Monday, April 22nd 2013

AMD Radeon HD 7990 Clock Speeds and Core Config Confirmed, Tested

Ahead of its April 24 launch, AMD board vendors has been distributing marketing materials to their retail partners. One such retailer in Japan revealed the flagship graphics card's specifications sheet, revealing details such as clock speeds and GPU core configuration.

To begin with, AMD isn't compromising much on clock speeds on the HD 7990 "Malta," in an effort to lower power draw. The card features GPU core clock speed of 1000 MHz, which puts it above the single-GPU Radeon HD 7970, but not much lower than the Radeon HD 7970 GHz Edition, with its 1050 MHz. The memory is clocked at 6.00 GHz, on par with the HD 7970 GHz Edition, which yields a cumulative memory bandwidth of 576 GB/s.
Moving on to the core configuration, the Radeon HD 7990 "Malta" doesn't disable anything on the "Tahiti" silicon. It uses a pair of "Tahiti" chips, with 2048 stream processors each, 128 TMUs each, 32 ROPs each, and 384-bit GDDR5 memory interfaces, each. The card features a total of 6 GB of memory, 3 GB per GPU.

In another slide, AMD gave retailers performance numbers for the Radeon HD 7990 "Malta" from its own testing. The HD 7990 is pitted against GeForce GTX 690 and GeForce GTX TITAN, and the three cards were put through 3DMark "Fire Strike" in performance and extreme presets. In performance preset, the Radeon HD 7990 scored 16 percent higher than GeForce GTX 690, and 29 percent higher than the GTX TITAN. In the extreme preset, it kept the trend up, scoring 15 percent higher than the GTX 690, and 31 percent higher than the GTX TITAN.
Source: Hermitage Akihabara
Add your own comment

62 Comments on AMD Radeon HD 7990 Clock Speeds and Core Config Confirmed, Tested

#1
Mathragh
Am I the only one that noticed that both texture and fillrate have more than doubled compared to the 7970 GHz Ed with the same amount of ROPs and a higher clockspeed?
If these are Tahiti chips like the ones used in the 7950 en 7970 series, they should have a lower texture and pixel fillrate at 1GHz.

I wonder what they did to the chip for this to be possible(two typos seem highly improbable to me, unless the table is fake ofc)

Edit: Humansmoke gave me the somewhat obvious answer :D
Posted on Reply
#2
the54thvoid
Super Intoxicated Moderator
What on earth have they done to the power configuration to allow such good output with 'speculated' reduced TDP? It'll be interesting to see the reviews.
Posted on Reply
#3
d1nky
im just picturing the asus version! cant wait to see what all these look like!
Posted on Reply
#4
HumanSmoke
MathraghAm I the only one that noticed that both texture and fillrate have more than doubled compared to the 7970 GHz Ed with the same amount of ROPs and a higher clockspeed?
If these are Tahiti chips like the ones used in the 7950 en 7970 series, they should have a lower texture and pixel fillrate at 1GHz.
That's easily solved. The 7970 GHz numbers in the picture are wrong- they are for the vanilla 7970.
Pixel fillrate is Core freq * ROP count
7970 is 925 x 32 = 29.6 GPixels/sec ( a GHz Edition would be 1000 x 32 = 32 GPixels/sec, or 33.6 GTex/sec under boost clock)
Likewise, texture fillrate is Core freq * TMUs
7970 is 925 x 128 = 118.4 GTexels/sec (GHz Edition would be 1000 x 128 = 128 GTex/sec, or 134.4 GTex/sec under boost clock)
Posted on Reply
#5
adulaamin
d1nkyim just picturing the asus version! cant wait to see what all these look like!
They're all gonna look the same just with different stickers/decals. :)
Posted on Reply
#6
d1nky
really?? that's shit!
Posted on Reply
#7
SIGSEGV
adulaaminThey're all gonna look the same just with different stickers/decals. :)
i'll take your info with a pinch of salt :)
as far as i know, amd are allowing their partner to custom its cooler and modified of its voltage to gain more clocks higher than average. :)
Posted on Reply
#8
qubit
Overclocked quantum bit
Don't forget that CrossFire runt frame problem. This means that these figures will be totally false and hence real world performance much lower until AMD fix this problem.
Posted on Reply
#9
Ghost
qubitDon't forget that CrossFire runt frame problem. This means that these figures will be totally false and hence real world performance much lower until AMD fix this problem.
There was an article posted here about how they fixed it. But it got pulled. It's been 1.5 years since release of 7970, so AMD should fix that problem anytime soon now. Perhaps during 2nd anniversary.
Posted on Reply
#10
qubit
Overclocked quantum bit
GhostThere was an article posted here about how they fixed it. But it got pulled. It's been 1.5 years since release of 7970, so AMD should fix that problem anytime soon now. Perhaps during 2nd anniversary.
Yes, it's a bit of a scandal how AMD have been happy to take your money for expensive cards while keeping quiet about a serious problem like this. :shadedshu Some enthusiasts who've been using CF for a long have been selling their cards and going nvidia over this issue.

