With nan Gamescom normal taking spot successful Germany this week, AMD is utilizing Europe’s largest video crippled waste and acquisition show arsenic nan venue for their latest Radeon video paper announcements. This greeting nan institution is announcing their long-awaited mediate members of nan Radeon RX 7000 series, nan Radeon RX 7800 XT and Radeon RX 7700 XT. Aimed astatine nan 1440p gaming marketplace and based connected AMD’s caller Navi 32 GPU, nan caller cards are designed to slot successful to nan mediate of AMD’s merchandise stack, offering a group of potent RDNA 3 architecture video cards for gamers who don’t request nan bleeding-edge capacity (and wallet-bleeding costs) of nan Radeon RX 7900 bid cards.
With today’s announcement being conscionable that – an announcement – nan unit motorboat of these cards, as good arsenic AMD's caller Hypr-RX enabled drivers, will travel successful a mates of weeks, connected Wednesday, September 6th. That day besides not-so-coincidentally happens to beryllium nan merchandise day for Bethesda’s ARPG Starfield, for which AMD is nan game’s exclusive PC hardware partner, and which AMD will beryllium offering pinch nan caller Radeon cards arsenic portion of nan company’s latest crippled bundle. So for AMD, nan stars are aligning to make September 6th a large time for nan company’s GPU division.
But arsenic for us, we’re present to talk astir hardware, truthful let’s talk nan Radeon RX 7800 XT and 7700 XT, arsenic good arsenic nan Navi 32 GPU that underpins them.
|AMD Radeon RX 7000 Series Specification Comparison|
|AMD Radeon RX 7800 XT||AMD Radeon RX 7700 XT||AMD Radeon RX 6800 XT||AMD Radeon RX 6700 XT|
|Throughput (FP32)||32.6 TFLOPS||30 TFLOPS||18.6 TFLOPS||12.4 TFLOPS|
|Memory Clock||19.5 Gbps GDDR6||18 Gbps GDDR6||16 Gbps GDDR6||16 Gbps GDDR6|
|Memory Bus Width||256-bit||192-bit||256-bit||192-bit|
|Total Board Power||263W||245W||300W||230W|
|Manufacturing Process||GCD: TSMC 5nm
MCD: TSMC 6nm
|GCD: TSMC 5nm
MCD: TSMC 6nm
|TSMC 7nm||TSMC 7nm|
|Transistor Count||GCD: ?
MCD: 2B x4
MCD: 2B x4
|GPU||Navi 32||Navi 32||Navi 21||Navi 22|
For this procreation of video cards, AMD opted to eschew nan accustomed top-to-bottom GPU and video paper strategy successful favour of releasing nan apical video cards (Radeon RX 7900 XTX & 7900 XT), and pursuing that up pinch nan bottommost of nan stack (Radeon RX 7600). As a result, this is 1 of those uncommon launches wherever nan mediate parts, broadly encompassing nan “enthusiast” segment, are nan last conception to launch. The precise bid is mostly immaterial, but it intends we’re coming into nan RX 7800 XT motorboat already knowing really nan AMD cards supra it and beneath it execute – and frankincense wherever it needs to beryllium to champion fresh successful to nan merchandise stack.
And pinch rival NVIDIA having already launched their complete stack of video cards for this generation, AMD besides gets nan advantage of knowing wherever to champion position their cards against nan competition. In this case, AMD is positioning nan Radeon RX 7800 XT astir against nan GeForce RTX 4070, while its smaller sibling, nan Radeon RX 7700 XT, will spell up against nan GeForce RTX 4060 Ti. NVIDIA has already shown their manus for 1440p gaming, truthful pinch this motorboat it’s up to AMD to determine really to champion hit it.
Radeon RX 7800 XT & Radeon RX 7700 XT: By The Numbers
Leading things disconnected is nan Radeon RX 7800 XT, which AMD will beryllium positioning arsenic their premiere 1440p gaming card. The RX 7800 XT has a near-full Navi 32 configuration (more later) pinch 60 CUs integrated, affording 7680 ALUs (3840 watercourse processors). AMD is clocking this portion a small little aggressively than what we’ve seen pinch nan flagship RX 7900 XTX, giving america a highest boost timepiece of 2430MHz and an mean crippled timepiece of 2124MHz. On insubstantial this gives nan paper an mean FP32 compute throughput of 32.6 TFLOPS, though arsenic has been nan lawsuit pinch nan remainder of nan RDNA 3 family, nan architectural changes make throughput comparisons against anterior AMD cards tricky.
By nan numbers, nan RX 7800 XT offers astir 76% of nan compute throughput of nan adjacent paper up, nan RX 7900 XT. Notably, however, nan number of ROPs has been halved versus some RX 7900 bid cards, truthful nan existent rendered pixel throughput is person to 50%. Few games are wholly ROP-bound, truthful nan existent capacity spread betwixt this paper and nan 7900 XT shouldn’t beryllium astir arsenic big, but it underscores nan truth that we’re not looking astatine thing arsenic straightforward arsenic 75% of a 7900 XT.
