Mhz on water (benching only).

As to the 390X I hope it brings some big improvements in the IPC department. It would be great if PI and Maxwell were neck and neck.

Are those the actual clocks? Is there a 'boost' in effect?

 

[el etro]

Are those the actual clocks? Is there a 'boost' in effect?

Yes, but the boosting on clocks depend on many Driver+Afterburner conditions and settings. At least most of this 780Ti cards will boost 40-50Mhz.

 

[el etro]

I think the way Kepler and Maxwell are Fermi derivatives, the next AMD architectures will be derivatives of GCN1.0 basic aspects.


The greatest gains in Hawaii arch is in compute workloads(see HD 7790 review for more info).