I'm surprised an article like that got pulled from TPU as censorship isn't something they do too much of around here I'm glad to say.

Finally, frame rating is a complex test methodology to explain and understand the results from, so my article summarizes it, with PC Perspective's original in depth version linked to at the bottom.
Posted on Reply
#11
adulaamin
d1nkyreally?? that's shit!
SIGSEGVi'll take your info with a pinch of salt :)
as far as i know, amd are allowing their partner to custom its cooler and modified of its voltage to gain more clocks higher than average. :)
You guys are right. Asus is gonna come out with a uber custom cooled 7990 on launch date. Also MSI, Sapphire, Gigabyte. Reference coolers? Naaahhh :)
Posted on Reply
#12
the54thvoid
Super Intoxicated Moderator
qubitSome enthusiasts who've been using CF for a long have been selling their cards and going nvidia over this issue...
I'd hardly call myself an enthusiast but I will take the title you give gladly ;)

I can't speak for the rest of the folk that ditched crossfire for Titan but if Nvidia release a higher specced Titan as is rumoured (which i don't really believe) I'll not buy Nvidia again as long as AMD have competitive offerings. I left the AMD crossfire problems behind for a single super card. If Nvidia release something better a few months later they will not get my money again. AMD on the other hand have always said they'll bring out a 7990, just a bit late is all.

So if AMD finally iron out their software and release the 7990 with near perfect drivers and Nvidia release a higher specced Titan to boot, it is Nvidia who don't give shit about their customers. :shadedshu
Posted on Reply
#13
EarthDog
qubitYes, it's a bit of a scandal how AMD have been happy to take your money for expensive cards while keeping quiet about a serious problem like this. :shadedshu Some enthusiasts who've been using CF for a long have been selling their cards and going nvidia over this issue.

I'm surprised an article like that got pulled from TPU as censorship isn't something they do too much of around here I'm glad to say.

Finally, frame rating is a complex test methodology to explain and understand the results from, so my article summarizes it, with PC Perspective's original in depth version linked to at the bottom.
Its certainly not as big a problem as some here made/make it out to be...

First, AMD cards are cheaper at the high end...by a lot. They were not initially, but massive price drops have them squarely below NVIDIA whos prices haven't budged for the same exact performance.

Second, how is it a scandal when they never tested for it in the first place?

Please, TPU, stop making mountains out of molehills...the lemming just fall right off the cliff... :nutkick:
Posted on Reply
#14
qubit
Overclocked quantum bit
EarthDogIts certainly not as big a problem as some here made/make it out to be...

First, AMD cards are cheaper at the high end...by a lot. They were not initially, but massive price drops have them squarely below NVIDIA whos prices haven't budged for the same exact performance.

Second, how is it a scandal when they never tested for it in the first place?