With 4 progressive MCDs, nan RX 7800 XT is attached to a 256-bit GDDR6 representation bus. While we’ve seen nan 7900 bid cards clocked astatine 20Gbps, AMD has taken nan overseas prime to dial down nan clocks connected nan 7800 XT conscionable a bit, to nan noticeably fractional 19.5Gbps. This gives nan resulting paper a full of 624GB/second of representation bandwidth. That, successful turn, is further augmented pinch 64MB of Infinity Cache dispersed among nan MCDs, two-thirds of nan cache recovered connected AMD’s flagship cards.
The usage of a 256-bit representation autobus intends that AMD is easy capable to equip nan paper pinch 16GB of VRAM, which is going to beryllium a awesome trading constituent from nan institution successful competing pinch nan GeForce RTX 4070 and its 12GB of VRAM sitting connected a 192-bit representation bus. Consequently, AMD has nan separator successful some representation capacity and representation bandwidth, though arsenic we’ve seen successful nan past, that won’t needfully construe into a capacity advantage for AMD.
Meanwhile, pinch nan wide creation of Navi 32 favoring costs complete powerfulness efficiency, nan powerfulness depletion of nan 7800 XT is going to tally astatine a comparatively precocious 263 Watts, 37W little than nan 7900 XT. AMD is not going to beryllium capable to triumph nan powerfulness ratio title pinch NVIDIA present (and arguably they ne'er moreover intended to try) truthful alternatively nan RX 7800 XT is configured to maximize its performance. Note that these are AMD’s baseline specifications, and mill overclocked AIB partner cards tin spell higher.
Making generational comparisons present is simply a spot tricky, some owed to nan architectural changes from RDNA 2 to RDNA 3, arsenic good arsenic immoderate quirks successful AMD’s ain paper positioning for this generation. While nan RX 7800 XT is AMD’s 3rd/4th tier paper wide (behind nan 3 7900 bid cards), its nonstop predecessor, nan RX 6800 XT, was AMD’s second-tier card, down only nan flagship 6900 XT. Long communicative short, nan 6800 XT/7800 XT comparison isn’t particularly flattering for AMD – nan number of CUs has gone down by a spot – though truthful has powerfulness depletion and price. The much apt comparison is arguably nan RX 6800 vanilla, but I digress.
In immoderate case, nan large generational upgrade push from AMD is going to beryllium pitting nan RX 7800 XT against cards that are now coming up connected 4 years old, specified arsenic nan Radeon RX 5700 XT and GeForce RTX 2070. The RX 7800 XT should astir double nan capacity of those cards, making them a much palatable upgrade for strategy owners whose existent 1440p cards are nary longer keeping up.
Sapphire Pulse RX 7700 XT
Finally, opinionated successful statement down nan RX 7800 XT, we person nan RX 7700 XT. This is simply a cut-down Navi 32-based paper that, contempt nan little bid number, is besides aimed astatine 1440p gaming.
Hardware-wise, we’re looking astatine 54 CUs, pinch a maximum boost timepiece of 2544MHz and an mean crippled timepiece of 2171MHz. As a result, nan compute capacity of nan RX 7700 XT is, connected paper, rather adjacent to its bigger brother, pinch nan smaller paper slated to present 92% of nan RX 7800 XT’s compute performance. And, notably, location is nary trim to nan number of ROPs here, truthful nan wide pixel pushing powerfulness is technically ever truthful somewhat successful nan RX 7700 XT’s favor.
Instead, what will beryllium driving nan bulk of nan capacity differences betwixt nan 2 cards will beryllium representation bandwidth. The RX 7700 XT only features 3 progressive MCDs (like nan 7900 XT, nan inactive MCD is kept arsenic a dummy spacer), which intends it’s losing a 4th of its representation bus, arsenic good arsenic a 4th of its Infinity Cache. Coupled pinch that, AMD has trim nan representation clocks a spot more, bringing them down to 18Gbps/pin. The nett consequence is that nan RX 7700 XT offers 69% of nan RX 7800 XT’s representation bandwidth, and 75% (48MB) of nan Infinity Cache capacity. All of which intends that it will beryllium conscionable really representation and cache bandwidth-bound a fixed crippled is that determines really adjacent nan RX 7700 XT tin get to nan RX 7800 XT.
The smaller representation autobus besides intends that nan paper comes pinch little VRAM overall. In this case, we’re looking astatine 12GB of GDDR6. With AMD positioning this paper against NVIDIA’s 128-bit GeForce RTX 4060 Ti, that intends they person a representation capacity and bandwidth against nan baseline 8GB model, though NVIDIA’s partners are besides offering a 16GB exemplary that leaves nan RX 7700 XT smack dab successful nan middle.
Otherwise, pinch clockspeeds moving moreover higher than nan RX 7800 XT, nan RX 7700 XT picks up very small successful powerfulness savings. AMD has nan card’s charismatic TBP astatine 245W, conscionable 18W little than nan RX 7800 XT. Assuming what real-world powerfulness depletion matches this fig (which is not ever nan case), past it intends that nan RX 7700 XT is going to beryllium little powerfulness businesslike of nan 2 cards overall. And, for that matter, acold little powerfulness businesslike than NVIDIA’s 160W RTX 4060 Ti.
The RX 7700 XT will beryllium replacing nan RX 6700 XT wrong AMD’s merchandise stack, which astatine nan clip of its introduction, was nan apical Navi 22 card. While we don’t person immoderate comparative capacity figures, nan RX 7700 XT wide enjoys a 35% bump successful CU count, albeit astatine a ~10% regression successful mean clockspeeds. All nan while powerfulness depletion has trickled up a spot from 230W. So it should make for a humble generational jump complete its nonstop predecessor.
Navi 32: A Second, Smaller Chiplet GPU
Driving some cards is AMD’s caller RDNA 3 architecture-based Navi 32 GPU. This is nan 2nd user chiplet GPU from nan company, and is fundamentally a smaller return connected Navi 31 that powers nan RX 7900 series. In this lawsuit “smaller take” intends that AMD is utilizing a caller Graphics Compute Die (GCD) pinch less functional blocks and transistors, trading disconnected capacity for little powerfulness depletion and manufacturing costs. Meanwhile nan GCD is being paired pinch 4 of AMD’s Memory Cache Dies (MCD), which are being reused from Navi 31.
As pinch Navi 31, nan intricate GCD is fabbed connected TSMC’s 5nm process, building nan scalable and capacity captious parts of nan GPU connected a leading-edge TSMC process. Meanwhile nan MCDs proceed to beryllium produced connected TMSC’s 6nm process, lodging parts of nan GPU that don’t standard good pinch smaller nodes, specified arsenic nan Infinity Cache and GDDR6 representation controllers. So Navi 32 is simply a operation of caller and old(er) process nodes pinch regards to its fabrication, while it’s connected nan cutting separator of packaging.
Broadly speaking, nan full Navi 3x GPU lineup is designed to beryllium cost-efficient from a manufacturing standpoint, and Navi 32 is nary different. AMD is only producing nan captious bits connected a starring separator 5nm node – and moreover then, not utilizing TSMC’s bleeding-edge 4nm version – while producing nan remainder connected a cheaper 6nm node. Without going excessively acold down nan rabbit spread for this announcement, this has capacity implications that go a facet successful really AMD configures and positions their cards. But astatine a precocious level, nan takeaway is that nan novelty of a chiplet-based GPU aside, Navi 32 follows a notably different creation strategy than past enthusiast-class AMD GPUs, aliases for that matter rival NVIDIA’s monolithic designs. As we’ve seen pinch nan RX 7900 series, this is going to beryllium nan astir noticeable successful powerfulness consumption, and is apt nan biggest facet successful why some Navi 32 cards person TBP specs good supra their NVIDIA competition.
For now, we’re still waiting connected AMD to corroborate immoderate of nan specifications of Navi 32, successful peculiar nan dice size of nan GCD and its transistor count. Based connected AMD’s off-angle dice photo, we’re presently estimating a unsmooth 211mm2 based connected nan known size of nan MCDs (37.5mm2), but that’s a unsmooth conjecture astatine best. That would put it astatine astir two-thirds nan size of nan Navi 31 GCD.
At this constituent we’re besides awaiting confirmation connected nan full number of CUs disposable wrong nan GPU. While nan RX 7800 XT is nan apical tier part, AMD has not confirmed whether Navi 32 itself besides only has 60 CUs, aliases if it has more. Given that Navi 31 was 96 CUs and Navi 33 was 32 CUs, it’s apt that Navi 32 is simply a 64 CU design, but alien things person happened.
Otherwise, Navi 32 is RDNA 3 done and through. This intends it incorporates nan aforesaid architectural changes we’ve seen successful nan remainder of nan RX 7000 bid lineup, specified arsenic AV1 encode/decode support, improved ray tracing performance, AMD’s caller DisplayPort UHBR13.5-capable Radiance show controller, and nan 2nd slope of ALUs for dual-issuing suitable instructions and improving AI performance.
Reference Cards, Partner Cards, & Product Positioning
As noted earlier, some cards will beryllium launching successful a mates of weeks, connected September 6th. With regards to reference cards, AMD is fundamentally splitting nan quality – they will beryllium trading a reference Radeon RX 7800 XT done their website, but not a reference Radeon RX 7700 XT. Instead, AMD’s committee partners will beryllium swooping successful to support nan motorboat of some cards, pinch AIB designs slated to beryllium disposable for some cards connected nan motorboat day.
Taking a little look astatine AMD’s reference card, it’s a reasonably sizable twin-fan unfastened aerial design, copying nan creation aesthetics from AMD’s reference RX 7900 bid cards. At 2 fans it’s smaller overall, but not immensely so. Notably, this is simply a 2.5 slot card, which not each of nan pictures show very well, truthful it will request immoderate other room to breathe.
Powering nan paper is simply a brace of 8-pin PCIe powerfulness sockets. Strictly speaking, this is technically overkill, since a 6+8 configuration would still beryllium bully for 300W. But much and much we’re seeing 6-pin sockets spell distant entirely, truthful it’s not wholly astonishing to spot AMD spell for 8+8. As noted earlier, nan TDP of nan reference paper is 265W, nevertheless AIB partners tin (and will) timepiece immoderate of their cards higher pinch higher TDPs to match.
Meanwhile, AMD’s larboard configuration has taken a spot of a departure from nan RX 7900 reference designs. While still offering 4 ports, AMD has thrown retired nan USB-C larboard successful favour of different full-size DisplayPort. The consequence is simply a trio of DisplayPorts that support DP 2.1 astatine information rates of up to UHBR13.5, arsenic good arsenic a sole HDMI larboard supporting HDMI 2.1.
Sapphire's Nitro RX 7800 XT & RX 7700 XT
As for nan partner cards, each of nan accustomed suspects person RX 7800/7700 designs successful nan works, including ASUS, GIGABYTE, Sapphire, PowerColor, and more. AMD sent complete a mates of twelve photos showing disconnected nan different designs, a batch of which are triple-fan cards that look for illustration their existing RX 7900 designs.
Unfortunately, astir of these photos are head-on shots, truthful we can’t show whether galore (if any) are existent dual-slot designs (versus nan 2.5 slot creation of nan AMD reference card). Due to nan comparatively precocious TBPs, nan RX 7800/7700 bid puts AMD and its partners successful a spot of a reliable spot erstwhile it comes to producing SFF-sized cards. We’ll person much accusation erstwhile nan committee partners station their charismatic merchandise pages, but astatine first blush, it doesn’t look for illustration we’ll spot overmuch successful nan measurement of mini cards – astatine slightest initially.
In position of competition, AMD is mounting up nan RX 7800 XT and RX 7700 XT astir against nan GeForce RTX 4070 and GeForce RTX 4060 Ti respectively – pinch AMD believing they’re successful nan lead, of course. It’s a spot of an overseas match-up since nan RTX 4060 Ti is not chiefly aimed astatine 1440p gaming nan measurement nan RX 7700 XT is, but this is hardly nan first clip AMD and NVIDIA’s cards don’t perfectly statement up.
Just really good they compare, however, remains to beryllium seen. Not conscionable for independent capacity benchmarks, but because AMD is not disclosing nan value of either paper to nan property up of time. So whether AMD is going for iso-pricing aliases if they intend to undercut NVIDIA successful immoderate manner remains to beryllium seen.
Update: Now that nan charismatic announcement has gone out, AMD has disclosed nan value of nan cards. The Radeon RX 7800 XT will commencement astatine $499, while nan Radeon RX 7700 XT will commencement astatine $449. The remainder of nan original article follows arsenic below.
But, successful a motion that AMD apt isn’t going to beryllium too fierce connected pricing, nan institution is extending their existent Starfield promotional crippled bundle to screen nan caller Radeon cards; truthful some nan RX 7800 XT and RX 7700 XT will beryllium coming pinch a crippled bundle correct disconnected nan bat. Starfield is AMD’s large crippled business for nan year, truthful nan institution is keen to beforehand what they spot arsenic 1 of nan astir important games of nan year. At nan aforesaid time, crippled bundles are traditionally utilized to prop up video paper prices successful lieu of value cuts by adding worth to nan wide package.
Ultimately, conscionable really successful nan Radeon RX 7800 XT and RX 7700 XT are will beryllium wholly successful AMD’s hands. RDNA 3 is simply a coagulated architecture, and acknowledgment to nan company’s cost-optimized chiplet-based creation strategy, they shouldn’t person immoderate problem hitting value points that make some cards competitive. The only existent mobility is conscionable really galore cards AMD wants to waste – and frankincense what value points they target.
Wrapping things up, nan Radeon RX 7800 XT and RX 7700 XT will beryllium launching connected September 6th. AMD will beryllium carrying their reference RX 7800 XT straight connected their website, meantime Newegg, Amazon, and different retailers and etailers will beryllium carrying nan various AIB cards that nan partners person put together.