Please, TPU, stop making mountains out of molehills...the lemming just fall right off the cliff... :nutkick:
Have you seen the frame rating test results at PC Per? It's a scandal all right, or at least naughty, lol.
Posted on Reply
#15
d1nky
adulaaminYou guys are right. Asus is gonna come out with a uber custom cooled 7990 on launch date. Also MSI, Sapphire, Gigabyte. Reference coolers? Naaahhh :)
that's what I like to hear!! I was thinking all standard and reference cards wouldn't sell as well as own branded remakes!

ooooh I cant wait to see some triple slot coolers and some supersized slap my bitch up coolers..... wish I was rich!! :p
Posted on Reply
#16
the54thvoid
Super Intoxicated Moderator
qubitHave you seen the frame rating test results at PC Per? It's a scandal all right, or at least naughty, lol.
You and Earthdog are about to enter an endless tango about driver issues and latency. I'd bailout now m8 before it's too late. ;)
Posted on Reply
#17
Xzibit
qubitHave you seen the frame rating test results at PC Per? It's a scandal all right, or at least naughty, lol.
Its the just about the same thing Nvidia was dealing with 1year ago and for all there efforts its still not solved on their side either after a year of trial and error with driver updates.
Posted on Reply
#18
EarthDog
qubitHave you seen the frame rating test results at PC Per? It's a scandal all right, or at least naughty, lol.
Yes, and No. Disagree. No scandal, no nothing like that at all. A lot of people have bought this hook line and sinker for WAY worse than it really is.

I digress.
Posted on Reply
#19
cadaveca
My name is Dave
EarthDogYes, and No. Disagree. No scandal, no nothing like that at all. A lot of people have bought this hook line and sinker for WAY worse than it really is.

I digress.
Says the dude who only runs a single 7970 or GTX690. :laugh:

Do keep in mind there are a huge number of Crossfire users here on TPU that DO agree it's an issue, and as big of an issue as related, so you'll have a hard time convincing those that use Crossfire there's no problems, and that's it's no big deal.

See, I can be a broken record, just like you.

:slap:

EarthDog: Not a big problem. No I don't run Crossfire.


Xzibit: Yeah, it's an issue....but look, NVidia has the same problems!


Also keep in mind, that W1zz does relate the same issues as being evident, in his reviews, before the PCPer stuff came out. Frankly, nobody really cares about PCPer, as all they showed is WHY we have issues in some instances, but doesn't provide a fix.



Why not just put your opinion on the subject in your signature, and not post? :roll: The less you post about it not being an issue, the less you'll find others posting that it is.
Posted on Reply
#20
Mathragh
I think its important for people to understand what the problems exactly are.

Most of the worst issues are rather specific, with some examples being:
Crossfire (both scaling and stuttering/frametimes)
Bugs while overclocking
Bugs when using a multi-monitor set-up
Bugs with low power modes
Bugs with the use of OpenCL.
Bugs with the use of displayport

As you can tell by the nature of those problems, they do not apply to someone not using that specific feature, and furthermore, as is usually the case with bugs: their occurence is usually not pervasive, and sometimes unpredictable.

The fact of the matter is: There are problems, but they are simply not applicable to everyone, either because the person in question use their hardware in such a way that they never engage in bug-ridden useage, or they simply don't notice the fact that something is not working as intended.

I myself as a single card user have noticed a lot of bugs, mainly overclocking, displayport, and multi-monitor related ones, but also some OpenCL weirdness. Furthermore I also got a friend who's running crossfire, and sometimes he's better off just running on 1 card, plain and simple, and while i probably wouldnt mind dealing with some issues that arise, the state of things at the moment is that in peoples heads, they generally just have to jump through too many hoops in order to enjoy something that is advertised as working.

For people like Earthdog I've got only one tip: Try to look beyond your own situation, as you're only one person compared to the numerous people in this thread alone posting something different(although this tip will probably be in vain, judging by his repetitive posts).

Edit: While there are definitely problems, I can also see things changing, and i've already felt some of those changes; like the improvements they've made to games like skyrim and borderlands. I actually have quite a bit of confidence that things will get better, confidence i had a whole lot less of half a year ago.
Posted on Reply
#21
Frick
Fishfaced Nincompoop
Oohhhh goooooooooddd not the frame stuff again. Can't you just wait for the drivers damnit? You must have emptied that subject twenty times over by now.
Posted on Reply
#22
cadaveca
My name is Dave
FrickOohhhh goooooooooddd not the frame stuff again. Can't you just wait for the drivers damnit? You must have emptied that subject twenty times over by now.
Sure, shall we move on to OpenCL issues, Flash acceleration, or game issues? Which other problem would you like us to talk about?


:p






I have months and months of problems, and no properly working WHQL Win8 drivers. Frame latency can eat it, Crossfire simply causes my system to crash at random intervals, and THAT is what's really bugging me about AMD's drivers.


Frame latency only explains why playing games makes me physically ill. Now that I know why, I got rid of Crossfire, and can game again!


:roll:


However, AMD still has other problems, too. MANY of them. And, no the pictured system is NOT overclocked.
Posted on Reply
#23
the54thvoid
Super Intoxicated Moderator
Back on topic....

If the thing is powered by 2 8 pins, that makes it a 375watt TDP card (well, power draw anyhow). If they have managed to do that, what the hell magic have they performed on the same silicon?

2 Fully fledged 1GHz chips on 2x3GB memory on less than 400 watts. That's pretty bloody impressive. Very, very impatient for full, warts and all reviews to come out in two days....
Posted on Reply
#24
cadaveca
My name is Dave
the54thvoidBack on topic....

If the thing is powered by 2 8 pins, that makes it a 375watt TDP card (well, power draw anyhow). If they have managed to do that, what the hell magic have the performed on the same silicon?

2 Fully fledged 1GHz chips on 2x3GB memory on less than 400 watts. That's pretty bloody impressive. Very, very impatient for full, warts and all reviews to come out in two days....
My full 3770k and 2x 7950 system pulls ~400W from the wall(three card pull less, because Crossfire is broken). I don't call 7990's power consumption all that amazing, really.
Posted on Reply
#25
EarthDog
cadavecaSays the dude who only runs a single 7970 or GTX690. :laugh:

Do keep in mind there are a huge number of Crossfire users here on TPU that DO agree it's an issue, and as big of an issue as related, so you'll have a hard time convincing those that use Crossfire there's no problems, and that's it's no big deal.

See, I can be a broken record, just like you.

:slap:

EarthDog: Not a big problem. No I don't run Crossfire.


Xzibit: Yeah, it's an issue....but look, NVidia has the same problems!


Also keep in mind, that W1zz does relate the same issues as being evident, in his reviews, before the PCPer stuff came out. Frankly, nobody really cares about PCPer, as all they showed is WHY we have issues in some instances, but doesn't provide a fix.



Why not just put your opinion on the subject in your signature, and not post? :roll: The less you post about it not being an issue, the less you'll find others posting that it is.
Ran it.. dont need it. 690 was essentially a gift so that is why I run it as opposed to any 1,2,3, or 4 7970's that I have had in the past. I would be running 7970's in CFx if the ones I had fit in my case. ;)

Frankly, nobody cares about people blowing this out of proportion. As I said, is it an issue, yes it is, for some people, not all, not on all titles at or most even. CFx is not "useless" as some fearmongers have stated.

Here is my exact take on this in fact. As a team, instead of fearmongering like was done at many sites/articles...: www.overclockers.com/nvidias-fcat-gpu-testing-pursuing/

@ Math - I have always noted this is a problem, just not as bad as A LOT of people have made it out to be is all. :)
Posted on Reply
Add your own comment
Dec 22nd, 2024 10:26 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts